ResNet152网络复现（Caffe）

一、准备数据集

1)  下载数据集

Imagnet网站上下载了三类图片，分别是big cat、dog、fish，其中训练集的图片数一共是4149，测试集的图片数是1003，训练集和测试集的图片数比例4:1，将训练集的图片保存在train文件夹下，测试集图片保存在val文件夹下.

train、val文件夹下面均有bigcat、dog、fish三个文件夹，分别存放着对应类别的图片.

 

2) 利用python代码，生成train.txt、val.txt

train.txt、val.txt分别存储着训练集和测试集图片的文件名及其类别标签（注：bigcat:0、 dog:1、fish:2），格式如下：

n02084071_9865.JPEG 1
n02512053_3388.JPEG 2
n02512053_6294.JPEG 2
n02512053_2413.JPEG 2
n02084071_5655.JPEG 1
n02127808_9965.JPEG 0
n02127808_8206.JPEG 0
n02127808_4887.JPEG 0
n02512053_1952.JPEG 2

     Python代码如下(可以先在windows环境下利用以下的代码生成txt)：

import os
import random


trainPath = 'F:\Resnet152\train\' 
valPath = 'F:\Resnet152\val\' 

train = {}
val = {}

    
# add
for name in os.listdir(trainPath + "bigcat\"):
    train[name] = 0
     
# add 
for name in os.listdir(trainPath + "dog\"):
    train[name] = 1
    
# add 
for name in os.listdir(trainPath + "fish\"):
    train[name] = 2


# add
for name in os.listdir(valPath + "bigcat\"):
    val[name] = 0
     
# add
for name in os.listdir(valPath + "dog\"):
    val[name] = 1
    

# add
for name in os.listdir(valPath + "fish\"):
    val[name] = 2
    
    
ftrain = open("F:\Resnet152\train\train.txt", 'w')
fval = open("F:\Resnet152\val\val.txt", 'w')

trainName = []
valName = []
for (item) in train:
    trainName.append(item)

for item in val:
    valName.append(item)

random.shuffle(trainName)
random.shuffle(valName)

for name in trainName:
    label = train[name]
    ftrain.write(name + " " + str(label) + "
")

for name in valName:
    label = val[name]
    fval.write(name + " " + str(label) + "
")

ftrain.close() 
fval.close()   

 

3) 生成数据集

编写lmdb.sh脚本文件，利用train.txt、val.txt生成train_lmdb、val_lmdb

lmdb.sh脚本代码如下

#!/usr/bin/env sh
# Create the face_48 lmdb inputs
# N.B. set the path to the face_48 train + val data dirs

EXAMPLE=/home/wy/ResNet152    #lmdb生成后存放目录
DATA=/home/wy/ResNet152          #train.txt、val.txt存放目录
TOOLS=/home/wy/caffe/build/tools #caffe安装目录

TRAIN_DATA_ROOT=/home/wy/ResNet152/train/
VAL_DATA_ROOT=/home/wy/ResNet152/val/

# Set RESIZE=true to resize the images to 224 x 224. Leave as false if images have
# already been resized using another tool.
RESIZE=true
if $RESIZE; then
  RESIZE_HEIGHT=224   #resize图片大小
  RESIZE_WIDTH=224
else
  RESIZE_HEIGHT=0
  RESIZE_WIDTH=0
fi

if [ ! -d "$TRAIN_DATA_ROOT" ]; then
  echo "Error: TRAIN_DATA_ROOT is not a path to a directory: $TRAIN_DATA_ROOT"
  echo "Set the TRAIN_DATA_ROOT variable in create_face_48.sh to the path" 
       "where the face_48 training data is stored."
  exit 1
fi

if [ ! -d "$VAL_DATA_ROOT" ]; then
  echo "Error: VAL_DATA_ROOT is not a path to a directory: $VAL_DATA_ROOT"
  echo "Set the VAL_DATA_ROOT variable in create_face_48.sh to the path" 
       "where the face_48 validation data is stored."
  exit 1
fi

echo "Creating train lmdb..."

GLOG_logtostderr=1 $TOOLS/convert_imageset 
    --resize_height=$RESIZE_HEIGHT 
    --resize_width=$RESIZE_WIDTH 
    --shuffle 
    $TRAIN_DATA_ROOT 
    $DATA/train.txt 
    $EXAMPLE/train_lmdb

echo "Creating val lmdb..."

GLOG_logtostderr=1 $TOOLS/convert_imageset 
    --resize_height=$RESIZE_HEIGHT 
    --resize_width=$RESIZE_WIDTH 
    --shuffle 
    $VAL_DATA_ROOT 
    $DATA/val.txt 
    $EXAMPLE/val_lmdb

echo "Done."
Status API Training Shop Blog About

注：上述代码中有RESIZE_HEIGHT=224、RESIZE_WIDTH=224，因为使用的是ResNet网络，

需要统一图片大小，如果使用的网络对图片大小无要求，或者已经是224*224大小的图片，可不进行resize操作

linux系统终端中执行sh lmdb.sh即可生成lmdb文件，完成数据集的生成工作

执行结束后，会在ResNet152目录下，生成两个文件夹,分别是train_lmdb、val_lmdb

 

4) 计算图片平均值

训练网络图片时，需要对图片做减均值处理，先将图片平均值保存到文件中，在训练网络中直接引用，先介绍如何生成图片平均值文件，后面引用的时候会特别注明，命令如下：

/home/wy/caffe/build/tools/compute_image_mean /home/wy/ResNet152/train_lmdb /home/wy/ResNet152/mean.binaryproto

      利用Caffe安装目录下面的compute_image_mean.bin进行图片平均值的计算，输入的是train_lmdb，输出的为mean.binaryproto

      

     

二、修改网络配置文件

修改配置之前，先列出需要的文件：train_val.prototxt、deploy.prototxt、solver.prototxt（三个文件的具体内容会在博客最后给出）

1)  train_val.prototxt

文件名train_val包括train和val，可以先这样理解每进行500次训练的时候，会进行一次验证，方便输出供训练者观察情况，既然这样，就需要在这个文件里面指定train数据集和val数据集

      

       

       关于train_val.prototxt中其余部分的代码均是ResNet152的网络结构代码，无需修改

 

2)  deploy.prototxt

deploy.prototxt主要是当模型训练出来以后，利用模型对用户提交的一张图片进行分类应用的时候使用的网络文件，只需要进行一次前向传播计算输入图片所属类别的概率值，所以，此文件里面没有损失函数层的定义

 

3) solver.prototxt

文件指定训练的相关规则和参数，具体内容介绍见下图中的标注

 

三、训练网络

网络文件编辑完后，即可开始训练网络，可直接在linux终端输入以下命令

/home/wy/caffe/build/tools/caffe train --solver=/home/wy/ResNet152/solver.prototxt

前者是caffe安装目录路径，后者是solver.prototxt存放路径，回车后即可开始训练；也可创建train.sh脚本文件，文件中保存上述指令

此时，在linux终端(导航到train.sh存放的路径)，输入sh  train.sh即可

训练过程中的一张图

 

四、测试图片

用训练出的model进行图片的预测

/home/wy/caffe/build/examples/cpp_classification/classification.bin /home/wy/ResNet152/deploy.prototxt /home/wy/ResNet152/model/solver_iter_60000.caffemodel /home/wy/ResNet152/mean.binaryproto /home/wy/ResNet152/class_name.txt /home/wy/ResNet152/testImages/ISIC_0000001.jpg

在linux下输入上述命令，即可对图片’ISIC_0000001.jpg’进行预测

其中，class_name.txt文件中内容（和trainx.txt中标签要对应）：

0 bigcat  
1 dog  
2 fish 

测试图片：

看下测试结果（测试的时候换了台电脑,所以发现目录不太一样）：

所以，属于bigcat的概率为0.9999，属于fish的概率为0.0001，属于dog的概率为0

 

train_val.prototxt

name: "ResNet-152"
layer {
    name: "data"
    type: "Data"
    top: "data"
    top: "label"
    include {
        phase: TRAIN
    }
    transform_param {
        mirror: true
        crop_size: 224
        mean_file: "/home/wy/ResNet152/mean.binaryproto"
    }
    data_param {
        source: "/home/wy/ResNet152/train_lmdb"
        batch_size: 1
        backend: LMDB
    }
}
layer {
    name: "data"
    type: "Data"
    top: "data"
    top: "label"
    include {
        phase: TEST
    }
    transform_param {
        mirror: false
        crop_size: 224
        mean_file:"/usr/develop/repertory/ResNet152/mean.binaryproto"
    }
    data_param {
        source: "/usr/develop/repertory/ResNet152/val_lmdb"
        batch_size: 1
        backend: LMDB
    }
}

layer {
    bottom: "data"
    top: "conv1"
    name: "conv1"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 7
        pad: 3
        stride: 2
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "conv1"
    top: "conv1"
    name: "bn_conv1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "conv1"
    top: "conv1"
    name: "scale_conv1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "conv1"
    top: "conv1"
    name: "conv1_relu"
    type: "ReLU"
}

layer {
    bottom: "conv1"
    top: "pool1"
    name: "pool1"
    type: "Pooling"
    pooling_param {
        kernel_size: 3
        stride: 2
        pool: MAX
    }
}

layer {
    bottom: "pool1"
    top: "res2a_branch1"
    name: "res2a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2a_branch1"
    top: "res2a_branch1"
    name: "bn2a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2a_branch1"
    top: "res2a_branch1"
    name: "scale2a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "pool1"
    top: "res2a_branch2a"
    name: "res2a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2a_branch2a"
    top: "res2a_branch2a"
    name: "bn2a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2a_branch2a"
    top: "res2a_branch2a"
    name: "scale2a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2a_branch2a"
    top: "res2a_branch2a"
    name: "res2a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2a_branch2a"
    top: "res2a_branch2b"
    name: "res2a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2a_branch2b"
    top: "res2a_branch2b"
    name: "bn2a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2a_branch2b"
    top: "res2a_branch2b"
    name: "scale2a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2a_branch2b"
    top: "res2a_branch2b"
    name: "res2a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2a_branch2b"
    top: "res2a_branch2c"
    name: "res2a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2a_branch2c"
    top: "res2a_branch2c"
    name: "bn2a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2a_branch2c"
    top: "res2a_branch2c"
    name: "scale2a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2a_branch1"
    bottom: "res2a_branch2c"
    top: "res2a"
    name: "res2a"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res2a"
    top: "res2a"
    name: "res2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2a"
    top: "res2b_branch2a"
    name: "res2b_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2b_branch2a"
    top: "res2b_branch2a"
    name: "bn2b_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2b_branch2a"
    top: "res2b_branch2a"
    name: "scale2b_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2b_branch2a"
    top: "res2b_branch2a"
    name: "res2b_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2b_branch2a"
    top: "res2b_branch2b"
    name: "res2b_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2b_branch2b"
    top: "res2b_branch2b"
    name: "bn2b_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2b_branch2b"
    top: "res2b_branch2b"
    name: "scale2b_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2b_branch2b"
    top: "res2b_branch2b"
    name: "res2b_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2b_branch2b"
    top: "res2b_branch2c"
    name: "res2b_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2b_branch2c"
    top: "res2b_branch2c"
    name: "bn2b_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2b_branch2c"
    top: "res2b_branch2c"
    name: "scale2b_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2a"
    bottom: "res2b_branch2c"
    top: "res2b"
    name: "res2b"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res2b"
    top: "res2b"
    name: "res2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2b"
    top: "res2c_branch2a"
    name: "res2c_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2c_branch2a"
    top: "res2c_branch2a"
    name: "bn2c_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2c_branch2a"
    top: "res2c_branch2a"
    name: "scale2c_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2c_branch2a"
    top: "res2c_branch2a"
    name: "res2c_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2c_branch2a"
    top: "res2c_branch2b"
    name: "res2c_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2c_branch2b"
    top: "res2c_branch2b"
    name: "bn2c_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2c_branch2b"
    top: "res2c_branch2b"
    name: "scale2c_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2c_branch2b"
    top: "res2c_branch2b"
    name: "res2c_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2c_branch2b"
    top: "res2c_branch2c"
    name: "res2c_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res2c_branch2c"
    top: "res2c_branch2c"
    name: "bn2c_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res2c_branch2c"
    top: "res2c_branch2c"
    name: "scale2c_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2b"
    bottom: "res2c_branch2c"
    top: "res2c"
    name: "res2c"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res2c"
    top: "res2c"
    name: "res2c_relu"
    type: "ReLU"
}

layer {
    bottom: "res2c"
    top: "res3a_branch1"
    name: "res3a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 2
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3a_branch1"
    top: "res3a_branch1"
    name: "bn3a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3a_branch1"
    top: "res3a_branch1"
    name: "scale3a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2c"
    top: "res3a_branch2a"
    name: "res3a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 2
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3a_branch2a"
    top: "res3a_branch2a"
    name: "bn3a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3a_branch2a"
    top: "res3a_branch2a"
    name: "scale3a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3a_branch2a"
    top: "res3a_branch2a"
    name: "res3a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3a_branch2a"
    top: "res3a_branch2b"
    name: "res3a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3a_branch2b"
    top: "res3a_branch2b"
    name: "bn3a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3a_branch2b"
    top: "res3a_branch2b"
    name: "scale3a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3a_branch2b"
    top: "res3a_branch2b"
    name: "res3a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3a_branch2b"
    top: "res3a_branch2c"
    name: "res3a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3a_branch2c"
    top: "res3a_branch2c"
    name: "bn3a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3a_branch2c"
    top: "res3a_branch2c"
    name: "scale3a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3a_branch1"
    bottom: "res3a_branch2c"
    top: "res3a"
    name: "res3a"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3a"
    top: "res3a"
    name: "res3a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3a"
    top: "res3b1_branch2a"
    name: "res3b1_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b1_branch2a"
    top: "res3b1_branch2a"
    name: "bn3b1_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b1_branch2a"
    top: "res3b1_branch2a"
    name: "scale3b1_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b1_branch2a"
    top: "res3b1_branch2a"
    name: "res3b1_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b1_branch2a"
    top: "res3b1_branch2b"
    name: "res3b1_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b1_branch2b"
    top: "res3b1_branch2b"
    name: "bn3b1_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b1_branch2b"
    top: "res3b1_branch2b"
    name: "scale3b1_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b1_branch2b"
    top: "res3b1_branch2b"
    name: "res3b1_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b1_branch2b"
    top: "res3b1_branch2c"
    name: "res3b1_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b1_branch2c"
    top: "res3b1_branch2c"
    name: "bn3b1_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b1_branch2c"
    top: "res3b1_branch2c"
    name: "scale3b1_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3a"
    bottom: "res3b1_branch2c"
    top: "res3b1"
    name: "res3b1"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3b1"
    top: "res3b1"
    name: "res3b1_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b1"
    top: "res3b2_branch2a"
    name: "res3b2_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b2_branch2a"
    top: "res3b2_branch2a"
    name: "bn3b2_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b2_branch2a"
    top: "res3b2_branch2a"
    name: "scale3b2_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b2_branch2a"
    top: "res3b2_branch2a"
    name: "res3b2_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b2_branch2a"
    top: "res3b2_branch2b"
    name: "res3b2_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b2_branch2b"
    top: "res3b2_branch2b"
    name: "bn3b2_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b2_branch2b"
    top: "res3b2_branch2b"
    name: "scale3b2_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b2_branch2b"
    top: "res3b2_branch2b"
    name: "res3b2_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b2_branch2b"
    top: "res3b2_branch2c"
    name: "res3b2_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b2_branch2c"
    top: "res3b2_branch2c"
    name: "bn3b2_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b2_branch2c"
    top: "res3b2_branch2c"
    name: "scale3b2_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b1"
    bottom: "res3b2_branch2c"
    top: "res3b2"
    name: "res3b2"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3b2"
    top: "res3b2"
    name: "res3b2_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b2"
    top: "res3b3_branch2a"
    name: "res3b3_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b3_branch2a"
    top: "res3b3_branch2a"
    name: "bn3b3_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b3_branch2a"
    top: "res3b3_branch2a"
    name: "scale3b3_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b3_branch2a"
    top: "res3b3_branch2a"
    name: "res3b3_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b3_branch2a"
    top: "res3b3_branch2b"
    name: "res3b3_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b3_branch2b"
    top: "res3b3_branch2b"
    name: "bn3b3_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b3_branch2b"
    top: "res3b3_branch2b"
    name: "scale3b3_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b3_branch2b"
    top: "res3b3_branch2b"
    name: "res3b3_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b3_branch2b"
    top: "res3b3_branch2c"
    name: "res3b3_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b3_branch2c"
    top: "res3b3_branch2c"
    name: "bn3b3_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b3_branch2c"
    top: "res3b3_branch2c"
    name: "scale3b3_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b2"
    bottom: "res3b3_branch2c"
    top: "res3b3"
    name: "res3b3"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3b3"
    top: "res3b3"
    name: "res3b3_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b3"
    top: "res3b4_branch2a"
    name: "res3b4_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b4_branch2a"
    top: "res3b4_branch2a"
    name: "bn3b4_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b4_branch2a"
    top: "res3b4_branch2a"
    name: "scale3b4_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b4_branch2a"
    top: "res3b4_branch2a"
    name: "res3b4_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b4_branch2a"
    top: "res3b4_branch2b"
    name: "res3b4_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b4_branch2b"
    top: "res3b4_branch2b"
    name: "bn3b4_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b4_branch2b"
    top: "res3b4_branch2b"
    name: "scale3b4_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b4_branch2b"
    top: "res3b4_branch2b"
    name: "res3b4_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b4_branch2b"
    top: "res3b4_branch2c"
    name: "res3b4_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b4_branch2c"
    top: "res3b4_branch2c"
    name: "bn3b4_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b4_branch2c"
    top: "res3b4_branch2c"
    name: "scale3b4_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b3"
    bottom: "res3b4_branch2c"
    top: "res3b4"
    name: "res3b4"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3b4"
    top: "res3b4"
    name: "res3b4_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b4"
    top: "res3b5_branch2a"
    name: "res3b5_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b5_branch2a"
    top: "res3b5_branch2a"
    name: "bn3b5_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b5_branch2a"
    top: "res3b5_branch2a"
    name: "scale3b5_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b5_branch2a"
    top: "res3b5_branch2a"
    name: "res3b5_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b5_branch2a"
    top: "res3b5_branch2b"
    name: "res3b5_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b5_branch2b"
    top: "res3b5_branch2b"
    name: "bn3b5_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b5_branch2b"
    top: "res3b5_branch2b"
    name: "scale3b5_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b5_branch2b"
    top: "res3b5_branch2b"
    name: "res3b5_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b5_branch2b"
    top: "res3b5_branch2c"
    name: "res3b5_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b5_branch2c"
    top: "res3b5_branch2c"
    name: "bn3b5_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b5_branch2c"
    top: "res3b5_branch2c"
    name: "scale3b5_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b4"
    bottom: "res3b5_branch2c"
    top: "res3b5"
    name: "res3b5"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3b5"
    top: "res3b5"
    name: "res3b5_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b5"
    top: "res3b6_branch2a"
    name: "res3b6_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b6_branch2a"
    top: "res3b6_branch2a"
    name: "bn3b6_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b6_branch2a"
    top: "res3b6_branch2a"
    name: "scale3b6_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b6_branch2a"
    top: "res3b6_branch2a"
    name: "res3b6_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b6_branch2a"
    top: "res3b6_branch2b"
    name: "res3b6_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b6_branch2b"
    top: "res3b6_branch2b"
    name: "bn3b6_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b6_branch2b"
    top: "res3b6_branch2b"
    name: "scale3b6_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b6_branch2b"
    top: "res3b6_branch2b"
    name: "res3b6_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b6_branch2b"
    top: "res3b6_branch2c"
    name: "res3b6_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b6_branch2c"
    top: "res3b6_branch2c"
    name: "bn3b6_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b6_branch2c"
    top: "res3b6_branch2c"
    name: "scale3b6_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b5"
    bottom: "res3b6_branch2c"
    top: "res3b6"
    name: "res3b6"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3b6"
    top: "res3b6"
    name: "res3b6_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b6"
    top: "res3b7_branch2a"
    name: "res3b7_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b7_branch2a"
    top: "res3b7_branch2a"
    name: "bn3b7_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b7_branch2a"
    top: "res3b7_branch2a"
    name: "scale3b7_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b7_branch2a"
    top: "res3b7_branch2a"
    name: "res3b7_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b7_branch2a"
    top: "res3b7_branch2b"
    name: "res3b7_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b7_branch2b"
    top: "res3b7_branch2b"
    name: "bn3b7_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b7_branch2b"
    top: "res3b7_branch2b"
    name: "scale3b7_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b7_branch2b"
    top: "res3b7_branch2b"
    name: "res3b7_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b7_branch2b"
    top: "res3b7_branch2c"
    name: "res3b7_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res3b7_branch2c"
    top: "res3b7_branch2c"
    name: "bn3b7_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res3b7_branch2c"
    top: "res3b7_branch2c"
    name: "scale3b7_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b6"
    bottom: "res3b7_branch2c"
    top: "res3b7"
    name: "res3b7"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res3b7"
    top: "res3b7"
    name: "res3b7_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b7"
    top: "res4a_branch1"
    name: "res4a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 2
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4a_branch1"
    top: "res4a_branch1"
    name: "bn4a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4a_branch1"
    top: "res4a_branch1"
    name: "scale4a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b7"
    top: "res4a_branch2a"
    name: "res4a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 2
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4a_branch2a"
    top: "res4a_branch2a"
    name: "bn4a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4a_branch2a"
    top: "res4a_branch2a"
    name: "scale4a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4a_branch2a"
    top: "res4a_branch2a"
    name: "res4a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4a_branch2a"
    top: "res4a_branch2b"
    name: "res4a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4a_branch2b"
    top: "res4a_branch2b"
    name: "bn4a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4a_branch2b"
    top: "res4a_branch2b"
    name: "scale4a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4a_branch2b"
    top: "res4a_branch2b"
    name: "res4a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4a_branch2b"
    top: "res4a_branch2c"
    name: "res4a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4a_branch2c"
    top: "res4a_branch2c"
    name: "bn4a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4a_branch2c"
    top: "res4a_branch2c"
    name: "scale4a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4a_branch1"
    bottom: "res4a_branch2c"
    top: "res4a"
    name: "res4a"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4a"
    top: "res4a"
    name: "res4a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4a"
    top: "res4b1_branch2a"
    name: "res4b1_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b1_branch2a"
    top: "res4b1_branch2a"
    name: "bn4b1_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b1_branch2a"
    top: "res4b1_branch2a"
    name: "scale4b1_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b1_branch2a"
    top: "res4b1_branch2a"
    name: "res4b1_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b1_branch2a"
    top: "res4b1_branch2b"
    name: "res4b1_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b1_branch2b"
    top: "res4b1_branch2b"
    name: "bn4b1_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b1_branch2b"
    top: "res4b1_branch2b"
    name: "scale4b1_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b1_branch2b"
    top: "res4b1_branch2b"
    name: "res4b1_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b1_branch2b"
    top: "res4b1_branch2c"
    name: "res4b1_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b1_branch2c"
    top: "res4b1_branch2c"
    name: "bn4b1_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b1_branch2c"
    top: "res4b1_branch2c"
    name: "scale4b1_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4a"
    bottom: "res4b1_branch2c"
    top: "res4b1"
    name: "res4b1"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b1"
    top: "res4b1"
    name: "res4b1_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b1"
    top: "res4b2_branch2a"
    name: "res4b2_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b2_branch2a"
    top: "res4b2_branch2a"
    name: "bn4b2_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b2_branch2a"
    top: "res4b2_branch2a"
    name: "scale4b2_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b2_branch2a"
    top: "res4b2_branch2a"
    name: "res4b2_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b2_branch2a"
    top: "res4b2_branch2b"
    name: "res4b2_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b2_branch2b"
    top: "res4b2_branch2b"
    name: "bn4b2_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b2_branch2b"
    top: "res4b2_branch2b"
    name: "scale4b2_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b2_branch2b"
    top: "res4b2_branch2b"
    name: "res4b2_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b2_branch2b"
    top: "res4b2_branch2c"
    name: "res4b2_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b2_branch2c"
    top: "res4b2_branch2c"
    name: "bn4b2_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b2_branch2c"
    top: "res4b2_branch2c"
    name: "scale4b2_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b1"
    bottom: "res4b2_branch2c"
    top: "res4b2"
    name: "res4b2"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b2"
    top: "res4b2"
    name: "res4b2_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b2"
    top: "res4b3_branch2a"
    name: "res4b3_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b3_branch2a"
    top: "res4b3_branch2a"
    name: "bn4b3_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b3_branch2a"
    top: "res4b3_branch2a"
    name: "scale4b3_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b3_branch2a"
    top: "res4b3_branch2a"
    name: "res4b3_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b3_branch2a"
    top: "res4b3_branch2b"
    name: "res4b3_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b3_branch2b"
    top: "res4b3_branch2b"
    name: "bn4b3_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b3_branch2b"
    top: "res4b3_branch2b"
    name: "scale4b3_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b3_branch2b"
    top: "res4b3_branch2b"
    name: "res4b3_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b3_branch2b"
    top: "res4b3_branch2c"
    name: "res4b3_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b3_branch2c"
    top: "res4b3_branch2c"
    name: "bn4b3_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b3_branch2c"
    top: "res4b3_branch2c"
    name: "scale4b3_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b2"
    bottom: "res4b3_branch2c"
    top: "res4b3"
    name: "res4b3"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b3"
    top: "res4b3"
    name: "res4b3_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b3"
    top: "res4b4_branch2a"
    name: "res4b4_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b4_branch2a"
    top: "res4b4_branch2a"
    name: "bn4b4_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b4_branch2a"
    top: "res4b4_branch2a"
    name: "scale4b4_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b4_branch2a"
    top: "res4b4_branch2a"
    name: "res4b4_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b4_branch2a"
    top: "res4b4_branch2b"
    name: "res4b4_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b4_branch2b"
    top: "res4b4_branch2b"
    name: "bn4b4_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b4_branch2b"
    top: "res4b4_branch2b"
    name: "scale4b4_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b4_branch2b"
    top: "res4b4_branch2b"
    name: "res4b4_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b4_branch2b"
    top: "res4b4_branch2c"
    name: "res4b4_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b4_branch2c"
    top: "res4b4_branch2c"
    name: "bn4b4_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b4_branch2c"
    top: "res4b4_branch2c"
    name: "scale4b4_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b3"
    bottom: "res4b4_branch2c"
    top: "res4b4"
    name: "res4b4"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b4"
    top: "res4b4"
    name: "res4b4_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b4"
    top: "res4b5_branch2a"
    name: "res4b5_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b5_branch2a"
    top: "res4b5_branch2a"
    name: "bn4b5_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b5_branch2a"
    top: "res4b5_branch2a"
    name: "scale4b5_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b5_branch2a"
    top: "res4b5_branch2a"
    name: "res4b5_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b5_branch2a"
    top: "res4b5_branch2b"
    name: "res4b5_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b5_branch2b"
    top: "res4b5_branch2b"
    name: "bn4b5_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b5_branch2b"
    top: "res4b5_branch2b"
    name: "scale4b5_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b5_branch2b"
    top: "res4b5_branch2b"
    name: "res4b5_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b5_branch2b"
    top: "res4b5_branch2c"
    name: "res4b5_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b5_branch2c"
    top: "res4b5_branch2c"
    name: "bn4b5_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b5_branch2c"
    top: "res4b5_branch2c"
    name: "scale4b5_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b4"
    bottom: "res4b5_branch2c"
    top: "res4b5"
    name: "res4b5"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b5"
    top: "res4b5"
    name: "res4b5_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b5"
    top: "res4b6_branch2a"
    name: "res4b6_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b6_branch2a"
    top: "res4b6_branch2a"
    name: "bn4b6_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b6_branch2a"
    top: "res4b6_branch2a"
    name: "scale4b6_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b6_branch2a"
    top: "res4b6_branch2a"
    name: "res4b6_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b6_branch2a"
    top: "res4b6_branch2b"
    name: "res4b6_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b6_branch2b"
    top: "res4b6_branch2b"
    name: "bn4b6_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b6_branch2b"
    top: "res4b6_branch2b"
    name: "scale4b6_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b6_branch2b"
    top: "res4b6_branch2b"
    name: "res4b6_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b6_branch2b"
    top: "res4b6_branch2c"
    name: "res4b6_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b6_branch2c"
    top: "res4b6_branch2c"
    name: "bn4b6_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b6_branch2c"
    top: "res4b6_branch2c"
    name: "scale4b6_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b5"
    bottom: "res4b6_branch2c"
    top: "res4b6"
    name: "res4b6"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b6"
    top: "res4b6"
    name: "res4b6_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b6"
    top: "res4b7_branch2a"
    name: "res4b7_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b7_branch2a"
    top: "res4b7_branch2a"
    name: "bn4b7_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b7_branch2a"
    top: "res4b7_branch2a"
    name: "scale4b7_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b7_branch2a"
    top: "res4b7_branch2a"
    name: "res4b7_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b7_branch2a"
    top: "res4b7_branch2b"
    name: "res4b7_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b7_branch2b"
    top: "res4b7_branch2b"
    name: "bn4b7_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b7_branch2b"
    top: "res4b7_branch2b"
    name: "scale4b7_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b7_branch2b"
    top: "res4b7_branch2b"
    name: "res4b7_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b7_branch2b"
    top: "res4b7_branch2c"
    name: "res4b7_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b7_branch2c"
    top: "res4b7_branch2c"
    name: "bn4b7_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b7_branch2c"
    top: "res4b7_branch2c"
    name: "scale4b7_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b6"
    bottom: "res4b7_branch2c"
    top: "res4b7"
    name: "res4b7"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b7"
    top: "res4b7"
    name: "res4b7_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b7"
    top: "res4b8_branch2a"
    name: "res4b8_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b8_branch2a"
    top: "res4b8_branch2a"
    name: "bn4b8_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b8_branch2a"
    top: "res4b8_branch2a"
    name: "scale4b8_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b8_branch2a"
    top: "res4b8_branch2a"
    name: "res4b8_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b8_branch2a"
    top: "res4b8_branch2b"
    name: "res4b8_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b8_branch2b"
    top: "res4b8_branch2b"
    name: "bn4b8_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b8_branch2b"
    top: "res4b8_branch2b"
    name: "scale4b8_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b8_branch2b"
    top: "res4b8_branch2b"
    name: "res4b8_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b8_branch2b"
    top: "res4b8_branch2c"
    name: "res4b8_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b8_branch2c"
    top: "res4b8_branch2c"
    name: "bn4b8_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b8_branch2c"
    top: "res4b8_branch2c"
    name: "scale4b8_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b7"
    bottom: "res4b8_branch2c"
    top: "res4b8"
    name: "res4b8"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b8"
    top: "res4b8"
    name: "res4b8_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b8"
    top: "res4b9_branch2a"
    name: "res4b9_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b9_branch2a"
    top: "res4b9_branch2a"
    name: "bn4b9_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b9_branch2a"
    top: "res4b9_branch2a"
    name: "scale4b9_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b9_branch2a"
    top: "res4b9_branch2a"
    name: "res4b9_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b9_branch2a"
    top: "res4b9_branch2b"
    name: "res4b9_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b9_branch2b"
    top: "res4b9_branch2b"
    name: "bn4b9_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b9_branch2b"
    top: "res4b9_branch2b"
    name: "scale4b9_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b9_branch2b"
    top: "res4b9_branch2b"
    name: "res4b9_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b9_branch2b"
    top: "res4b9_branch2c"
    name: "res4b9_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b9_branch2c"
    top: "res4b9_branch2c"
    name: "bn4b9_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b9_branch2c"
    top: "res4b9_branch2c"
    name: "scale4b9_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b8"
    bottom: "res4b9_branch2c"
    top: "res4b9"
    name: "res4b9"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b9"
    top: "res4b9"
    name: "res4b9_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b9"
    top: "res4b10_branch2a"
    name: "res4b10_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b10_branch2a"
    top: "res4b10_branch2a"
    name: "bn4b10_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b10_branch2a"
    top: "res4b10_branch2a"
    name: "scale4b10_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b10_branch2a"
    top: "res4b10_branch2a"
    name: "res4b10_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b10_branch2a"
    top: "res4b10_branch2b"
    name: "res4b10_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b10_branch2b"
    top: "res4b10_branch2b"
    name: "bn4b10_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b10_branch2b"
    top: "res4b10_branch2b"
    name: "scale4b10_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b10_branch2b"
    top: "res4b10_branch2b"
    name: "res4b10_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b10_branch2b"
    top: "res4b10_branch2c"
    name: "res4b10_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b10_branch2c"
    top: "res4b10_branch2c"
    name: "bn4b10_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b10_branch2c"
    top: "res4b10_branch2c"
    name: "scale4b10_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b9"
    bottom: "res4b10_branch2c"
    top: "res4b10"
    name: "res4b10"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b10"
    top: "res4b10"
    name: "res4b10_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b10"
    top: "res4b11_branch2a"
    name: "res4b11_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b11_branch2a"
    top: "res4b11_branch2a"
    name: "bn4b11_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b11_branch2a"
    top: "res4b11_branch2a"
    name: "scale4b11_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b11_branch2a"
    top: "res4b11_branch2a"
    name: "res4b11_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b11_branch2a"
    top: "res4b11_branch2b"
    name: "res4b11_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b11_branch2b"
    top: "res4b11_branch2b"
    name: "bn4b11_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b11_branch2b"
    top: "res4b11_branch2b"
    name: "scale4b11_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b11_branch2b"
    top: "res4b11_branch2b"
    name: "res4b11_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b11_branch2b"
    top: "res4b11_branch2c"
    name: "res4b11_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b11_branch2c"
    top: "res4b11_branch2c"
    name: "bn4b11_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b11_branch2c"
    top: "res4b11_branch2c"
    name: "scale4b11_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b10"
    bottom: "res4b11_branch2c"
    top: "res4b11"
    name: "res4b11"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b11"
    top: "res4b11"
    name: "res4b11_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b11"
    top: "res4b12_branch2a"
    name: "res4b12_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b12_branch2a"
    top: "res4b12_branch2a"
    name: "bn4b12_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b12_branch2a"
    top: "res4b12_branch2a"
    name: "scale4b12_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b12_branch2a"
    top: "res4b12_branch2a"
    name: "res4b12_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b12_branch2a"
    top: "res4b12_branch2b"
    name: "res4b12_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b12_branch2b"
    top: "res4b12_branch2b"
    name: "bn4b12_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b12_branch2b"
    top: "res4b12_branch2b"
    name: "scale4b12_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b12_branch2b"
    top: "res4b12_branch2b"
    name: "res4b12_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b12_branch2b"
    top: "res4b12_branch2c"
    name: "res4b12_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b12_branch2c"
    top: "res4b12_branch2c"
    name: "bn4b12_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b12_branch2c"
    top: "res4b12_branch2c"
    name: "scale4b12_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b11"
    bottom: "res4b12_branch2c"
    top: "res4b12"
    name: "res4b12"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b12"
    top: "res4b12"
    name: "res4b12_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b12"
    top: "res4b13_branch2a"
    name: "res4b13_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b13_branch2a"
    top: "res4b13_branch2a"
    name: "bn4b13_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b13_branch2a"
    top: "res4b13_branch2a"
    name: "scale4b13_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b13_branch2a"
    top: "res4b13_branch2a"
    name: "res4b13_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b13_branch2a"
    top: "res4b13_branch2b"
    name: "res4b13_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b13_branch2b"
    top: "res4b13_branch2b"
    name: "bn4b13_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b13_branch2b"
    top: "res4b13_branch2b"
    name: "scale4b13_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b13_branch2b"
    top: "res4b13_branch2b"
    name: "res4b13_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b13_branch2b"
    top: "res4b13_branch2c"
    name: "res4b13_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b13_branch2c"
    top: "res4b13_branch2c"
    name: "bn4b13_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b13_branch2c"
    top: "res4b13_branch2c"
    name: "scale4b13_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b12"
    bottom: "res4b13_branch2c"
    top: "res4b13"
    name: "res4b13"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b13"
    top: "res4b13"
    name: "res4b13_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b13"
    top: "res4b14_branch2a"
    name: "res4b14_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b14_branch2a"
    top: "res4b14_branch2a"
    name: "bn4b14_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b14_branch2a"
    top: "res4b14_branch2a"
    name: "scale4b14_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b14_branch2a"
    top: "res4b14_branch2a"
    name: "res4b14_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b14_branch2a"
    top: "res4b14_branch2b"
    name: "res4b14_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b14_branch2b"
    top: "res4b14_branch2b"
    name: "bn4b14_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b14_branch2b"
    top: "res4b14_branch2b"
    name: "scale4b14_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b14_branch2b"
    top: "res4b14_branch2b"
    name: "res4b14_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b14_branch2b"
    top: "res4b14_branch2c"
    name: "res4b14_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b14_branch2c"
    top: "res4b14_branch2c"
    name: "bn4b14_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b14_branch2c"
    top: "res4b14_branch2c"
    name: "scale4b14_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b13"
    bottom: "res4b14_branch2c"
    top: "res4b14"
    name: "res4b14"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b14"
    top: "res4b14"
    name: "res4b14_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b14"
    top: "res4b15_branch2a"
    name: "res4b15_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b15_branch2a"
    top: "res4b15_branch2a"
    name: "bn4b15_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b15_branch2a"
    top: "res4b15_branch2a"
    name: "scale4b15_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b15_branch2a"
    top: "res4b15_branch2a"
    name: "res4b15_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b15_branch2a"
    top: "res4b15_branch2b"
    name: "res4b15_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b15_branch2b"
    top: "res4b15_branch2b"
    name: "bn4b15_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b15_branch2b"
    top: "res4b15_branch2b"
    name: "scale4b15_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b15_branch2b"
    top: "res4b15_branch2b"
    name: "res4b15_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b15_branch2b"
    top: "res4b15_branch2c"
    name: "res4b15_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b15_branch2c"
    top: "res4b15_branch2c"
    name: "bn4b15_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b15_branch2c"
    top: "res4b15_branch2c"
    name: "scale4b15_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b14"
    bottom: "res4b15_branch2c"
    top: "res4b15"
    name: "res4b15"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b15"
    top: "res4b15"
    name: "res4b15_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b15"
    top: "res4b16_branch2a"
    name: "res4b16_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b16_branch2a"
    top: "res4b16_branch2a"
    name: "bn4b16_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b16_branch2a"
    top: "res4b16_branch2a"
    name: "scale4b16_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b16_branch2a"
    top: "res4b16_branch2a"
    name: "res4b16_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b16_branch2a"
    top: "res4b16_branch2b"
    name: "res4b16_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b16_branch2b"
    top: "res4b16_branch2b"
    name: "bn4b16_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b16_branch2b"
    top: "res4b16_branch2b"
    name: "scale4b16_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b16_branch2b"
    top: "res4b16_branch2b"
    name: "res4b16_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b16_branch2b"
    top: "res4b16_branch2c"
    name: "res4b16_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b16_branch2c"
    top: "res4b16_branch2c"
    name: "bn4b16_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b16_branch2c"
    top: "res4b16_branch2c"
    name: "scale4b16_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b15"
    bottom: "res4b16_branch2c"
    top: "res4b16"
    name: "res4b16"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b16"
    top: "res4b16"
    name: "res4b16_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b16"
    top: "res4b17_branch2a"
    name: "res4b17_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b17_branch2a"
    top: "res4b17_branch2a"
    name: "bn4b17_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b17_branch2a"
    top: "res4b17_branch2a"
    name: "scale4b17_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b17_branch2a"
    top: "res4b17_branch2a"
    name: "res4b17_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b17_branch2a"
    top: "res4b17_branch2b"
    name: "res4b17_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b17_branch2b"
    top: "res4b17_branch2b"
    name: "bn4b17_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b17_branch2b"
    top: "res4b17_branch2b"
    name: "scale4b17_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b17_branch2b"
    top: "res4b17_branch2b"
    name: "res4b17_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b17_branch2b"
    top: "res4b17_branch2c"
    name: "res4b17_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b17_branch2c"
    top: "res4b17_branch2c"
    name: "bn4b17_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b17_branch2c"
    top: "res4b17_branch2c"
    name: "scale4b17_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b16"
    bottom: "res4b17_branch2c"
    top: "res4b17"
    name: "res4b17"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b17"
    top: "res4b17"
    name: "res4b17_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b17"
    top: "res4b18_branch2a"
    name: "res4b18_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b18_branch2a"
    top: "res4b18_branch2a"
    name: "bn4b18_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b18_branch2a"
    top: "res4b18_branch2a"
    name: "scale4b18_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b18_branch2a"
    top: "res4b18_branch2a"
    name: "res4b18_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b18_branch2a"
    top: "res4b18_branch2b"
    name: "res4b18_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b18_branch2b"
    top: "res4b18_branch2b"
    name: "bn4b18_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b18_branch2b"
    top: "res4b18_branch2b"
    name: "scale4b18_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b18_branch2b"
    top: "res4b18_branch2b"
    name: "res4b18_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b18_branch2b"
    top: "res4b18_branch2c"
    name: "res4b18_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b18_branch2c"
    top: "res4b18_branch2c"
    name: "bn4b18_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b18_branch2c"
    top: "res4b18_branch2c"
    name: "scale4b18_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b17"
    bottom: "res4b18_branch2c"
    top: "res4b18"
    name: "res4b18"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b18"
    top: "res4b18"
    name: "res4b18_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b18"
    top: "res4b19_branch2a"
    name: "res4b19_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b19_branch2a"
    top: "res4b19_branch2a"
    name: "bn4b19_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b19_branch2a"
    top: "res4b19_branch2a"
    name: "scale4b19_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b19_branch2a"
    top: "res4b19_branch2a"
    name: "res4b19_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b19_branch2a"
    top: "res4b19_branch2b"
    name: "res4b19_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b19_branch2b"
    top: "res4b19_branch2b"
    name: "bn4b19_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b19_branch2b"
    top: "res4b19_branch2b"
    name: "scale4b19_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b19_branch2b"
    top: "res4b19_branch2b"
    name: "res4b19_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b19_branch2b"
    top: "res4b19_branch2c"
    name: "res4b19_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b19_branch2c"
    top: "res4b19_branch2c"
    name: "bn4b19_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b19_branch2c"
    top: "res4b19_branch2c"
    name: "scale4b19_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b18"
    bottom: "res4b19_branch2c"
    top: "res4b19"
    name: "res4b19"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b19"
    top: "res4b19"
    name: "res4b19_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b19"
    top: "res4b20_branch2a"
    name: "res4b20_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b20_branch2a"
    top: "res4b20_branch2a"
    name: "bn4b20_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b20_branch2a"
    top: "res4b20_branch2a"
    name: "scale4b20_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b20_branch2a"
    top: "res4b20_branch2a"
    name: "res4b20_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b20_branch2a"
    top: "res4b20_branch2b"
    name: "res4b20_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b20_branch2b"
    top: "res4b20_branch2b"
    name: "bn4b20_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b20_branch2b"
    top: "res4b20_branch2b"
    name: "scale4b20_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b20_branch2b"
    top: "res4b20_branch2b"
    name: "res4b20_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b20_branch2b"
    top: "res4b20_branch2c"
    name: "res4b20_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b20_branch2c"
    top: "res4b20_branch2c"
    name: "bn4b20_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b20_branch2c"
    top: "res4b20_branch2c"
    name: "scale4b20_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b19"
    bottom: "res4b20_branch2c"
    top: "res4b20"
    name: "res4b20"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b20"
    top: "res4b20"
    name: "res4b20_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b20"
    top: "res4b21_branch2a"
    name: "res4b21_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b21_branch2a"
    top: "res4b21_branch2a"
    name: "bn4b21_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b21_branch2a"
    top: "res4b21_branch2a"
    name: "scale4b21_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b21_branch2a"
    top: "res4b21_branch2a"
    name: "res4b21_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b21_branch2a"
    top: "res4b21_branch2b"
    name: "res4b21_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b21_branch2b"
    top: "res4b21_branch2b"
    name: "bn4b21_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b21_branch2b"
    top: "res4b21_branch2b"
    name: "scale4b21_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b21_branch2b"
    top: "res4b21_branch2b"
    name: "res4b21_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b21_branch2b"
    top: "res4b21_branch2c"
    name: "res4b21_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b21_branch2c"
    top: "res4b21_branch2c"
    name: "bn4b21_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b21_branch2c"
    top: "res4b21_branch2c"
    name: "scale4b21_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b20"
    bottom: "res4b21_branch2c"
    top: "res4b21"
    name: "res4b21"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b21"
    top: "res4b21"
    name: "res4b21_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b21"
    top: "res4b22_branch2a"
    name: "res4b22_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b22_branch2a"
    top: "res4b22_branch2a"
    name: "bn4b22_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b22_branch2a"
    top: "res4b22_branch2a"
    name: "scale4b22_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b22_branch2a"
    top: "res4b22_branch2a"
    name: "res4b22_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b22_branch2a"
    top: "res4b22_branch2b"
    name: "res4b22_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b22_branch2b"
    top: "res4b22_branch2b"
    name: "bn4b22_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b22_branch2b"
    top: "res4b22_branch2b"
    name: "scale4b22_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b22_branch2b"
    top: "res4b22_branch2b"
    name: "res4b22_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b22_branch2b"
    top: "res4b22_branch2c"
    name: "res4b22_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b22_branch2c"
    top: "res4b22_branch2c"
    name: "bn4b22_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b22_branch2c"
    top: "res4b22_branch2c"
    name: "scale4b22_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b21"
    bottom: "res4b22_branch2c"
    top: "res4b22"
    name: "res4b22"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b22"
    top: "res4b22"
    name: "res4b22_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b22"
    top: "res4b23_branch2a"
    name: "res4b23_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b23_branch2a"
    top: "res4b23_branch2a"
    name: "bn4b23_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b23_branch2a"
    top: "res4b23_branch2a"
    name: "scale4b23_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b23_branch2a"
    top: "res4b23_branch2a"
    name: "res4b23_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b23_branch2a"
    top: "res4b23_branch2b"
    name: "res4b23_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b23_branch2b"
    top: "res4b23_branch2b"
    name: "bn4b23_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b23_branch2b"
    top: "res4b23_branch2b"
    name: "scale4b23_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b23_branch2b"
    top: "res4b23_branch2b"
    name: "res4b23_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b23_branch2b"
    top: "res4b23_branch2c"
    name: "res4b23_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b23_branch2c"
    top: "res4b23_branch2c"
    name: "bn4b23_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b23_branch2c"
    top: "res4b23_branch2c"
    name: "scale4b23_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b22"
    bottom: "res4b23_branch2c"
    top: "res4b23"
    name: "res4b23"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b23"
    top: "res4b23"
    name: "res4b23_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b23"
    top: "res4b24_branch2a"
    name: "res4b24_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b24_branch2a"
    top: "res4b24_branch2a"
    name: "bn4b24_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b24_branch2a"
    top: "res4b24_branch2a"
    name: "scale4b24_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b24_branch2a"
    top: "res4b24_branch2a"
    name: "res4b24_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b24_branch2a"
    top: "res4b24_branch2b"
    name: "res4b24_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b24_branch2b"
    top: "res4b24_branch2b"
    name: "bn4b24_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b24_branch2b"
    top: "res4b24_branch2b"
    name: "scale4b24_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b24_branch2b"
    top: "res4b24_branch2b"
    name: "res4b24_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b24_branch2b"
    top: "res4b24_branch2c"
    name: "res4b24_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b24_branch2c"
    top: "res4b24_branch2c"
    name: "bn4b24_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b24_branch2c"
    top: "res4b24_branch2c"
    name: "scale4b24_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b23"
    bottom: "res4b24_branch2c"
    top: "res4b24"
    name: "res4b24"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b24"
    top: "res4b24"
    name: "res4b24_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b24"
    top: "res4b25_branch2a"
    name: "res4b25_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b25_branch2a"
    top: "res4b25_branch2a"
    name: "bn4b25_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b25_branch2a"
    top: "res4b25_branch2a"
    name: "scale4b25_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b25_branch2a"
    top: "res4b25_branch2a"
    name: "res4b25_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b25_branch2a"
    top: "res4b25_branch2b"
    name: "res4b25_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b25_branch2b"
    top: "res4b25_branch2b"
    name: "bn4b25_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b25_branch2b"
    top: "res4b25_branch2b"
    name: "scale4b25_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b25_branch2b"
    top: "res4b25_branch2b"
    name: "res4b25_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b25_branch2b"
    top: "res4b25_branch2c"
    name: "res4b25_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b25_branch2c"
    top: "res4b25_branch2c"
    name: "bn4b25_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b25_branch2c"
    top: "res4b25_branch2c"
    name: "scale4b25_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b24"
    bottom: "res4b25_branch2c"
    top: "res4b25"
    name: "res4b25"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b25"
    top: "res4b25"
    name: "res4b25_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b25"
    top: "res4b26_branch2a"
    name: "res4b26_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b26_branch2a"
    top: "res4b26_branch2a"
    name: "bn4b26_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b26_branch2a"
    top: "res4b26_branch2a"
    name: "scale4b26_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b26_branch2a"
    top: "res4b26_branch2a"
    name: "res4b26_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b26_branch2a"
    top: "res4b26_branch2b"
    name: "res4b26_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b26_branch2b"
    top: "res4b26_branch2b"
    name: "bn4b26_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b26_branch2b"
    top: "res4b26_branch2b"
    name: "scale4b26_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b26_branch2b"
    top: "res4b26_branch2b"
    name: "res4b26_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b26_branch2b"
    top: "res4b26_branch2c"
    name: "res4b26_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b26_branch2c"
    top: "res4b26_branch2c"
    name: "bn4b26_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b26_branch2c"
    top: "res4b26_branch2c"
    name: "scale4b26_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b25"
    bottom: "res4b26_branch2c"
    top: "res4b26"
    name: "res4b26"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b26"
    top: "res4b26"
    name: "res4b26_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b26"
    top: "res4b27_branch2a"
    name: "res4b27_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b27_branch2a"
    top: "res4b27_branch2a"
    name: "bn4b27_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b27_branch2a"
    top: "res4b27_branch2a"
    name: "scale4b27_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b27_branch2a"
    top: "res4b27_branch2a"
    name: "res4b27_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b27_branch2a"
    top: "res4b27_branch2b"
    name: "res4b27_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b27_branch2b"
    top: "res4b27_branch2b"
    name: "bn4b27_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b27_branch2b"
    top: "res4b27_branch2b"
    name: "scale4b27_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b27_branch2b"
    top: "res4b27_branch2b"
    name: "res4b27_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b27_branch2b"
    top: "res4b27_branch2c"
    name: "res4b27_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b27_branch2c"
    top: "res4b27_branch2c"
    name: "bn4b27_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b27_branch2c"
    top: "res4b27_branch2c"
    name: "scale4b27_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b26"
    bottom: "res4b27_branch2c"
    top: "res4b27"
    name: "res4b27"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b27"
    top: "res4b27"
    name: "res4b27_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b27"
    top: "res4b28_branch2a"
    name: "res4b28_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b28_branch2a"
    top: "res4b28_branch2a"
    name: "bn4b28_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b28_branch2a"
    top: "res4b28_branch2a"
    name: "scale4b28_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b28_branch2a"
    top: "res4b28_branch2a"
    name: "res4b28_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b28_branch2a"
    top: "res4b28_branch2b"
    name: "res4b28_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b28_branch2b"
    top: "res4b28_branch2b"
    name: "bn4b28_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b28_branch2b"
    top: "res4b28_branch2b"
    name: "scale4b28_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b28_branch2b"
    top: "res4b28_branch2b"
    name: "res4b28_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b28_branch2b"
    top: "res4b28_branch2c"
    name: "res4b28_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b28_branch2c"
    top: "res4b28_branch2c"
    name: "bn4b28_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b28_branch2c"
    top: "res4b28_branch2c"
    name: "scale4b28_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b27"
    bottom: "res4b28_branch2c"
    top: "res4b28"
    name: "res4b28"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b28"
    top: "res4b28"
    name: "res4b28_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b28"
    top: "res4b29_branch2a"
    name: "res4b29_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b29_branch2a"
    top: "res4b29_branch2a"
    name: "bn4b29_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b29_branch2a"
    top: "res4b29_branch2a"
    name: "scale4b29_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b29_branch2a"
    top: "res4b29_branch2a"
    name: "res4b29_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b29_branch2a"
    top: "res4b29_branch2b"
    name: "res4b29_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b29_branch2b"
    top: "res4b29_branch2b"
    name: "bn4b29_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b29_branch2b"
    top: "res4b29_branch2b"
    name: "scale4b29_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b29_branch2b"
    top: "res4b29_branch2b"
    name: "res4b29_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b29_branch2b"
    top: "res4b29_branch2c"
    name: "res4b29_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b29_branch2c"
    top: "res4b29_branch2c"
    name: "bn4b29_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b29_branch2c"
    top: "res4b29_branch2c"
    name: "scale4b29_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b28"
    bottom: "res4b29_branch2c"
    top: "res4b29"
    name: "res4b29"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b29"
    top: "res4b29"
    name: "res4b29_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b29"
    top: "res4b30_branch2a"
    name: "res4b30_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b30_branch2a"
    top: "res4b30_branch2a"
    name: "bn4b30_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b30_branch2a"
    top: "res4b30_branch2a"
    name: "scale4b30_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b30_branch2a"
    top: "res4b30_branch2a"
    name: "res4b30_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b30_branch2a"
    top: "res4b30_branch2b"
    name: "res4b30_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b30_branch2b"
    top: "res4b30_branch2b"
    name: "bn4b30_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b30_branch2b"
    top: "res4b30_branch2b"
    name: "scale4b30_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b30_branch2b"
    top: "res4b30_branch2b"
    name: "res4b30_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b30_branch2b"
    top: "res4b30_branch2c"
    name: "res4b30_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b30_branch2c"
    top: "res4b30_branch2c"
    name: "bn4b30_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b30_branch2c"
    top: "res4b30_branch2c"
    name: "scale4b30_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b29"
    bottom: "res4b30_branch2c"
    top: "res4b30"
    name: "res4b30"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b30"
    top: "res4b30"
    name: "res4b30_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b30"
    top: "res4b31_branch2a"
    name: "res4b31_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b31_branch2a"
    top: "res4b31_branch2a"
    name: "bn4b31_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b31_branch2a"
    top: "res4b31_branch2a"
    name: "scale4b31_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b31_branch2a"
    top: "res4b31_branch2a"
    name: "res4b31_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b31_branch2a"
    top: "res4b31_branch2b"
    name: "res4b31_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b31_branch2b"
    top: "res4b31_branch2b"
    name: "bn4b31_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b31_branch2b"
    top: "res4b31_branch2b"
    name: "scale4b31_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b31_branch2b"
    top: "res4b31_branch2b"
    name: "res4b31_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b31_branch2b"
    top: "res4b31_branch2c"
    name: "res4b31_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b31_branch2c"
    top: "res4b31_branch2c"
    name: "bn4b31_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b31_branch2c"
    top: "res4b31_branch2c"
    name: "scale4b31_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b30"
    bottom: "res4b31_branch2c"
    top: "res4b31"
    name: "res4b31"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b31"
    top: "res4b31"
    name: "res4b31_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b31"
    top: "res4b32_branch2a"
    name: "res4b32_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b32_branch2a"
    top: "res4b32_branch2a"
    name: "bn4b32_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b32_branch2a"
    top: "res4b32_branch2a"
    name: "scale4b32_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b32_branch2a"
    top: "res4b32_branch2a"
    name: "res4b32_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b32_branch2a"
    top: "res4b32_branch2b"
    name: "res4b32_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b32_branch2b"
    top: "res4b32_branch2b"
    name: "bn4b32_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b32_branch2b"
    top: "res4b32_branch2b"
    name: "scale4b32_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b32_branch2b"
    top: "res4b32_branch2b"
    name: "res4b32_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b32_branch2b"
    top: "res4b32_branch2c"
    name: "res4b32_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b32_branch2c"
    top: "res4b32_branch2c"
    name: "bn4b32_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b32_branch2c"
    top: "res4b32_branch2c"
    name: "scale4b32_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b31"
    bottom: "res4b32_branch2c"
    top: "res4b32"
    name: "res4b32"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b32"
    top: "res4b32"
    name: "res4b32_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b32"
    top: "res4b33_branch2a"
    name: "res4b33_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b33_branch2a"
    top: "res4b33_branch2a"
    name: "bn4b33_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b33_branch2a"
    top: "res4b33_branch2a"
    name: "scale4b33_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b33_branch2a"
    top: "res4b33_branch2a"
    name: "res4b33_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b33_branch2a"
    top: "res4b33_branch2b"
    name: "res4b33_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b33_branch2b"
    top: "res4b33_branch2b"
    name: "bn4b33_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b33_branch2b"
    top: "res4b33_branch2b"
    name: "scale4b33_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b33_branch2b"
    top: "res4b33_branch2b"
    name: "res4b33_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b33_branch2b"
    top: "res4b33_branch2c"
    name: "res4b33_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b33_branch2c"
    top: "res4b33_branch2c"
    name: "bn4b33_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b33_branch2c"
    top: "res4b33_branch2c"
    name: "scale4b33_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b32"
    bottom: "res4b33_branch2c"
    top: "res4b33"
    name: "res4b33"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b33"
    top: "res4b33"
    name: "res4b33_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b33"
    top: "res4b34_branch2a"
    name: "res4b34_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b34_branch2a"
    top: "res4b34_branch2a"
    name: "bn4b34_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b34_branch2a"
    top: "res4b34_branch2a"
    name: "scale4b34_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b34_branch2a"
    top: "res4b34_branch2a"
    name: "res4b34_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b34_branch2a"
    top: "res4b34_branch2b"
    name: "res4b34_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b34_branch2b"
    top: "res4b34_branch2b"
    name: "bn4b34_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b34_branch2b"
    top: "res4b34_branch2b"
    name: "scale4b34_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b34_branch2b"
    top: "res4b34_branch2b"
    name: "res4b34_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b34_branch2b"
    top: "res4b34_branch2c"
    name: "res4b34_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b34_branch2c"
    top: "res4b34_branch2c"
    name: "bn4b34_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b34_branch2c"
    top: "res4b34_branch2c"
    name: "scale4b34_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b33"
    bottom: "res4b34_branch2c"
    top: "res4b34"
    name: "res4b34"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b34"
    top: "res4b34"
    name: "res4b34_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b34"
    top: "res4b35_branch2a"
    name: "res4b35_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b35_branch2a"
    top: "res4b35_branch2a"
    name: "bn4b35_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b35_branch2a"
    top: "res4b35_branch2a"
    name: "scale4b35_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b35_branch2a"
    top: "res4b35_branch2a"
    name: "res4b35_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b35_branch2a"
    top: "res4b35_branch2b"
    name: "res4b35_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b35_branch2b"
    top: "res4b35_branch2b"
    name: "bn4b35_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b35_branch2b"
    top: "res4b35_branch2b"
    name: "scale4b35_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b35_branch2b"
    top: "res4b35_branch2b"
    name: "res4b35_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b35_branch2b"
    top: "res4b35_branch2c"
    name: "res4b35_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res4b35_branch2c"
    top: "res4b35_branch2c"
    name: "bn4b35_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res4b35_branch2c"
    top: "res4b35_branch2c"
    name: "scale4b35_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b34"
    bottom: "res4b35_branch2c"
    top: "res4b35"
    name: "res4b35"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res4b35"
    top: "res4b35"
    name: "res4b35_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b35"
    top: "res5a_branch1"
    name: "res5a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 2
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5a_branch1"
    top: "res5a_branch1"
    name: "bn5a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5a_branch1"
    top: "res5a_branch1"
    name: "scale5a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b35"
    top: "res5a_branch2a"
    name: "res5a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 2
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5a_branch2a"
    top: "res5a_branch2a"
    name: "bn5a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5a_branch2a"
    top: "res5a_branch2a"
    name: "scale5a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5a_branch2a"
    top: "res5a_branch2a"
    name: "res5a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5a_branch2a"
    top: "res5a_branch2b"
    name: "res5a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5a_branch2b"
    top: "res5a_branch2b"
    name: "bn5a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5a_branch2b"
    top: "res5a_branch2b"
    name: "scale5a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5a_branch2b"
    top: "res5a_branch2b"
    name: "res5a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5a_branch2b"
    top: "res5a_branch2c"
    name: "res5a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5a_branch2c"
    top: "res5a_branch2c"
    name: "bn5a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5a_branch2c"
    top: "res5a_branch2c"
    name: "scale5a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5a_branch1"
    bottom: "res5a_branch2c"
    top: "res5a"
    name: "res5a"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res5a"
    top: "res5a"
    name: "res5a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5a"
    top: "res5b_branch2a"
    name: "res5b_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5b_branch2a"
    top: "res5b_branch2a"
    name: "bn5b_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5b_branch2a"
    top: "res5b_branch2a"
    name: "scale5b_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5b_branch2a"
    top: "res5b_branch2a"
    name: "res5b_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5b_branch2a"
    top: "res5b_branch2b"
    name: "res5b_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5b_branch2b"
    top: "res5b_branch2b"
    name: "bn5b_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5b_branch2b"
    top: "res5b_branch2b"
    name: "scale5b_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5b_branch2b"
    top: "res5b_branch2b"
    name: "res5b_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5b_branch2b"
    top: "res5b_branch2c"
    name: "res5b_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5b_branch2c"
    top: "res5b_branch2c"
    name: "bn5b_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5b_branch2c"
    top: "res5b_branch2c"
    name: "scale5b_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5a"
    bottom: "res5b_branch2c"
    top: "res5b"
    name: "res5b"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res5b"
    top: "res5b"
    name: "res5b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5b"
    top: "res5c_branch2a"
    name: "res5c_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5c_branch2a"
    top: "res5c_branch2a"
    name: "bn5c_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5c_branch2a"
    top: "res5c_branch2a"
    name: "scale5c_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5c_branch2a"
    top: "res5c_branch2a"
    name: "res5c_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5c_branch2a"
    top: "res5c_branch2b"
    name: "res5c_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 3
        pad: 1
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5c_branch2b"
    top: "res5c_branch2b"
    name: "bn5c_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5c_branch2b"
    top: "res5c_branch2b"
    name: "scale5c_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5c_branch2b"
    top: "res5c_branch2b"
    name: "res5c_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5c_branch2b"
    top: "res5c_branch2c"
    name: "res5c_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 1
        weight_filler {
            type: "msra"
        }
        bias_term: false

    }
}

layer {
    bottom: "res5c_branch2c"
    top: "res5c_branch2c"
    name: "bn5c_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: false
    }
}

layer {
    bottom: "res5c_branch2c"
    top: "res5c_branch2c"
    name: "scale5c_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5b"
    bottom: "res5c_branch2c"
    top: "res5c"
    name: "res5c"
    type: "Eltwise"
    eltwise_param {
        operation: SUM
    }
}

layer {
    bottom: "res5c"
    top: "res5c"
    name: "res5c_relu"
    type: "ReLU"
}

layer {
    bottom: "res5c"
    top: "pool5"
    name: "pool5"
    type: "Pooling"
    pooling_param {
        kernel_size: 7
        stride: 1
        pool: AVE
    }
}

layer {
    bottom: "pool5"
    top: "fc3"
    name: "fc3"
    type: "InnerProduct"
    param {
        lr_mult: 1
        decay_mult: 1
    }
    param {
        lr_mult: 2
        decay_mult: 1
    }
    inner_product_param {
        num_output: 3
        weight_filler {
            type: "xavier"
        }
        bias_filler {
            type: "constant"
            value: 0
        }
    }
}

layer {
    bottom: "fc3"
    bottom: "label"
    name: "loss"
    type: "SoftmaxWithLoss"
    top: "loss"
}

layer {
       name: "probt"
       type: "Softmax"
       bottom: "fc3"
       top: "probt"
       include {
       phase: TEST
    }
}

layer {
  bottom: "fc3"
  bottom: "label"
  top: "accuracy@1"
  name: "accuracy/top1"
  type: "Accuracy"
  accuracy_param {
    top_k: 1
  }
}

       deploy.prototxt

name: "ResNet-152"
input: "data"
input_shape {
  dim: 1
  dim: 3
  dim: 224
  dim: 224
}

layer {
    bottom: "data"
    top: "conv1"
    name: "conv1"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 7
        pad: 3
        stride: 2
        bias_term: false
    }
}

layer {
    bottom: "conv1"
    top: "conv1"
    name: "bn_conv1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "conv1"
    top: "conv1"
    name: "scale_conv1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "conv1"
    bottom: "conv1"
    name: "conv1_relu"
    type: "ReLU"
}

layer {
    bottom: "conv1"
    top: "pool1"
    name: "pool1"
    type: "Pooling"
    pooling_param {
        kernel_size: 3
        stride: 2
        pool: MAX
    }
}

layer {
    bottom: "pool1"
    top: "res2a_branch1"
    name: "res2a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2a_branch1"
    top: "res2a_branch1"
    name: "bn2a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2a_branch1"
    top: "res2a_branch1"
    name: "scale2a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "pool1"
    top: "res2a_branch2a"
    name: "res2a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2a_branch2a"
    top: "res2a_branch2a"
    name: "bn2a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2a_branch2a"
    top: "res2a_branch2a"
    name: "scale2a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res2a_branch2a"
    bottom: "res2a_branch2a"
    name: "res2a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2a_branch2a"
    top: "res2a_branch2b"
    name: "res2a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2a_branch2b"
    top: "res2a_branch2b"
    name: "bn2a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2a_branch2b"
    top: "res2a_branch2b"
    name: "scale2a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res2a_branch2b"
    bottom: "res2a_branch2b"
    name: "res2a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2a_branch2b"
    top: "res2a_branch2c"
    name: "res2a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2a_branch2c"
    top: "res2a_branch2c"
    name: "bn2a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2a_branch2c"
    top: "res2a_branch2c"
    name: "scale2a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2a_branch1"
    bottom: "res2a_branch2c"
    top: "res2a"
    name: "res2a"
    type: "Eltwise"
}

layer {
    bottom: "res2a"
    top: "res2a"
    name: "res2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2a"
    top: "res2b_branch2a"
    name: "res2b_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2b_branch2a"
    top: "res2b_branch2a"
    name: "bn2b_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2b_branch2a"
    top: "res2b_branch2a"
    name: "scale2b_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res2b_branch2a"
    bottom: "res2b_branch2a"
    name: "res2b_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2b_branch2a"
    top: "res2b_branch2b"
    name: "res2b_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2b_branch2b"
    top: "res2b_branch2b"
    name: "bn2b_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2b_branch2b"
    top: "res2b_branch2b"
    name: "scale2b_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res2b_branch2b"
    bottom: "res2b_branch2b"
    name: "res2b_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2b_branch2b"
    top: "res2b_branch2c"
    name: "res2b_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2b_branch2c"
    top: "res2b_branch2c"
    name: "bn2b_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2b_branch2c"
    top: "res2b_branch2c"
    name: "scale2b_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2a"
    bottom: "res2b_branch2c"
    top: "res2b"
    name: "res2b"
    type: "Eltwise"
}

layer {
    bottom: "res2b"
    top: "res2b"
    name: "res2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2b"
    top: "res2c_branch2a"
    name: "res2c_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2c_branch2a"
    top: "res2c_branch2a"
    name: "bn2c_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2c_branch2a"
    top: "res2c_branch2a"
    name: "scale2c_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res2c_branch2a"
    bottom: "res2c_branch2a"
    name: "res2c_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res2c_branch2a"
    top: "res2c_branch2b"
    name: "res2c_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 64
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2c_branch2b"
    top: "res2c_branch2b"
    name: "bn2c_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2c_branch2b"
    top: "res2c_branch2b"
    name: "scale2c_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res2c_branch2b"
    bottom: "res2c_branch2b"
    name: "res2c_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res2c_branch2b"
    top: "res2c_branch2c"
    name: "res2c_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res2c_branch2c"
    top: "res2c_branch2c"
    name: "bn2c_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res2c_branch2c"
    top: "res2c_branch2c"
    name: "scale2c_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2b"
    bottom: "res2c_branch2c"
    top: "res2c"
    name: "res2c"
    type: "Eltwise"
}

layer {
    bottom: "res2c"
    top: "res2c"
    name: "res2c_relu"
    type: "ReLU"
}

layer {
    bottom: "res2c"
    top: "res3a_branch1"
    name: "res3a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 2
        bias_term: false
    }
}

layer {
    bottom: "res3a_branch1"
    top: "res3a_branch1"
    name: "bn3a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3a_branch1"
    top: "res3a_branch1"
    name: "scale3a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res2c"
    top: "res3a_branch2a"
    name: "res3a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 2
        bias_term: false
    }
}

layer {
    bottom: "res3a_branch2a"
    top: "res3a_branch2a"
    name: "bn3a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3a_branch2a"
    top: "res3a_branch2a"
    name: "scale3a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3a_branch2a"
    bottom: "res3a_branch2a"
    name: "res3a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3a_branch2a"
    top: "res3a_branch2b"
    name: "res3a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3a_branch2b"
    top: "res3a_branch2b"
    name: "bn3a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3a_branch2b"
    top: "res3a_branch2b"
    name: "scale3a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3a_branch2b"
    bottom: "res3a_branch2b"
    name: "res3a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3a_branch2b"
    top: "res3a_branch2c"
    name: "res3a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3a_branch2c"
    top: "res3a_branch2c"
    name: "bn3a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3a_branch2c"
    top: "res3a_branch2c"
    name: "scale3a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3a_branch1"
    bottom: "res3a_branch2c"
    top: "res3a"
    name: "res3a"
    type: "Eltwise"
}

layer {
    bottom: "res3a"
    top: "res3a"
    name: "res3a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3a"
    top: "res3b1_branch2a"
    name: "res3b1_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b1_branch2a"
    top: "res3b1_branch2a"
    name: "bn3b1_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b1_branch2a"
    top: "res3b1_branch2a"
    name: "scale3b1_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b1_branch2a"
    bottom: "res3b1_branch2a"
    name: "res3b1_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b1_branch2a"
    top: "res3b1_branch2b"
    name: "res3b1_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b1_branch2b"
    top: "res3b1_branch2b"
    name: "bn3b1_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b1_branch2b"
    top: "res3b1_branch2b"
    name: "scale3b1_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b1_branch2b"
    bottom: "res3b1_branch2b"
    name: "res3b1_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b1_branch2b"
    top: "res3b1_branch2c"
    name: "res3b1_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b1_branch2c"
    top: "res3b1_branch2c"
    name: "bn3b1_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b1_branch2c"
    top: "res3b1_branch2c"
    name: "scale3b1_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3a"
    bottom: "res3b1_branch2c"
    top: "res3b1"
    name: "res3b1"
    type: "Eltwise"
}

layer {
    bottom: "res3b1"
    top: "res3b1"
    name: "res3b1_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b1"
    top: "res3b2_branch2a"
    name: "res3b2_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b2_branch2a"
    top: "res3b2_branch2a"
    name: "bn3b2_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b2_branch2a"
    top: "res3b2_branch2a"
    name: "scale3b2_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b2_branch2a"
    bottom: "res3b2_branch2a"
    name: "res3b2_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b2_branch2a"
    top: "res3b2_branch2b"
    name: "res3b2_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b2_branch2b"
    top: "res3b2_branch2b"
    name: "bn3b2_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b2_branch2b"
    top: "res3b2_branch2b"
    name: "scale3b2_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b2_branch2b"
    bottom: "res3b2_branch2b"
    name: "res3b2_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b2_branch2b"
    top: "res3b2_branch2c"
    name: "res3b2_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b2_branch2c"
    top: "res3b2_branch2c"
    name: "bn3b2_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b2_branch2c"
    top: "res3b2_branch2c"
    name: "scale3b2_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b1"
    bottom: "res3b2_branch2c"
    top: "res3b2"
    name: "res3b2"
    type: "Eltwise"
}

layer {
    bottom: "res3b2"
    top: "res3b2"
    name: "res3b2_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b2"
    top: "res3b3_branch2a"
    name: "res3b3_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b3_branch2a"
    top: "res3b3_branch2a"
    name: "bn3b3_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b3_branch2a"
    top: "res3b3_branch2a"
    name: "scale3b3_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b3_branch2a"
    bottom: "res3b3_branch2a"
    name: "res3b3_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b3_branch2a"
    top: "res3b3_branch2b"
    name: "res3b3_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b3_branch2b"
    top: "res3b3_branch2b"
    name: "bn3b3_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b3_branch2b"
    top: "res3b3_branch2b"
    name: "scale3b3_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b3_branch2b"
    bottom: "res3b3_branch2b"
    name: "res3b3_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b3_branch2b"
    top: "res3b3_branch2c"
    name: "res3b3_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b3_branch2c"
    top: "res3b3_branch2c"
    name: "bn3b3_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b3_branch2c"
    top: "res3b3_branch2c"
    name: "scale3b3_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b2"
    bottom: "res3b3_branch2c"
    top: "res3b3"
    name: "res3b3"
    type: "Eltwise"
}

layer {
    bottom: "res3b3"
    top: "res3b3"
    name: "res3b3_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b3"
    top: "res3b4_branch2a"
    name: "res3b4_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b4_branch2a"
    top: "res3b4_branch2a"
    name: "bn3b4_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b4_branch2a"
    top: "res3b4_branch2a"
    name: "scale3b4_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b4_branch2a"
    bottom: "res3b4_branch2a"
    name: "res3b4_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b4_branch2a"
    top: "res3b4_branch2b"
    name: "res3b4_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b4_branch2b"
    top: "res3b4_branch2b"
    name: "bn3b4_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b4_branch2b"
    top: "res3b4_branch2b"
    name: "scale3b4_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b4_branch2b"
    bottom: "res3b4_branch2b"
    name: "res3b4_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b4_branch2b"
    top: "res3b4_branch2c"
    name: "res3b4_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b4_branch2c"
    top: "res3b4_branch2c"
    name: "bn3b4_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b4_branch2c"
    top: "res3b4_branch2c"
    name: "scale3b4_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b3"
    bottom: "res3b4_branch2c"
    top: "res3b4"
    name: "res3b4"
    type: "Eltwise"
}

layer {
    bottom: "res3b4"
    top: "res3b4"
    name: "res3b4_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b4"
    top: "res3b5_branch2a"
    name: "res3b5_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b5_branch2a"
    top: "res3b5_branch2a"
    name: "bn3b5_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b5_branch2a"
    top: "res3b5_branch2a"
    name: "scale3b5_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b5_branch2a"
    bottom: "res3b5_branch2a"
    name: "res3b5_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b5_branch2a"
    top: "res3b5_branch2b"
    name: "res3b5_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b5_branch2b"
    top: "res3b5_branch2b"
    name: "bn3b5_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b5_branch2b"
    top: "res3b5_branch2b"
    name: "scale3b5_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b5_branch2b"
    bottom: "res3b5_branch2b"
    name: "res3b5_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b5_branch2b"
    top: "res3b5_branch2c"
    name: "res3b5_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b5_branch2c"
    top: "res3b5_branch2c"
    name: "bn3b5_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b5_branch2c"
    top: "res3b5_branch2c"
    name: "scale3b5_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b4"
    bottom: "res3b5_branch2c"
    top: "res3b5"
    name: "res3b5"
    type: "Eltwise"
}

layer {
    bottom: "res3b5"
    top: "res3b5"
    name: "res3b5_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b5"
    top: "res3b6_branch2a"
    name: "res3b6_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b6_branch2a"
    top: "res3b6_branch2a"
    name: "bn3b6_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b6_branch2a"
    top: "res3b6_branch2a"
    name: "scale3b6_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b6_branch2a"
    bottom: "res3b6_branch2a"
    name: "res3b6_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b6_branch2a"
    top: "res3b6_branch2b"
    name: "res3b6_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b6_branch2b"
    top: "res3b6_branch2b"
    name: "bn3b6_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b6_branch2b"
    top: "res3b6_branch2b"
    name: "scale3b6_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b6_branch2b"
    bottom: "res3b6_branch2b"
    name: "res3b6_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b6_branch2b"
    top: "res3b6_branch2c"
    name: "res3b6_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b6_branch2c"
    top: "res3b6_branch2c"
    name: "bn3b6_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b6_branch2c"
    top: "res3b6_branch2c"
    name: "scale3b6_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b5"
    bottom: "res3b6_branch2c"
    top: "res3b6"
    name: "res3b6"
    type: "Eltwise"
}

layer {
    bottom: "res3b6"
    top: "res3b6"
    name: "res3b6_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b6"
    top: "res3b7_branch2a"
    name: "res3b7_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b7_branch2a"
    top: "res3b7_branch2a"
    name: "bn3b7_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b7_branch2a"
    top: "res3b7_branch2a"
    name: "scale3b7_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b7_branch2a"
    bottom: "res3b7_branch2a"
    name: "res3b7_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b7_branch2a"
    top: "res3b7_branch2b"
    name: "res3b7_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 128
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b7_branch2b"
    top: "res3b7_branch2b"
    name: "bn3b7_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b7_branch2b"
    top: "res3b7_branch2b"
    name: "scale3b7_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res3b7_branch2b"
    bottom: "res3b7_branch2b"
    name: "res3b7_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b7_branch2b"
    top: "res3b7_branch2c"
    name: "res3b7_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res3b7_branch2c"
    top: "res3b7_branch2c"
    name: "bn3b7_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res3b7_branch2c"
    top: "res3b7_branch2c"
    name: "scale3b7_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b6"
    bottom: "res3b7_branch2c"
    top: "res3b7"
    name: "res3b7"
    type: "Eltwise"
}

layer {
    bottom: "res3b7"
    top: "res3b7"
    name: "res3b7_relu"
    type: "ReLU"
}

layer {
    bottom: "res3b7"
    top: "res4a_branch1"
    name: "res4a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 2
        bias_term: false
    }
}

layer {
    bottom: "res4a_branch1"
    top: "res4a_branch1"
    name: "bn4a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4a_branch1"
    top: "res4a_branch1"
    name: "scale4a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res3b7"
    top: "res4a_branch2a"
    name: "res4a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 2
        bias_term: false
    }
}

layer {
    bottom: "res4a_branch2a"
    top: "res4a_branch2a"
    name: "bn4a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4a_branch2a"
    top: "res4a_branch2a"
    name: "scale4a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4a_branch2a"
    bottom: "res4a_branch2a"
    name: "res4a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4a_branch2a"
    top: "res4a_branch2b"
    name: "res4a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4a_branch2b"
    top: "res4a_branch2b"
    name: "bn4a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4a_branch2b"
    top: "res4a_branch2b"
    name: "scale4a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4a_branch2b"
    bottom: "res4a_branch2b"
    name: "res4a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4a_branch2b"
    top: "res4a_branch2c"
    name: "res4a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4a_branch2c"
    top: "res4a_branch2c"
    name: "bn4a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4a_branch2c"
    top: "res4a_branch2c"
    name: "scale4a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4a_branch1"
    bottom: "res4a_branch2c"
    top: "res4a"
    name: "res4a"
    type: "Eltwise"
}

layer {
    bottom: "res4a"
    top: "res4a"
    name: "res4a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4a"
    top: "res4b1_branch2a"
    name: "res4b1_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b1_branch2a"
    top: "res4b1_branch2a"
    name: "bn4b1_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b1_branch2a"
    top: "res4b1_branch2a"
    name: "scale4b1_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b1_branch2a"
    bottom: "res4b1_branch2a"
    name: "res4b1_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b1_branch2a"
    top: "res4b1_branch2b"
    name: "res4b1_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b1_branch2b"
    top: "res4b1_branch2b"
    name: "bn4b1_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b1_branch2b"
    top: "res4b1_branch2b"
    name: "scale4b1_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b1_branch2b"
    bottom: "res4b1_branch2b"
    name: "res4b1_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b1_branch2b"
    top: "res4b1_branch2c"
    name: "res4b1_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b1_branch2c"
    top: "res4b1_branch2c"
    name: "bn4b1_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b1_branch2c"
    top: "res4b1_branch2c"
    name: "scale4b1_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4a"
    bottom: "res4b1_branch2c"
    top: "res4b1"
    name: "res4b1"
    type: "Eltwise"
}

layer {
    bottom: "res4b1"
    top: "res4b1"
    name: "res4b1_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b1"
    top: "res4b2_branch2a"
    name: "res4b2_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b2_branch2a"
    top: "res4b2_branch2a"
    name: "bn4b2_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b2_branch2a"
    top: "res4b2_branch2a"
    name: "scale4b2_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b2_branch2a"
    bottom: "res4b2_branch2a"
    name: "res4b2_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b2_branch2a"
    top: "res4b2_branch2b"
    name: "res4b2_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b2_branch2b"
    top: "res4b2_branch2b"
    name: "bn4b2_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b2_branch2b"
    top: "res4b2_branch2b"
    name: "scale4b2_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b2_branch2b"
    bottom: "res4b2_branch2b"
    name: "res4b2_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b2_branch2b"
    top: "res4b2_branch2c"
    name: "res4b2_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b2_branch2c"
    top: "res4b2_branch2c"
    name: "bn4b2_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b2_branch2c"
    top: "res4b2_branch2c"
    name: "scale4b2_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b1"
    bottom: "res4b2_branch2c"
    top: "res4b2"
    name: "res4b2"
    type: "Eltwise"
}

layer {
    bottom: "res4b2"
    top: "res4b2"
    name: "res4b2_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b2"
    top: "res4b3_branch2a"
    name: "res4b3_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b3_branch2a"
    top: "res4b3_branch2a"
    name: "bn4b3_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b3_branch2a"
    top: "res4b3_branch2a"
    name: "scale4b3_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b3_branch2a"
    bottom: "res4b3_branch2a"
    name: "res4b3_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b3_branch2a"
    top: "res4b3_branch2b"
    name: "res4b3_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b3_branch2b"
    top: "res4b3_branch2b"
    name: "bn4b3_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b3_branch2b"
    top: "res4b3_branch2b"
    name: "scale4b3_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b3_branch2b"
    bottom: "res4b3_branch2b"
    name: "res4b3_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b3_branch2b"
    top: "res4b3_branch2c"
    name: "res4b3_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b3_branch2c"
    top: "res4b3_branch2c"
    name: "bn4b3_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b3_branch2c"
    top: "res4b3_branch2c"
    name: "scale4b3_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b2"
    bottom: "res4b3_branch2c"
    top: "res4b3"
    name: "res4b3"
    type: "Eltwise"
}

layer {
    bottom: "res4b3"
    top: "res4b3"
    name: "res4b3_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b3"
    top: "res4b4_branch2a"
    name: "res4b4_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b4_branch2a"
    top: "res4b4_branch2a"
    name: "bn4b4_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b4_branch2a"
    top: "res4b4_branch2a"
    name: "scale4b4_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b4_branch2a"
    bottom: "res4b4_branch2a"
    name: "res4b4_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b4_branch2a"
    top: "res4b4_branch2b"
    name: "res4b4_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b4_branch2b"
    top: "res4b4_branch2b"
    name: "bn4b4_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b4_branch2b"
    top: "res4b4_branch2b"
    name: "scale4b4_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b4_branch2b"
    bottom: "res4b4_branch2b"
    name: "res4b4_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b4_branch2b"
    top: "res4b4_branch2c"
    name: "res4b4_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b4_branch2c"
    top: "res4b4_branch2c"
    name: "bn4b4_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b4_branch2c"
    top: "res4b4_branch2c"
    name: "scale4b4_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b3"
    bottom: "res4b4_branch2c"
    top: "res4b4"
    name: "res4b4"
    type: "Eltwise"
}

layer {
    bottom: "res4b4"
    top: "res4b4"
    name: "res4b4_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b4"
    top: "res4b5_branch2a"
    name: "res4b5_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b5_branch2a"
    top: "res4b5_branch2a"
    name: "bn4b5_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b5_branch2a"
    top: "res4b5_branch2a"
    name: "scale4b5_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b5_branch2a"
    bottom: "res4b5_branch2a"
    name: "res4b5_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b5_branch2a"
    top: "res4b5_branch2b"
    name: "res4b5_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b5_branch2b"
    top: "res4b5_branch2b"
    name: "bn4b5_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b5_branch2b"
    top: "res4b5_branch2b"
    name: "scale4b5_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b5_branch2b"
    bottom: "res4b5_branch2b"
    name: "res4b5_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b5_branch2b"
    top: "res4b5_branch2c"
    name: "res4b5_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b5_branch2c"
    top: "res4b5_branch2c"
    name: "bn4b5_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b5_branch2c"
    top: "res4b5_branch2c"
    name: "scale4b5_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b4"
    bottom: "res4b5_branch2c"
    top: "res4b5"
    name: "res4b5"
    type: "Eltwise"
}

layer {
    bottom: "res4b5"
    top: "res4b5"
    name: "res4b5_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b5"
    top: "res4b6_branch2a"
    name: "res4b6_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b6_branch2a"
    top: "res4b6_branch2a"
    name: "bn4b6_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b6_branch2a"
    top: "res4b6_branch2a"
    name: "scale4b6_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b6_branch2a"
    bottom: "res4b6_branch2a"
    name: "res4b6_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b6_branch2a"
    top: "res4b6_branch2b"
    name: "res4b6_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b6_branch2b"
    top: "res4b6_branch2b"
    name: "bn4b6_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b6_branch2b"
    top: "res4b6_branch2b"
    name: "scale4b6_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b6_branch2b"
    bottom: "res4b6_branch2b"
    name: "res4b6_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b6_branch2b"
    top: "res4b6_branch2c"
    name: "res4b6_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b6_branch2c"
    top: "res4b6_branch2c"
    name: "bn4b6_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b6_branch2c"
    top: "res4b6_branch2c"
    name: "scale4b6_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b5"
    bottom: "res4b6_branch2c"
    top: "res4b6"
    name: "res4b6"
    type: "Eltwise"
}

layer {
    bottom: "res4b6"
    top: "res4b6"
    name: "res4b6_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b6"
    top: "res4b7_branch2a"
    name: "res4b7_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b7_branch2a"
    top: "res4b7_branch2a"
    name: "bn4b7_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b7_branch2a"
    top: "res4b7_branch2a"
    name: "scale4b7_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b7_branch2a"
    bottom: "res4b7_branch2a"
    name: "res4b7_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b7_branch2a"
    top: "res4b7_branch2b"
    name: "res4b7_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b7_branch2b"
    top: "res4b7_branch2b"
    name: "bn4b7_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b7_branch2b"
    top: "res4b7_branch2b"
    name: "scale4b7_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b7_branch2b"
    bottom: "res4b7_branch2b"
    name: "res4b7_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b7_branch2b"
    top: "res4b7_branch2c"
    name: "res4b7_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b7_branch2c"
    top: "res4b7_branch2c"
    name: "bn4b7_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b7_branch2c"
    top: "res4b7_branch2c"
    name: "scale4b7_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b6"
    bottom: "res4b7_branch2c"
    top: "res4b7"
    name: "res4b7"
    type: "Eltwise"
}

layer {
    bottom: "res4b7"
    top: "res4b7"
    name: "res4b7_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b7"
    top: "res4b8_branch2a"
    name: "res4b8_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b8_branch2a"
    top: "res4b8_branch2a"
    name: "bn4b8_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b8_branch2a"
    top: "res4b8_branch2a"
    name: "scale4b8_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b8_branch2a"
    bottom: "res4b8_branch2a"
    name: "res4b8_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b8_branch2a"
    top: "res4b8_branch2b"
    name: "res4b8_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b8_branch2b"
    top: "res4b8_branch2b"
    name: "bn4b8_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b8_branch2b"
    top: "res4b8_branch2b"
    name: "scale4b8_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b8_branch2b"
    bottom: "res4b8_branch2b"
    name: "res4b8_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b8_branch2b"
    top: "res4b8_branch2c"
    name: "res4b8_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b8_branch2c"
    top: "res4b8_branch2c"
    name: "bn4b8_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b8_branch2c"
    top: "res4b8_branch2c"
    name: "scale4b8_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b7"
    bottom: "res4b8_branch2c"
    top: "res4b8"
    name: "res4b8"
    type: "Eltwise"
}

layer {
    bottom: "res4b8"
    top: "res4b8"
    name: "res4b8_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b8"
    top: "res4b9_branch2a"
    name: "res4b9_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b9_branch2a"
    top: "res4b9_branch2a"
    name: "bn4b9_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b9_branch2a"
    top: "res4b9_branch2a"
    name: "scale4b9_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b9_branch2a"
    bottom: "res4b9_branch2a"
    name: "res4b9_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b9_branch2a"
    top: "res4b9_branch2b"
    name: "res4b9_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b9_branch2b"
    top: "res4b9_branch2b"
    name: "bn4b9_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b9_branch2b"
    top: "res4b9_branch2b"
    name: "scale4b9_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b9_branch2b"
    bottom: "res4b9_branch2b"
    name: "res4b9_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b9_branch2b"
    top: "res4b9_branch2c"
    name: "res4b9_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b9_branch2c"
    top: "res4b9_branch2c"
    name: "bn4b9_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b9_branch2c"
    top: "res4b9_branch2c"
    name: "scale4b9_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b8"
    bottom: "res4b9_branch2c"
    top: "res4b9"
    name: "res4b9"
    type: "Eltwise"
}

layer {
    bottom: "res4b9"
    top: "res4b9"
    name: "res4b9_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b9"
    top: "res4b10_branch2a"
    name: "res4b10_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b10_branch2a"
    top: "res4b10_branch2a"
    name: "bn4b10_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b10_branch2a"
    top: "res4b10_branch2a"
    name: "scale4b10_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b10_branch2a"
    bottom: "res4b10_branch2a"
    name: "res4b10_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b10_branch2a"
    top: "res4b10_branch2b"
    name: "res4b10_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b10_branch2b"
    top: "res4b10_branch2b"
    name: "bn4b10_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b10_branch2b"
    top: "res4b10_branch2b"
    name: "scale4b10_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b10_branch2b"
    bottom: "res4b10_branch2b"
    name: "res4b10_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b10_branch2b"
    top: "res4b10_branch2c"
    name: "res4b10_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b10_branch2c"
    top: "res4b10_branch2c"
    name: "bn4b10_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b10_branch2c"
    top: "res4b10_branch2c"
    name: "scale4b10_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b9"
    bottom: "res4b10_branch2c"
    top: "res4b10"
    name: "res4b10"
    type: "Eltwise"
}

layer {
    bottom: "res4b10"
    top: "res4b10"
    name: "res4b10_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b10"
    top: "res4b11_branch2a"
    name: "res4b11_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b11_branch2a"
    top: "res4b11_branch2a"
    name: "bn4b11_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b11_branch2a"
    top: "res4b11_branch2a"
    name: "scale4b11_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b11_branch2a"
    bottom: "res4b11_branch2a"
    name: "res4b11_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b11_branch2a"
    top: "res4b11_branch2b"
    name: "res4b11_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b11_branch2b"
    top: "res4b11_branch2b"
    name: "bn4b11_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b11_branch2b"
    top: "res4b11_branch2b"
    name: "scale4b11_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b11_branch2b"
    bottom: "res4b11_branch2b"
    name: "res4b11_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b11_branch2b"
    top: "res4b11_branch2c"
    name: "res4b11_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b11_branch2c"
    top: "res4b11_branch2c"
    name: "bn4b11_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b11_branch2c"
    top: "res4b11_branch2c"
    name: "scale4b11_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b10"
    bottom: "res4b11_branch2c"
    top: "res4b11"
    name: "res4b11"
    type: "Eltwise"
}

layer {
    bottom: "res4b11"
    top: "res4b11"
    name: "res4b11_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b11"
    top: "res4b12_branch2a"
    name: "res4b12_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b12_branch2a"
    top: "res4b12_branch2a"
    name: "bn4b12_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b12_branch2a"
    top: "res4b12_branch2a"
    name: "scale4b12_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b12_branch2a"
    bottom: "res4b12_branch2a"
    name: "res4b12_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b12_branch2a"
    top: "res4b12_branch2b"
    name: "res4b12_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b12_branch2b"
    top: "res4b12_branch2b"
    name: "bn4b12_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b12_branch2b"
    top: "res4b12_branch2b"
    name: "scale4b12_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b12_branch2b"
    bottom: "res4b12_branch2b"
    name: "res4b12_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b12_branch2b"
    top: "res4b12_branch2c"
    name: "res4b12_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b12_branch2c"
    top: "res4b12_branch2c"
    name: "bn4b12_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b12_branch2c"
    top: "res4b12_branch2c"
    name: "scale4b12_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b11"
    bottom: "res4b12_branch2c"
    top: "res4b12"
    name: "res4b12"
    type: "Eltwise"
}

layer {
    bottom: "res4b12"
    top: "res4b12"
    name: "res4b12_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b12"
    top: "res4b13_branch2a"
    name: "res4b13_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b13_branch2a"
    top: "res4b13_branch2a"
    name: "bn4b13_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b13_branch2a"
    top: "res4b13_branch2a"
    name: "scale4b13_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b13_branch2a"
    bottom: "res4b13_branch2a"
    name: "res4b13_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b13_branch2a"
    top: "res4b13_branch2b"
    name: "res4b13_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b13_branch2b"
    top: "res4b13_branch2b"
    name: "bn4b13_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b13_branch2b"
    top: "res4b13_branch2b"
    name: "scale4b13_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b13_branch2b"
    bottom: "res4b13_branch2b"
    name: "res4b13_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b13_branch2b"
    top: "res4b13_branch2c"
    name: "res4b13_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b13_branch2c"
    top: "res4b13_branch2c"
    name: "bn4b13_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b13_branch2c"
    top: "res4b13_branch2c"
    name: "scale4b13_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b12"
    bottom: "res4b13_branch2c"
    top: "res4b13"
    name: "res4b13"
    type: "Eltwise"
}

layer {
    bottom: "res4b13"
    top: "res4b13"
    name: "res4b13_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b13"
    top: "res4b14_branch2a"
    name: "res4b14_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b14_branch2a"
    top: "res4b14_branch2a"
    name: "bn4b14_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b14_branch2a"
    top: "res4b14_branch2a"
    name: "scale4b14_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b14_branch2a"
    bottom: "res4b14_branch2a"
    name: "res4b14_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b14_branch2a"
    top: "res4b14_branch2b"
    name: "res4b14_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b14_branch2b"
    top: "res4b14_branch2b"
    name: "bn4b14_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b14_branch2b"
    top: "res4b14_branch2b"
    name: "scale4b14_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b14_branch2b"
    bottom: "res4b14_branch2b"
    name: "res4b14_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b14_branch2b"
    top: "res4b14_branch2c"
    name: "res4b14_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b14_branch2c"
    top: "res4b14_branch2c"
    name: "bn4b14_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b14_branch2c"
    top: "res4b14_branch2c"
    name: "scale4b14_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b13"
    bottom: "res4b14_branch2c"
    top: "res4b14"
    name: "res4b14"
    type: "Eltwise"
}

layer {
    bottom: "res4b14"
    top: "res4b14"
    name: "res4b14_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b14"
    top: "res4b15_branch2a"
    name: "res4b15_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b15_branch2a"
    top: "res4b15_branch2a"
    name: "bn4b15_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b15_branch2a"
    top: "res4b15_branch2a"
    name: "scale4b15_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b15_branch2a"
    bottom: "res4b15_branch2a"
    name: "res4b15_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b15_branch2a"
    top: "res4b15_branch2b"
    name: "res4b15_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b15_branch2b"
    top: "res4b15_branch2b"
    name: "bn4b15_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b15_branch2b"
    top: "res4b15_branch2b"
    name: "scale4b15_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b15_branch2b"
    bottom: "res4b15_branch2b"
    name: "res4b15_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b15_branch2b"
    top: "res4b15_branch2c"
    name: "res4b15_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b15_branch2c"
    top: "res4b15_branch2c"
    name: "bn4b15_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b15_branch2c"
    top: "res4b15_branch2c"
    name: "scale4b15_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b14"
    bottom: "res4b15_branch2c"
    top: "res4b15"
    name: "res4b15"
    type: "Eltwise"
}

layer {
    bottom: "res4b15"
    top: "res4b15"
    name: "res4b15_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b15"
    top: "res4b16_branch2a"
    name: "res4b16_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b16_branch2a"
    top: "res4b16_branch2a"
    name: "bn4b16_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b16_branch2a"
    top: "res4b16_branch2a"
    name: "scale4b16_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b16_branch2a"
    bottom: "res4b16_branch2a"
    name: "res4b16_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b16_branch2a"
    top: "res4b16_branch2b"
    name: "res4b16_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b16_branch2b"
    top: "res4b16_branch2b"
    name: "bn4b16_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b16_branch2b"
    top: "res4b16_branch2b"
    name: "scale4b16_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b16_branch2b"
    bottom: "res4b16_branch2b"
    name: "res4b16_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b16_branch2b"
    top: "res4b16_branch2c"
    name: "res4b16_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b16_branch2c"
    top: "res4b16_branch2c"
    name: "bn4b16_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b16_branch2c"
    top: "res4b16_branch2c"
    name: "scale4b16_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b15"
    bottom: "res4b16_branch2c"
    top: "res4b16"
    name: "res4b16"
    type: "Eltwise"
}

layer {
    bottom: "res4b16"
    top: "res4b16"
    name: "res4b16_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b16"
    top: "res4b17_branch2a"
    name: "res4b17_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b17_branch2a"
    top: "res4b17_branch2a"
    name: "bn4b17_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b17_branch2a"
    top: "res4b17_branch2a"
    name: "scale4b17_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b17_branch2a"
    bottom: "res4b17_branch2a"
    name: "res4b17_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b17_branch2a"
    top: "res4b17_branch2b"
    name: "res4b17_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b17_branch2b"
    top: "res4b17_branch2b"
    name: "bn4b17_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b17_branch2b"
    top: "res4b17_branch2b"
    name: "scale4b17_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b17_branch2b"
    bottom: "res4b17_branch2b"
    name: "res4b17_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b17_branch2b"
    top: "res4b17_branch2c"
    name: "res4b17_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b17_branch2c"
    top: "res4b17_branch2c"
    name: "bn4b17_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b17_branch2c"
    top: "res4b17_branch2c"
    name: "scale4b17_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b16"
    bottom: "res4b17_branch2c"
    top: "res4b17"
    name: "res4b17"
    type: "Eltwise"
}

layer {
    bottom: "res4b17"
    top: "res4b17"
    name: "res4b17_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b17"
    top: "res4b18_branch2a"
    name: "res4b18_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b18_branch2a"
    top: "res4b18_branch2a"
    name: "bn4b18_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b18_branch2a"
    top: "res4b18_branch2a"
    name: "scale4b18_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b18_branch2a"
    bottom: "res4b18_branch2a"
    name: "res4b18_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b18_branch2a"
    top: "res4b18_branch2b"
    name: "res4b18_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b18_branch2b"
    top: "res4b18_branch2b"
    name: "bn4b18_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b18_branch2b"
    top: "res4b18_branch2b"
    name: "scale4b18_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b18_branch2b"
    bottom: "res4b18_branch2b"
    name: "res4b18_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b18_branch2b"
    top: "res4b18_branch2c"
    name: "res4b18_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b18_branch2c"
    top: "res4b18_branch2c"
    name: "bn4b18_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b18_branch2c"
    top: "res4b18_branch2c"
    name: "scale4b18_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b17"
    bottom: "res4b18_branch2c"
    top: "res4b18"
    name: "res4b18"
    type: "Eltwise"
}

layer {
    bottom: "res4b18"
    top: "res4b18"
    name: "res4b18_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b18"
    top: "res4b19_branch2a"
    name: "res4b19_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b19_branch2a"
    top: "res4b19_branch2a"
    name: "bn4b19_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b19_branch2a"
    top: "res4b19_branch2a"
    name: "scale4b19_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b19_branch2a"
    bottom: "res4b19_branch2a"
    name: "res4b19_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b19_branch2a"
    top: "res4b19_branch2b"
    name: "res4b19_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b19_branch2b"
    top: "res4b19_branch2b"
    name: "bn4b19_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b19_branch2b"
    top: "res4b19_branch2b"
    name: "scale4b19_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b19_branch2b"
    bottom: "res4b19_branch2b"
    name: "res4b19_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b19_branch2b"
    top: "res4b19_branch2c"
    name: "res4b19_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b19_branch2c"
    top: "res4b19_branch2c"
    name: "bn4b19_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b19_branch2c"
    top: "res4b19_branch2c"
    name: "scale4b19_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b18"
    bottom: "res4b19_branch2c"
    top: "res4b19"
    name: "res4b19"
    type: "Eltwise"
}

layer {
    bottom: "res4b19"
    top: "res4b19"
    name: "res4b19_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b19"
    top: "res4b20_branch2a"
    name: "res4b20_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b20_branch2a"
    top: "res4b20_branch2a"
    name: "bn4b20_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b20_branch2a"
    top: "res4b20_branch2a"
    name: "scale4b20_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b20_branch2a"
    bottom: "res4b20_branch2a"
    name: "res4b20_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b20_branch2a"
    top: "res4b20_branch2b"
    name: "res4b20_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b20_branch2b"
    top: "res4b20_branch2b"
    name: "bn4b20_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b20_branch2b"
    top: "res4b20_branch2b"
    name: "scale4b20_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b20_branch2b"
    bottom: "res4b20_branch2b"
    name: "res4b20_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b20_branch2b"
    top: "res4b20_branch2c"
    name: "res4b20_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b20_branch2c"
    top: "res4b20_branch2c"
    name: "bn4b20_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b20_branch2c"
    top: "res4b20_branch2c"
    name: "scale4b20_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b19"
    bottom: "res4b20_branch2c"
    top: "res4b20"
    name: "res4b20"
    type: "Eltwise"
}

layer {
    bottom: "res4b20"
    top: "res4b20"
    name: "res4b20_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b20"
    top: "res4b21_branch2a"
    name: "res4b21_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b21_branch2a"
    top: "res4b21_branch2a"
    name: "bn4b21_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b21_branch2a"
    top: "res4b21_branch2a"
    name: "scale4b21_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b21_branch2a"
    bottom: "res4b21_branch2a"
    name: "res4b21_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b21_branch2a"
    top: "res4b21_branch2b"
    name: "res4b21_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b21_branch2b"
    top: "res4b21_branch2b"
    name: "bn4b21_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b21_branch2b"
    top: "res4b21_branch2b"
    name: "scale4b21_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b21_branch2b"
    bottom: "res4b21_branch2b"
    name: "res4b21_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b21_branch2b"
    top: "res4b21_branch2c"
    name: "res4b21_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b21_branch2c"
    top: "res4b21_branch2c"
    name: "bn4b21_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b21_branch2c"
    top: "res4b21_branch2c"
    name: "scale4b21_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b20"
    bottom: "res4b21_branch2c"
    top: "res4b21"
    name: "res4b21"
    type: "Eltwise"
}

layer {
    bottom: "res4b21"
    top: "res4b21"
    name: "res4b21_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b21"
    top: "res4b22_branch2a"
    name: "res4b22_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b22_branch2a"
    top: "res4b22_branch2a"
    name: "bn4b22_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b22_branch2a"
    top: "res4b22_branch2a"
    name: "scale4b22_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b22_branch2a"
    bottom: "res4b22_branch2a"
    name: "res4b22_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b22_branch2a"
    top: "res4b22_branch2b"
    name: "res4b22_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b22_branch2b"
    top: "res4b22_branch2b"
    name: "bn4b22_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b22_branch2b"
    top: "res4b22_branch2b"
    name: "scale4b22_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b22_branch2b"
    bottom: "res4b22_branch2b"
    name: "res4b22_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b22_branch2b"
    top: "res4b22_branch2c"
    name: "res4b22_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b22_branch2c"
    top: "res4b22_branch2c"
    name: "bn4b22_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b22_branch2c"
    top: "res4b22_branch2c"
    name: "scale4b22_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b21"
    bottom: "res4b22_branch2c"
    top: "res4b22"
    name: "res4b22"
    type: "Eltwise"
}

layer {
    bottom: "res4b22"
    top: "res4b22"
    name: "res4b22_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b22"
    top: "res4b23_branch2a"
    name: "res4b23_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b23_branch2a"
    top: "res4b23_branch2a"
    name: "bn4b23_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b23_branch2a"
    top: "res4b23_branch2a"
    name: "scale4b23_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b23_branch2a"
    bottom: "res4b23_branch2a"
    name: "res4b23_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b23_branch2a"
    top: "res4b23_branch2b"
    name: "res4b23_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b23_branch2b"
    top: "res4b23_branch2b"
    name: "bn4b23_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b23_branch2b"
    top: "res4b23_branch2b"
    name: "scale4b23_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b23_branch2b"
    bottom: "res4b23_branch2b"
    name: "res4b23_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b23_branch2b"
    top: "res4b23_branch2c"
    name: "res4b23_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b23_branch2c"
    top: "res4b23_branch2c"
    name: "bn4b23_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b23_branch2c"
    top: "res4b23_branch2c"
    name: "scale4b23_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b22"
    bottom: "res4b23_branch2c"
    top: "res4b23"
    name: "res4b23"
    type: "Eltwise"
}

layer {
    bottom: "res4b23"
    top: "res4b23"
    name: "res4b23_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b23"
    top: "res4b24_branch2a"
    name: "res4b24_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b24_branch2a"
    top: "res4b24_branch2a"
    name: "bn4b24_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b24_branch2a"
    top: "res4b24_branch2a"
    name: "scale4b24_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b24_branch2a"
    bottom: "res4b24_branch2a"
    name: "res4b24_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b24_branch2a"
    top: "res4b24_branch2b"
    name: "res4b24_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b24_branch2b"
    top: "res4b24_branch2b"
    name: "bn4b24_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b24_branch2b"
    top: "res4b24_branch2b"
    name: "scale4b24_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b24_branch2b"
    bottom: "res4b24_branch2b"
    name: "res4b24_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b24_branch2b"
    top: "res4b24_branch2c"
    name: "res4b24_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b24_branch2c"
    top: "res4b24_branch2c"
    name: "bn4b24_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b24_branch2c"
    top: "res4b24_branch2c"
    name: "scale4b24_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b23"
    bottom: "res4b24_branch2c"
    top: "res4b24"
    name: "res4b24"
    type: "Eltwise"
}

layer {
    bottom: "res4b24"
    top: "res4b24"
    name: "res4b24_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b24"
    top: "res4b25_branch2a"
    name: "res4b25_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b25_branch2a"
    top: "res4b25_branch2a"
    name: "bn4b25_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b25_branch2a"
    top: "res4b25_branch2a"
    name: "scale4b25_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b25_branch2a"
    bottom: "res4b25_branch2a"
    name: "res4b25_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b25_branch2a"
    top: "res4b25_branch2b"
    name: "res4b25_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b25_branch2b"
    top: "res4b25_branch2b"
    name: "bn4b25_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b25_branch2b"
    top: "res4b25_branch2b"
    name: "scale4b25_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b25_branch2b"
    bottom: "res4b25_branch2b"
    name: "res4b25_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b25_branch2b"
    top: "res4b25_branch2c"
    name: "res4b25_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b25_branch2c"
    top: "res4b25_branch2c"
    name: "bn4b25_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b25_branch2c"
    top: "res4b25_branch2c"
    name: "scale4b25_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b24"
    bottom: "res4b25_branch2c"
    top: "res4b25"
    name: "res4b25"
    type: "Eltwise"
}

layer {
    bottom: "res4b25"
    top: "res4b25"
    name: "res4b25_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b25"
    top: "res4b26_branch2a"
    name: "res4b26_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b26_branch2a"
    top: "res4b26_branch2a"
    name: "bn4b26_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b26_branch2a"
    top: "res4b26_branch2a"
    name: "scale4b26_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b26_branch2a"
    bottom: "res4b26_branch2a"
    name: "res4b26_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b26_branch2a"
    top: "res4b26_branch2b"
    name: "res4b26_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b26_branch2b"
    top: "res4b26_branch2b"
    name: "bn4b26_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b26_branch2b"
    top: "res4b26_branch2b"
    name: "scale4b26_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b26_branch2b"
    bottom: "res4b26_branch2b"
    name: "res4b26_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b26_branch2b"
    top: "res4b26_branch2c"
    name: "res4b26_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b26_branch2c"
    top: "res4b26_branch2c"
    name: "bn4b26_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b26_branch2c"
    top: "res4b26_branch2c"
    name: "scale4b26_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b25"
    bottom: "res4b26_branch2c"
    top: "res4b26"
    name: "res4b26"
    type: "Eltwise"
}

layer {
    bottom: "res4b26"
    top: "res4b26"
    name: "res4b26_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b26"
    top: "res4b27_branch2a"
    name: "res4b27_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b27_branch2a"
    top: "res4b27_branch2a"
    name: "bn4b27_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b27_branch2a"
    top: "res4b27_branch2a"
    name: "scale4b27_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b27_branch2a"
    bottom: "res4b27_branch2a"
    name: "res4b27_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b27_branch2a"
    top: "res4b27_branch2b"
    name: "res4b27_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b27_branch2b"
    top: "res4b27_branch2b"
    name: "bn4b27_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b27_branch2b"
    top: "res4b27_branch2b"
    name: "scale4b27_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b27_branch2b"
    bottom: "res4b27_branch2b"
    name: "res4b27_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b27_branch2b"
    top: "res4b27_branch2c"
    name: "res4b27_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b27_branch2c"
    top: "res4b27_branch2c"
    name: "bn4b27_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b27_branch2c"
    top: "res4b27_branch2c"
    name: "scale4b27_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b26"
    bottom: "res4b27_branch2c"
    top: "res4b27"
    name: "res4b27"
    type: "Eltwise"
}

layer {
    bottom: "res4b27"
    top: "res4b27"
    name: "res4b27_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b27"
    top: "res4b28_branch2a"
    name: "res4b28_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b28_branch2a"
    top: "res4b28_branch2a"
    name: "bn4b28_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b28_branch2a"
    top: "res4b28_branch2a"
    name: "scale4b28_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b28_branch2a"
    bottom: "res4b28_branch2a"
    name: "res4b28_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b28_branch2a"
    top: "res4b28_branch2b"
    name: "res4b28_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b28_branch2b"
    top: "res4b28_branch2b"
    name: "bn4b28_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b28_branch2b"
    top: "res4b28_branch2b"
    name: "scale4b28_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b28_branch2b"
    bottom: "res4b28_branch2b"
    name: "res4b28_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b28_branch2b"
    top: "res4b28_branch2c"
    name: "res4b28_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b28_branch2c"
    top: "res4b28_branch2c"
    name: "bn4b28_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b28_branch2c"
    top: "res4b28_branch2c"
    name: "scale4b28_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b27"
    bottom: "res4b28_branch2c"
    top: "res4b28"
    name: "res4b28"
    type: "Eltwise"
}

layer {
    bottom: "res4b28"
    top: "res4b28"
    name: "res4b28_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b28"
    top: "res4b29_branch2a"
    name: "res4b29_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b29_branch2a"
    top: "res4b29_branch2a"
    name: "bn4b29_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b29_branch2a"
    top: "res4b29_branch2a"
    name: "scale4b29_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b29_branch2a"
    bottom: "res4b29_branch2a"
    name: "res4b29_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b29_branch2a"
    top: "res4b29_branch2b"
    name: "res4b29_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b29_branch2b"
    top: "res4b29_branch2b"
    name: "bn4b29_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b29_branch2b"
    top: "res4b29_branch2b"
    name: "scale4b29_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b29_branch2b"
    bottom: "res4b29_branch2b"
    name: "res4b29_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b29_branch2b"
    top: "res4b29_branch2c"
    name: "res4b29_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b29_branch2c"
    top: "res4b29_branch2c"
    name: "bn4b29_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b29_branch2c"
    top: "res4b29_branch2c"
    name: "scale4b29_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b28"
    bottom: "res4b29_branch2c"
    top: "res4b29"
    name: "res4b29"
    type: "Eltwise"
}

layer {
    bottom: "res4b29"
    top: "res4b29"
    name: "res4b29_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b29"
    top: "res4b30_branch2a"
    name: "res4b30_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b30_branch2a"
    top: "res4b30_branch2a"
    name: "bn4b30_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b30_branch2a"
    top: "res4b30_branch2a"
    name: "scale4b30_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b30_branch2a"
    bottom: "res4b30_branch2a"
    name: "res4b30_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b30_branch2a"
    top: "res4b30_branch2b"
    name: "res4b30_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b30_branch2b"
    top: "res4b30_branch2b"
    name: "bn4b30_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b30_branch2b"
    top: "res4b30_branch2b"
    name: "scale4b30_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b30_branch2b"
    bottom: "res4b30_branch2b"
    name: "res4b30_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b30_branch2b"
    top: "res4b30_branch2c"
    name: "res4b30_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b30_branch2c"
    top: "res4b30_branch2c"
    name: "bn4b30_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b30_branch2c"
    top: "res4b30_branch2c"
    name: "scale4b30_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b29"
    bottom: "res4b30_branch2c"
    top: "res4b30"
    name: "res4b30"
    type: "Eltwise"
}

layer {
    bottom: "res4b30"
    top: "res4b30"
    name: "res4b30_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b30"
    top: "res4b31_branch2a"
    name: "res4b31_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b31_branch2a"
    top: "res4b31_branch2a"
    name: "bn4b31_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b31_branch2a"
    top: "res4b31_branch2a"
    name: "scale4b31_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b31_branch2a"
    bottom: "res4b31_branch2a"
    name: "res4b31_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b31_branch2a"
    top: "res4b31_branch2b"
    name: "res4b31_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b31_branch2b"
    top: "res4b31_branch2b"
    name: "bn4b31_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b31_branch2b"
    top: "res4b31_branch2b"
    name: "scale4b31_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b31_branch2b"
    bottom: "res4b31_branch2b"
    name: "res4b31_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b31_branch2b"
    top: "res4b31_branch2c"
    name: "res4b31_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b31_branch2c"
    top: "res4b31_branch2c"
    name: "bn4b31_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b31_branch2c"
    top: "res4b31_branch2c"
    name: "scale4b31_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b30"
    bottom: "res4b31_branch2c"
    top: "res4b31"
    name: "res4b31"
    type: "Eltwise"
}

layer {
    bottom: "res4b31"
    top: "res4b31"
    name: "res4b31_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b31"
    top: "res4b32_branch2a"
    name: "res4b32_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b32_branch2a"
    top: "res4b32_branch2a"
    name: "bn4b32_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b32_branch2a"
    top: "res4b32_branch2a"
    name: "scale4b32_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b32_branch2a"
    bottom: "res4b32_branch2a"
    name: "res4b32_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b32_branch2a"
    top: "res4b32_branch2b"
    name: "res4b32_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b32_branch2b"
    top: "res4b32_branch2b"
    name: "bn4b32_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b32_branch2b"
    top: "res4b32_branch2b"
    name: "scale4b32_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b32_branch2b"
    bottom: "res4b32_branch2b"
    name: "res4b32_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b32_branch2b"
    top: "res4b32_branch2c"
    name: "res4b32_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b32_branch2c"
    top: "res4b32_branch2c"
    name: "bn4b32_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b32_branch2c"
    top: "res4b32_branch2c"
    name: "scale4b32_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b31"
    bottom: "res4b32_branch2c"
    top: "res4b32"
    name: "res4b32"
    type: "Eltwise"
}

layer {
    bottom: "res4b32"
    top: "res4b32"
    name: "res4b32_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b32"
    top: "res4b33_branch2a"
    name: "res4b33_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b33_branch2a"
    top: "res4b33_branch2a"
    name: "bn4b33_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b33_branch2a"
    top: "res4b33_branch2a"
    name: "scale4b33_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b33_branch2a"
    bottom: "res4b33_branch2a"
    name: "res4b33_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b33_branch2a"
    top: "res4b33_branch2b"
    name: "res4b33_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b33_branch2b"
    top: "res4b33_branch2b"
    name: "bn4b33_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b33_branch2b"
    top: "res4b33_branch2b"
    name: "scale4b33_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b33_branch2b"
    bottom: "res4b33_branch2b"
    name: "res4b33_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b33_branch2b"
    top: "res4b33_branch2c"
    name: "res4b33_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b33_branch2c"
    top: "res4b33_branch2c"
    name: "bn4b33_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b33_branch2c"
    top: "res4b33_branch2c"
    name: "scale4b33_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b32"
    bottom: "res4b33_branch2c"
    top: "res4b33"
    name: "res4b33"
    type: "Eltwise"
}

layer {
    bottom: "res4b33"
    top: "res4b33"
    name: "res4b33_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b33"
    top: "res4b34_branch2a"
    name: "res4b34_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b34_branch2a"
    top: "res4b34_branch2a"
    name: "bn4b34_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b34_branch2a"
    top: "res4b34_branch2a"
    name: "scale4b34_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b34_branch2a"
    bottom: "res4b34_branch2a"
    name: "res4b34_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b34_branch2a"
    top: "res4b34_branch2b"
    name: "res4b34_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b34_branch2b"
    top: "res4b34_branch2b"
    name: "bn4b34_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b34_branch2b"
    top: "res4b34_branch2b"
    name: "scale4b34_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b34_branch2b"
    bottom: "res4b34_branch2b"
    name: "res4b34_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b34_branch2b"
    top: "res4b34_branch2c"
    name: "res4b34_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b34_branch2c"
    top: "res4b34_branch2c"
    name: "bn4b34_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b34_branch2c"
    top: "res4b34_branch2c"
    name: "scale4b34_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b33"
    bottom: "res4b34_branch2c"
    top: "res4b34"
    name: "res4b34"
    type: "Eltwise"
}

layer {
    bottom: "res4b34"
    top: "res4b34"
    name: "res4b34_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b34"
    top: "res4b35_branch2a"
    name: "res4b35_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b35_branch2a"
    top: "res4b35_branch2a"
    name: "bn4b35_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b35_branch2a"
    top: "res4b35_branch2a"
    name: "scale4b35_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b35_branch2a"
    bottom: "res4b35_branch2a"
    name: "res4b35_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b35_branch2a"
    top: "res4b35_branch2b"
    name: "res4b35_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 256
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b35_branch2b"
    top: "res4b35_branch2b"
    name: "bn4b35_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b35_branch2b"
    top: "res4b35_branch2b"
    name: "scale4b35_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res4b35_branch2b"
    bottom: "res4b35_branch2b"
    name: "res4b35_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b35_branch2b"
    top: "res4b35_branch2c"
    name: "res4b35_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 1024
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res4b35_branch2c"
    top: "res4b35_branch2c"
    name: "bn4b35_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res4b35_branch2c"
    top: "res4b35_branch2c"
    name: "scale4b35_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b34"
    bottom: "res4b35_branch2c"
    top: "res4b35"
    name: "res4b35"
    type: "Eltwise"
}

layer {
    bottom: "res4b35"
    top: "res4b35"
    name: "res4b35_relu"
    type: "ReLU"
}

layer {
    bottom: "res4b35"
    top: "res5a_branch1"
    name: "res5a_branch1"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 2
        bias_term: false
    }
}

layer {
    bottom: "res5a_branch1"
    top: "res5a_branch1"
    name: "bn5a_branch1"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5a_branch1"
    top: "res5a_branch1"
    name: "scale5a_branch1"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res4b35"
    top: "res5a_branch2a"
    name: "res5a_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 2
        bias_term: false
    }
}

layer {
    bottom: "res5a_branch2a"
    top: "res5a_branch2a"
    name: "bn5a_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5a_branch2a"
    top: "res5a_branch2a"
    name: "scale5a_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res5a_branch2a"
    bottom: "res5a_branch2a"
    name: "res5a_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5a_branch2a"
    top: "res5a_branch2b"
    name: "res5a_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5a_branch2b"
    top: "res5a_branch2b"
    name: "bn5a_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5a_branch2b"
    top: "res5a_branch2b"
    name: "scale5a_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res5a_branch2b"
    bottom: "res5a_branch2b"
    name: "res5a_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5a_branch2b"
    top: "res5a_branch2c"
    name: "res5a_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5a_branch2c"
    top: "res5a_branch2c"
    name: "bn5a_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5a_branch2c"
    top: "res5a_branch2c"
    name: "scale5a_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5a_branch1"
    bottom: "res5a_branch2c"
    top: "res5a"
    name: "res5a"
    type: "Eltwise"
}

layer {
    bottom: "res5a"
    top: "res5a"
    name: "res5a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5a"
    top: "res5b_branch2a"
    name: "res5b_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5b_branch2a"
    top: "res5b_branch2a"
    name: "bn5b_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5b_branch2a"
    top: "res5b_branch2a"
    name: "scale5b_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res5b_branch2a"
    bottom: "res5b_branch2a"
    name: "res5b_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5b_branch2a"
    top: "res5b_branch2b"
    name: "res5b_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5b_branch2b"
    top: "res5b_branch2b"
    name: "bn5b_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5b_branch2b"
    top: "res5b_branch2b"
    name: "scale5b_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res5b_branch2b"
    bottom: "res5b_branch2b"
    name: "res5b_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5b_branch2b"
    top: "res5b_branch2c"
    name: "res5b_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5b_branch2c"
    top: "res5b_branch2c"
    name: "bn5b_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5b_branch2c"
    top: "res5b_branch2c"
    name: "scale5b_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5a"
    bottom: "res5b_branch2c"
    top: "res5b"
    name: "res5b"
    type: "Eltwise"
}

layer {
    bottom: "res5b"
    top: "res5b"
    name: "res5b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5b"
    top: "res5c_branch2a"
    name: "res5c_branch2a"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5c_branch2a"
    top: "res5c_branch2a"
    name: "bn5c_branch2a"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5c_branch2a"
    top: "res5c_branch2a"
    name: "scale5c_branch2a"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res5c_branch2a"
    bottom: "res5c_branch2a"
    name: "res5c_branch2a_relu"
    type: "ReLU"
}

layer {
    bottom: "res5c_branch2a"
    top: "res5c_branch2b"
    name: "res5c_branch2b"
    type: "Convolution"
    convolution_param {
        num_output: 512
        kernel_size: 3
        pad: 1
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5c_branch2b"
    top: "res5c_branch2b"
    name: "bn5c_branch2b"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5c_branch2b"
    top: "res5c_branch2b"
    name: "scale5c_branch2b"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    top: "res5c_branch2b"
    bottom: "res5c_branch2b"
    name: "res5c_branch2b_relu"
    type: "ReLU"
}

layer {
    bottom: "res5c_branch2b"
    top: "res5c_branch2c"
    name: "res5c_branch2c"
    type: "Convolution"
    convolution_param {
        num_output: 2048
        kernel_size: 1
        pad: 0
        stride: 1
        bias_term: false
    }
}

layer {
    bottom: "res5c_branch2c"
    top: "res5c_branch2c"
    name: "bn5c_branch2c"
    type: "BatchNorm"
    batch_norm_param {
        use_global_stats: true
    }
}

layer {
    bottom: "res5c_branch2c"
    top: "res5c_branch2c"
    name: "scale5c_branch2c"
    type: "Scale"
    scale_param {
        bias_term: true
    }
}

layer {
    bottom: "res5b"
    bottom: "res5c_branch2c"
    top: "res5c"
    name: "res5c"
    type: "Eltwise"
}

layer {
    bottom: "res5c"
    top: "res5c"
    name: "res5c_relu"
    type: "ReLU"
}

layer {
    bottom: "res5c"
    top: "pool5"
    name: "pool5"
    type: "Pooling"
    pooling_param {
        kernel_size: 7
        stride: 1
        pool: AVE
    }
}

layer {
    bottom: "pool5"
    top: "fc3"
    name: "fc3"
    type: "InnerProduct"
    inner_product_param {
        num_output: 3
    }
}

layer {
    bottom: "fc3"
    top: "prob"
    name: "prob"
    type: "Softmax"
}

   

      solver.prototxt

net: "/home/wy/ResNet152/train_val.prototxt"
test_iter: 1003
test_interval: 4000
test_initialization: false
display: 100
average_loss: 100
base_lr: 0.05
lr_policy: "step"
stepsize: 150000
gamma: 0.1
max_iter: 600000
momentum: 0.9
weight_decay: 0.0001
snapshot: 40000
snapshot_prefix: "/home/wy/ResNet152/model/"
solver_mode: GPU

作为一枚技术小白，写这篇笔记的时候参考了很多博客论文，在这里表示感谢，同时，未经同意，请勿转载....