首先总结前一节的内容。
简单的讲，就是训练并测试了快速解决方案。
转换数据格式：
convert_cifar_data.exe data/cifar10 examples/cifar10 lmdb
计算平均值
compute_image_mean.exe -backend=lmdb examples/cifar10/cifar10_train_lmdb examples/cifar10/mean.binaryproto
训练网络--快速解决方案
caffe train --solver=examples/cifar10/cifar10_quick_solver.prototxt
继续训练网络--再多迭代1000次
caffe train --solver=examples/cifar10/cifar10_quick_solver_lr1.prototxt --snapshot=examples/cifar10/cifar10_quick_iter_4000.solverstate
测试模型的准确率
caffe test -model examples/cifar10/cifar10_quick_train_test.prototxt -weights examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5 -iterations 100
识别一只猫
classification.exe examples/cifar10/cifar10_quick.prototxt examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5 examples/cifar10/mean.binaryproto data/cifar10/synset_words.txt examples/images/cat.jpg
识别人鱼自行车
classification.exe examples/cifar10/cifar10_quick.prototxt examples/cifar10/cifar10_quick_iter_5000.caffemodel.h5 examples/cifar10/mean.binaryproto data/cifar10/synset_words.txt examples/images/fish-bike
===========================================华丽的分割线=============================================
现在开始 尝试 完整解决方案 cifar10_full_solver.prototxt
1. 训练
caffe train --solver=examples/cifar10/cifar10_full_solver.prototxt
原来以为半个小时能跑完的，结果发现，40分钟只跑了10000次迭代。总数是6万。所以，跑了很久。
好在最后还是跑完了。以后事先要先预估一下计算量才好，不然，跑了也是白跑。
2.测试
测试：做10次迭代的结果
caffe test -model examples/cifar10/cifar10_full_train_test.prototxt -weights examples/cifar10
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$ caffe test -model examples/cifar10/cifar10_full_train_test.prototxt -weights examples/cifar10/cifar10_full_iter_60000.caffemodel.h5 -iterations 10
------------------------------------------------------------------------------
I0704 12:26:43.571471 5465 caffe.cpp:284] Use CPU.
I0704 12:26:45.302641 5465 net.cpp:294] The NetState phase (1) differed from the phase (0) specified by a rule in layer cifar
I0704 12:26:45.302878 5465 net.cpp:51] Initializing net from parameters:
name: "CIFAR10_full"
state {
phase: TEST
level: 0
stage: ""
}
layer {
name: "cifar"
type: "Data"
top: "data"
top: "label"
include {
phase: TEST
}
transform_param {
mean_file: "examples/cifar10/mean.binaryproto"
}
data_param {
source: "examples/cifar10/cifar10_test_lmdb"
batch_size: 100
backend: LMDB
}
}
layer {
name: "conv1"
type: "Convolution"
bottom: "data"
top: "conv1"
param {
lr_mult: 1
}
param {
lr_mult: 2
}
convolution_param {
num_output: 32
pad: 2
kernel_size: 5
stride: 1
weight_filler {
type: "gaussian"
std: 0.0001
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "pool1"
type: "Pooling"
bottom: "conv1"
top: "pool1"
pooling_param {
pool: MAX
kernel_size: 3
stride: 2
}
}
layer {
name: "relu1"
type: "ReLU"
bottom: "pool1"
top: "pool1"
}
layer {
name: "norm1"
type: "LRN"
bottom: "pool1"
top: "norm1"
lrn_param {
local_size: 3
alpha: 5e-05
beta: 0.75
norm_region: WITHIN_CHANNEL
}
}
layer {
name: "conv2"
type: "Convolution"
bottom: "norm1"
top: "conv2"
param {
lr_mult: 1
}
param {
lr_mult: 2
}
convolution_param {
num_output: 32
pad: 2
kernel_size: 5
stride: 1
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "relu2"
type: "ReLU"
bottom: "conv2"
top: "conv2"
}
layer {
name: "pool2"
type: "Pooling"
bottom: "conv2"
top: "pool2"
pooling_param {
pool: AVE
kernel_size: 3
stride: 2
}
}
layer {
name: "norm2"
type: "LRN"
bottom: "pool2"
top: "norm2"
lrn_param {
local_size: 3
alpha: 5e-05
beta: 0.75
norm_region: WITHIN_CHANNEL
}
}
layer {
name: "conv3"
type: "Convolution"
bottom: "norm2"
top: "conv3"
convolution_param {
num_output: 64
pad: 2
kernel_size: 5
stride: 1
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "relu3"
type: "ReLU"
bottom: "conv3"
top: "conv3"
}
layer {
name: "pool3"
type: "Pooling"
bottom: "conv3"
top: "pool3"
pooling_param {
pool: AVE
kernel_size: 3
stride: 2
}
}
layer {
name: "ip1"
type: "InnerProduct"
bottom: "pool3"
top: "ip1"
param {
lr_mult: 1
decay_mult: 250
}
param {
lr_mult: 2
decay_mult: 0
}
inner_product_param {
num_output: 10
weight_filler {
type: "gaussian"
std: 0.01
}
bias_filler {
type: "constant"
}
}
}
layer {
name: "accuracy"
type: "Accuracy"
bottom: "ip1"
bottom: "label"
top: "accuracy"
include {
phase: TEST
}
}
layer {
name: "loss"
type: "SoftmaxWithLoss"
bottom: "ip1"
bottom: "label"
top: "loss"
}
I0704 12:26:45.329610 5465 layer_factory.hpp:77] Creating layer cifar
I0704 12:26:45.329771 5465 db_lmdb.cpp:35] Opened lmdb examples/cifar10/cifar10_test_lmdb
I0704 12:26:45.329805 5465 net.cpp:84] Creating Layer cifar
I0704 12:26:45.329823 5465 net.cpp:380] cifar -> data
I0704 12:26:45.329852 5465 net.cpp:380] cifar -> label
I0704 12:26:45.329874 5465 data_transformer.cpp:25] Loading mean file from: examples/cifar10/mean.binaryproto
I0704 12:26:45.329970 5465 data_layer.cpp:45] output data size: 100,3,32,32
I0704 12:26:45.342496 5465 net.cpp:122] Setting up cifar
I0704 12:26:45.342577 5465 net.cpp:129] Top shape: 100 3 32 32 (307200)
I0704 12:26:45.342600 5465 net.cpp:129] Top shape: 100 (100)
I0704 12:26:45.342612 5465 net.cpp:137] Memory required for data: 1229200
I0704 12:26:45.342634 5465 layer_factory.hpp:77] Creating layer label_cifar_1_split
I0704 12:26:45.352715 5465 net.cpp:84] Creating Layer label_cifar_1_split
I0704 12:26:45.352736 5465 net.cpp:406] label_cifar_1_split <- label
I0704 12:26:45.352809 5465 net.cpp:380] label_cifar_1_split -> label_cifar_1_split_0
I0704 12:26:45.352833 5465 net.cpp:380] label_cifar_1_split -> label_cifar_1_split_1
I0704 12:26:45.352854 5465 net.cpp:122] Setting up label_cifar_1_split
I0704 12:26:45.352866 5465 net.cpp:129] Top shape: 100 (100)
I0704 12:26:45.352890 5465 net.cpp:129] Top shape: 100 (100)
I0704 12:26:45.352898 5465 net.cpp:137] Memory required for data: 1230000
I0704 12:26:45.352908 5465 layer_factory.hpp:77] Creating layer conv1
I0704 12:26:45.352936 5465 net.cpp:84] Creating Layer conv1
I0704 12:26:45.352947 5465 net.cpp:406] conv1 <- data
I0704 12:26:45.352963 5465 net.cpp:380] conv1 -> conv1
I0704 12:26:47.343575 5465 net.cpp:122] Setting up conv1
I0704 12:26:47.343633 5465 net.cpp:129] Top shape: 100 32 32 32 (3276800)
I0704 12:26:47.343646 5465 net.cpp:137] Memory required for data: 14337200
I0704 12:26:47.343694 5465 layer_factory.hpp:77] Creating layer pool1
I0704 12:26:47.343719 5465 net.cpp:84] Creating Layer pool1
I0704 12:26:47.343730 5465 net.cpp:406] pool1 <- conv1
I0704 12:26:47.343745 5465 net.cpp:380] pool1 -> pool1
I0704 12:26:47.343776 5465 net.cpp:122] Setting up pool1
I0704 12:26:47.343789 5465 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0704 12:26:47.343799 5465 net.cpp:137] Memory required for data: 17614000
I0704 12:26:47.343809 5465 layer_factory.hpp:77] Creating layer relu1
I0704 12:26:47.343827 5465 net.cpp:84] Creating Layer relu1
I0704 12:26:47.343838 5465 net.cpp:406] relu1 <- pool1
I0704 12:26:47.343852 5465 net.cpp:367] relu1 -> pool1 (in-place)
I0704 12:26:47.344049 5465 net.cpp:122] Setting up relu1
I0704 12:26:47.344064 5465 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0704 12:26:47.344074 5465 net.cpp:137] Memory required for data: 20890800
I0704 12:26:47.344084 5465 layer_factory.hpp:77] Creating layer norm1
I0704 12:26:47.344103 5465 net.cpp:84] Creating Layer norm1
I0704 12:26:47.344115 5465 net.cpp:406] norm1 <- pool1
I0704 12:26:47.344127 5465 net.cpp:380] norm1 -> norm1
I0704 12:26:47.344894 5465 net.cpp:122] Setting up norm1
I0704 12:26:47.344913 5465 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0704 12:26:47.344923 5465 net.cpp:137] Memory required for data: 24167600
I0704 12:26:47.344933 5465 layer_factory.hpp:77] Creating layer conv2
I0704 12:26:47.344955 5465 net.cpp:84] Creating Layer conv2
I0704 12:26:47.344966 5465 net.cpp:406] conv2 <- norm1
I0704 12:26:47.344981 5465 net.cpp:380] conv2 -> conv2
I0704 12:26:47.346536 5465 net.cpp:122] Setting up conv2
I0704 12:26:47.346554 5465 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0704 12:26:47.346565 5465 net.cpp:137] Memory required for data: 27444400
I0704 12:26:47.346583 5465 layer_factory.hpp:77] Creating layer relu2
I0704 12:26:47.346596 5465 net.cpp:84] Creating Layer relu2
I0704 12:26:47.346607 5465 net.cpp:406] relu2 <- conv2
I0704 12:26:47.346621 5465 net.cpp:367] relu2 -> conv2 (in-place)
I0704 12:26:47.346993 5465 net.cpp:122] Setting up relu2
I0704 12:26:47.347010 5465 net.cpp:129] Top shape: 100 32 16 16 (819200)
I0704 12:26:47.347021 5465 net.cpp:137] Memory required for data: 30721200
I0704 12:26:47.347031 5465 layer_factory.hpp:77] Creating layer pool2
I0704 12:26:47.347048 5465 net.cpp:84] Creating Layer pool2
I0704 12:26:47.347059 5465 net.cpp:406] pool2 <- conv2
I0704 12:26:47.347071 5465 net.cpp:380] pool2 -> pool2
I0704 12:26:47.347445 5465 net.cpp:122] Setting up pool2
I0704 12:26:47.347462 5465 net.cpp:129] Top shape: 100 32 8 8 (204800)
I0704 12:26:47.347472 5465 net.cpp:137] Memory required for data: 31540400
I0704 12:26:47.347482 5465 layer_factory.hpp:77] Creating layer norm2
I0704 12:26:47.347497 5465 net.cpp:84] Creating Layer norm2
I0704 12:26:47.347507 5465 net.cpp:406] norm2 <- pool2
I0704 12:26:47.347522 5465 net.cpp:380] norm2 -> norm2
I0704 12:26:47.348095 5465 net.cpp:122] Setting up norm2
I0704 12:26:47.348112 5465 net.cpp:129] Top shape: 100 32 8 8 (204800)
I0704 12:26:47.348122 5465 net.cpp:137] Memory required for data: 32359600
I0704 12:26:47.348132 5465 layer_factory.hpp:77] Creating layer conv3
I0704 12:26:47.348186 5465 net.cpp:84] Creating Layer conv3
I0704 12:26:47.348197 5465 net.cpp:406] conv3 <- norm2
I0704 12:26:47.348212 5465 net.cpp:380] conv3 -> conv3
I0704 12:26:47.358871 5465 net.cpp:122] Setting up conv3
I0704 12:26:47.358929 5465 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0704 12:26:47.358940 5465 net.cpp:137] Memory required for data: 33998000
I0704 12:26:47.358973 5465 layer_factory.hpp:77] Creating layer relu3
I0704 12:26:47.358996 5465 net.cpp:84] Creating Layer relu3
I0704 12:26:47.359007 5465 net.cpp:406] relu3 <- conv3
I0704 12:26:47.359025 5465 net.cpp:367] relu3 -> conv3 (in-place)
I0704 12:26:47.359354 5465 net.cpp:122] Setting up relu3
I0704 12:26:47.359377 5465 net.cpp:129] Top shape: 100 64 8 8 (409600)
I0704 12:26:47.359387 5465 net.cpp:137] Memory required for data: 35636400
I0704 12:26:47.359397 5465 layer_factory.hpp:77] Creating layer pool3
I0704 12:26:47.359411 5465 net.cpp:84] Creating Layer pool3
I0704 12:26:47.359422 5465 net.cpp:406] pool3 <- conv3
I0704 12:26:47.359439 5465 net.cpp:380] pool3 -> pool3
I0704 12:26:47.359905 5465 net.cpp:122] Setting up pool3
I0704 12:26:47.359925 5465 net.cpp:129] Top shape: 100 64 4 4 (102400)
I0704 12:26:47.359935 5465 net.cpp:137] Memory required for data: 36046000
I0704 12:26:47.359944 5465 layer_factory.hpp:77] Creating layer ip1
I0704 12:26:47.359964 5465 net.cpp:84] Creating Layer ip1
I0704 12:26:47.359975 5465 net.cpp:406] ip1 <- pool3
I0704 12:26:47.359992 5465 net.cpp:380] ip1 -> ip1
I0704 12:26:47.360213 5465 net.cpp:122] Setting up ip1
I0704 12:26:47.360225 5465 net.cpp:129] Top shape: 100 10 (1000)
I0704 12:26:47.360234 5465 net.cpp:137] Memory required for data: 36050000
I0704 12:26:47.360249 5465 layer_factory.hpp:77] Creating layer ip1_ip1_0_split
I0704 12:26:47.360265 5465 net.cpp:84] Creating Layer ip1_ip1_0_split
I0704 12:26:47.360276 5465 net.cpp:406] ip1_ip1_0_split <- ip1
I0704 12:26:47.360291 5465 net.cpp:380] ip1_ip1_0_split -> ip1_ip1_0_split_0
I0704 12:26:47.360307 5465 net.cpp:380] ip1_ip1_0_split -> ip1_ip1_0_split_1
I0704 12:26:47.360324 5465 net.cpp:122] Setting up ip1_ip1_0_split
I0704 12:26:47.360337 5465 net.cpp:129] Top shape: 100 10 (1000)
I0704 12:26:47.360347 5465 net.cpp:129] Top shape: 100 10 (1000)
I0704 12:26:47.360355 5465 net.cpp:137] Memory required for data: 36058000
I0704 12:26:47.360365 5465 layer_factory.hpp:77] Creating layer accuracy
I0704 12:26:47.360388 5465 net.cpp:84] Creating Layer accuracy
I0704 12:26:47.360399 5465 net.cpp:406] accuracy <- ip1_ip1_0_split_0
I0704 12:26:47.360410 5465 net.cpp:406] accuracy <- label_cifar_1_split_0
I0704 12:26:47.360426 5465 net.cpp:380] accuracy -> accuracy
I0704 12:26:47.360442 5465 net.cpp:122] Setting up accuracy
I0704 12:26:47.360455 5465 net.cpp:129] Top shape: (1)
I0704 12:26:47.360463 5465 net.cpp:137] Memory required for data: 36058004
I0704 12:26:47.360472 5465 layer_factory.hpp:77] Creating layer loss
I0704 12:26:47.360492 5465 net.cpp:84] Creating Layer loss
I0704 12:26:47.360503 5465 net.cpp:406] loss <- ip1_ip1_0_split_1
I0704 12:26:47.360514 5465 net.cpp:406] loss <- label_cifar_1_split_1
I0704 12:26:47.360528 5465 net.cpp:380] loss -> loss
I0704 12:26:47.360553 5465 layer_factory.hpp:77] Creating layer loss
I0704 12:26:47.360777 5465 net.cpp:122] Setting up loss
I0704 12:26:47.360795 5465 net.cpp:129] Top shape: (1)
I0704 12:26:47.360805 5465 net.cpp:132] with loss weight 1
I0704 12:26:47.360834 5465 net.cpp:137] Memory required for data: 36058008
I0704 12:26:47.360846 5465 net.cpp:198] loss needs backward computation.
I0704 12:26:47.360860 5465 net.cpp:200] accuracy does not need backward computation.
I0704 12:26:47.360872 5465 net.cpp:198] ip1_ip1_0_split needs backward computation.
I0704 12:26:47.360882 5465 net.cpp:198] ip1 needs backward computation.
I0704 12:26:47.360891 5465 net.cpp:198] pool3 needs backward computation.
I0704 12:26:47.360901 5465 net.cpp:198] relu3 needs backward computation.
I0704 12:26:47.360911 5465 net.cpp:198] conv3 needs backward computation.
I0704 12:26:47.360954 5465 net.cpp:198] norm2 needs backward computation.
I0704 12:26:47.360965 5465 net.cpp:198] pool2 needs backward computation.
I0704 12:26:47.360975 5465 net.cpp:198] relu2 needs backward computation.
I0704 12:26:47.360985 5465 net.cpp:198] conv2 needs backward computation.
I0704 12:26:47.360996 5465 net.cpp:198] norm1 needs backward computation.
I0704 12:26:47.361006 5465 net.cpp:198] relu1 needs backward computation.
I0704 12:26:47.361016 5465 net.cpp:198] pool1 needs backward computation.
I0704 12:26:47.361026 5465 net.cpp:198] conv1 needs backward computation.
I0704 12:26:47.361037 5465 net.cpp:200] label_cifar_1_split does not need backward computation.
I0704 12:26:47.361047 5465 net.cpp:200] cifar does not need backward computation.
I0704 12:26:47.361057 5465 net.cpp:242] This network produces output accuracy
I0704 12:26:47.361068 5465 net.cpp:242] This network produces output loss
I0704 12:26:47.361099 5465 net.cpp:255] Network initialization done.
I0704 12:26:47.440634 5465 hdf5.cpp:32] Datatype class: H5T_FLOAT
I0704 12:26:47.442189 5465 caffe.cpp:290] Running for 10 iterations.
I0704 12:26:48.122941 5465 caffe.cpp:313] Batch 0, accuracy = 0.81
I0704 12:26:48.123001 5465 caffe.cpp:313] Batch 0, loss = 0.669872
I0704 12:26:48.722308 5465 caffe.cpp:313] Batch 1, accuracy = 0.81
I0704 12:26:48.722373 5465 caffe.cpp:313] Batch 1, loss = 0.600901
I0704 12:26:49.322628 5465 caffe.cpp:313] Batch 2, accuracy = 0.72
I0704 12:26:49.322688 5465 caffe.cpp:313] Batch 2, loss = 0.712268
I0704 12:26:49.923393 5465 caffe.cpp:313] Batch 3, accuracy = 0.79
I0704 12:26:49.923454 5465 caffe.cpp:313] Batch 3, loss = 0.630273
I0704 12:26:50.522907 5465 caffe.cpp:313] Batch 4, accuracy = 0.8
I0704 12:26:50.522970 5465 caffe.cpp:313] Batch 4, loss = 0.519629
I0704 12:26:51.123852 5465 caffe.cpp:313] Batch 5, accuracy = 0.84
I0704 12:26:51.123913 5465 caffe.cpp:313] Batch 5, loss = 0.398108
I0704 12:26:51.724124 5465 caffe.cpp:313] Batch 6, accuracy = 0.73
I0704 12:26:51.724195 5465 caffe.cpp:313] Batch 6, loss = 0.712703
I0704 12:26:52.333739 5465 caffe.cpp:313] Batch 7, accuracy = 0.76
I0704 12:26:52.333806 5465 caffe.cpp:313] Batch 7, loss = 0.711946
I0704 12:26:52.934726 5465 caffe.cpp:313] Batch 8, accuracy = 0.8
I0704 12:26:52.934793 5465 caffe.cpp:313] Batch 8, loss = 0.67638
I0704 12:26:53.534425 5465 caffe.cpp:313] Batch 9, accuracy = 0.73
I0704 12:26:53.534492 5465 caffe.cpp:313] Batch 9, loss = 0.703718
I0704 12:26:53.534503 5465 caffe.cpp:318] Loss: 0.63358
I0704 12:26:53.534533 5465 caffe.cpp:330] accuracy = 0.779
I0704 12:26:53.534556 5465 caffe.cpp:330] loss = 0.63358 (* 1 = 0.63358 loss)
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$
-------------------------------------------------------------------------------------------------------------------------------------------------------
测试：做100次迭代的结果
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$ caffe test -model examples/cifar10/cifar10_full_train_test.prototxt -weights examples/cifar10/cifar10_full_iter_60000.caffemodel.h5 -iterations 100
I0704 13:12:44.414038 6622 caffe.cpp:284] Use CPU.
I0704 13:12:44.681273 6622 net.cpp:294] The NetState phase (1) differed from the phase (0) specified by a rule in layer cifar
I0704 13:12:44.681501 6622 net.cpp:51] Initializing net from parameters:
..............................
..............................
I0704 13:12:44.998669 6622 caffe.cpp:290] Running for 100 iterations.
I0704 13:12:45.619782 6622 caffe.cpp:313] Batch 0, accuracy = 0.81
I0704 13:12:45.619843 6622 caffe.cpp:313] Batch 0, loss = 0.669872
I0704 13:12:46.217348 6622 caffe.cpp:313] Batch 1, accuracy = 0.81
I0704 13:12:46.217408 6622 caffe.cpp:313] Batch 1, loss = 0.600901
I0704 13:12:46.818264 6622 caffe.cpp:313] Batch 2, accuracy = 0.72
I0704 13:12:46.818323 6622 caffe.cpp:313] Batch 2, loss = 0.712268
I0704 13:12:47.415946 6622 caffe.cpp:313] Batch 3, accuracy = 0.79
I0704 13:12:47.416007 6622 caffe.cpp:313] Batch 3, loss = 0.630273
I0704 13:12:48.015729 6622 caffe.cpp:313] Batch 4, accuracy = 0.8
I0704 13:12:48.015792 6622 caffe.cpp:313] Batch 4, loss = 0.519629
I0704 13:12:48.621882 6622 caffe.cpp:313] Batch 5, accuracy = 0.84
I0704 13:12:48.621939 6622 caffe.cpp:313] Batch 5, loss = 0.398108
I0704 13:12:49.227767 6622 caffe.cpp:313] Batch 6, accuracy = 0.73
I0704 13:12:49.227831 6622 caffe.cpp:313] Batch 6, loss = 0.712703
I0704 13:12:49.833628 6622 caffe.cpp:313] Batch 7, accuracy = 0.76
I0704 13:12:49.833700 6622 caffe.cpp:313] Batch 7, loss = 0.711946
I0704 13:12:50.431000 6622 caffe.cpp:313] Batch 8, accuracy = 0.8
I0704 13:12:50.431063 6622 caffe.cpp:313] Batch 8, loss = 0.67638
I0704 13:12:51.036695 6622 caffe.cpp:313] Batch 9, accuracy = 0.73
I0704 13:12:51.036754 6622 caffe.cpp:313] Batch 9, loss = 0.703718
I0704 13:12:51.654356 6622 caffe.cpp:313] Batch 10, accuracy = 0.81
I0704 13:12:51.654417 6622 caffe.cpp:313] Batch 10, loss = 0.628615
I0704 13:12:52.271215 6622 caffe.cpp:313] Batch 11, accuracy = 0.77
I0704 13:12:52.271272 6622 caffe.cpp:313] Batch 11, loss = 0.640812
I0704 13:12:52.871871 6622 caffe.cpp:313] Batch 12, accuracy = 0.84
I0704 13:12:52.871930 6622 caffe.cpp:313] Batch 12, loss = 0.48857
I0704 13:12:53.500851 6622 caffe.cpp:313] Batch 13, accuracy = 0.82
I0704 13:12:53.500921 6622 caffe.cpp:313] Batch 13, loss = 0.545125
I0704 13:12:54.106851 6622 caffe.cpp:313] Batch 14, accuracy = 0.83
I0704 13:12:54.106916 6622 caffe.cpp:313] Batch 14, loss = 0.454256
I0704 13:12:54.712241 6622 caffe.cpp:313] Batch 15, accuracy = 0.8
I0704 13:12:54.712308 6622 caffe.cpp:313] Batch 15, loss = 0.604123
I0704 13:12:55.317904 6622 caffe.cpp:313] Batch 16, accuracy = 0.82
I0704 13:12:55.317970 6622 caffe.cpp:313] Batch 16, loss = 0.602975
I0704 13:12:55.915468 6622 caffe.cpp:313] Batch 17, accuracy = 0.78
I0704 13:12:55.915534 6622 caffe.cpp:313] Batch 17, loss = 0.653725
I0704 13:12:56.521464 6622 caffe.cpp:313] Batch 18, accuracy = 0.74
I0704 13:12:56.521531 6622 caffe.cpp:313] Batch 18, loss = 0.775862
I0704 13:12:57.127270 6622 caffe.cpp:313] Batch 19, accuracy = 0.7
I0704 13:12:57.127341 6622 caffe.cpp:313] Batch 19, loss = 0.927896
I0704 13:12:57.724392 6622 caffe.cpp:313] Batch 20, accuracy = 0.75
I0704 13:12:57.724458 6622 caffe.cpp:313] Batch 20, loss = 0.631937
I0704 13:12:58.330080 6622 caffe.cpp:313] Batch 21, accuracy = 0.75
I0704 13:12:58.330200 6622 caffe.cpp:313] Batch 21, loss = 0.689163
I0704 13:12:58.927534 6622 caffe.cpp:313] Batch 22, accuracy = 0.79
I0704 13:12:58.927600 6622 caffe.cpp:313] Batch 22, loss = 0.718018
I0704 13:12:59.533542 6622 caffe.cpp:313] Batch 23, accuracy = 0.76
I0704 13:12:59.533630 6622 caffe.cpp:313] Batch 23, loss = 0.772289
I0704 13:13:00.131110 6622 caffe.cpp:313] Batch 24, accuracy = 0.78
I0704 13:13:00.131175 6622 caffe.cpp:313] Batch 24, loss = 0.616908
I0704 13:13:00.727895 6622 caffe.cpp:313] Batch 25, accuracy = 0.69
I0704 13:13:00.727962 6622 caffe.cpp:313] Batch 25, loss = 0.921352
I0704 13:13:01.324116 6622 caffe.cpp:313] Batch 26, accuracy = 0.86
I0704 13:13:01.324188 6622 caffe.cpp:313] Batch 26, loss = 0.43752
I0704 13:13:01.920688 6622 caffe.cpp:313] Batch 27, accuracy = 0.76
I0704 13:13:01.920760 6622 caffe.cpp:313] Batch 27, loss = 0.694387
I0704 13:13:02.518180 6622 caffe.cpp:313] Batch 28, accuracy = 0.88
I0704 13:13:02.518250 6622 caffe.cpp:313] Batch 28, loss = 0.503546
I0704 13:13:03.123641 6622 caffe.cpp:313] Batch 29, accuracy = 0.76
I0704 13:13:03.123706 6622 caffe.cpp:313] Batch 29, loss = 0.675348
I0704 13:13:03.729601 6622 caffe.cpp:313] Batch 30, accuracy = 0.75
I0704 13:13:03.729679 6622 caffe.cpp:313] Batch 30, loss = 0.641321
I0704 13:13:04.326668 6622 caffe.cpp:313] Batch 31, accuracy = 0.79
I0704 13:13:04.326732 6622 caffe.cpp:313] Batch 31, loss = 0.668134
I0704 13:13:04.932490 6622 caffe.cpp:313] Batch 32, accuracy = 0.78
I0704 13:13:04.932554 6622 caffe.cpp:313] Batch 32, loss = 0.587462
I0704 13:13:05.529435 6622 caffe.cpp:313] Batch 33, accuracy = 0.69
I0704 13:13:05.529508 6622 caffe.cpp:313] Batch 33, loss = 0.809499
I0704 13:13:06.125875 6622 caffe.cpp:313] Batch 34, accuracy = 0.72
I0704 13:13:06.125944 6622 caffe.cpp:313] Batch 34, loss = 0.841917
I0704 13:13:06.722806 6622 caffe.cpp:313] Batch 35, accuracy = 0.81
I0704 13:13:06.722877 6622 caffe.cpp:313] Batch 35, loss = 0.65033
I0704 13:13:07.328688 6622 caffe.cpp:313] Batch 36, accuracy = 0.74
I0704 13:13:07.328760 6622 caffe.cpp:313] Batch 36, loss = 0.73518
I0704 13:13:07.926215 6622 caffe.cpp:313] Batch 37, accuracy = 0.77
I0704 13:13:07.926282 6622 caffe.cpp:313] Batch 37, loss = 0.626204
I0704 13:13:08.531821 6622 caffe.cpp:313] Batch 38, accuracy = 0.84
I0704 13:13:08.531885 6622 caffe.cpp:313] Batch 38, loss = 0.50705
I0704 13:13:09.128522 6622 caffe.cpp:313] Batch 39, accuracy = 0.86
I0704 13:13:09.128587 6622 caffe.cpp:313] Batch 39, loss = 0.45618
I0704 13:13:09.725127 6622 caffe.cpp:313] Batch 40, accuracy = 0.82
I0704 13:13:09.725200 6622 caffe.cpp:313] Batch 40, loss = 0.594011
I0704 13:13:10.321892 6622 caffe.cpp:313] Batch 41, accuracy = 0.83
I0704 13:13:10.321964 6622 caffe.cpp:313] Batch 41, loss = 0.673196
I0704 13:13:10.918488 6622 caffe.cpp:313] Batch 42, accuracy = 0.85
I0704 13:13:10.918555 6622 caffe.cpp:313] Batch 42, loss = 0.45519
I0704 13:13:11.524237 6622 caffe.cpp:313] Batch 43, accuracy = 0.79
I0704 13:13:11.524307 6622 caffe.cpp:313] Batch 43, loss = 0.628985
I0704 13:13:12.129947 6622 caffe.cpp:313] Batch 44, accuracy = 0.78
I0704 13:13:12.130018 6622 caffe.cpp:313] Batch 44, loss = 0.704623
I0704 13:13:12.735232 6622 caffe.cpp:313] Batch 45, accuracy = 0.71
I0704 13:13:12.735304 6622 caffe.cpp:313] Batch 45, loss = 0.695174
I0704 13:13:13.340771 6622 caffe.cpp:313] Batch 46, accuracy = 0.84
I0704 13:13:13.340842 6622 caffe.cpp:313] Batch 46, loss = 0.552928
I0704 13:13:13.947043 6622 caffe.cpp:313] Batch 47, accuracy = 0.76
I0704 13:13:13.947109 6622 caffe.cpp:313] Batch 47, loss = 0.651739
I0704 13:13:14.553036 6622 caffe.cpp:313] Batch 48, accuracy = 0.82
I0704 13:13:14.553324 6622 caffe.cpp:313] Batch 48, loss = 0.441534
I0704 13:13:15.180112 6622 caffe.cpp:313] Batch 49, accuracy = 0.76
I0704 13:13:15.180176 6622 caffe.cpp:313] Batch 49, loss = 0.729064
I0704 13:13:15.779243 6622 caffe.cpp:313] Batch 50, accuracy = 0.8
I0704 13:13:15.779307 6622 caffe.cpp:313] Batch 50, loss = 0.584773
I0704 13:13:16.378615 6622 caffe.cpp:313] Batch 51, accuracy = 0.79
I0704 13:13:16.378679 6622 caffe.cpp:313] Batch 51, loss = 0.541237
I0704 13:13:16.978050 6622 caffe.cpp:313] Batch 52, accuracy = 0.79
I0704 13:13:16.978111 6622 caffe.cpp:313] Batch 52, loss = 0.656132
I0704 13:13:17.577879 6622 caffe.cpp:313] Batch 53, accuracy = 0.79
I0704 13:13:17.577941 6622 caffe.cpp:313] Batch 53, loss = 0.568454
I0704 13:13:18.177491 6622 caffe.cpp:313] Batch 54, accuracy = 0.72
I0704 13:13:18.177562 6622 caffe.cpp:313] Batch 54, loss = 0.797033
I0704 13:13:18.776235 6622 caffe.cpp:313] Batch 55, accuracy = 0.82
I0704 13:13:18.776298 6622 caffe.cpp:313] Batch 55, loss = 0.679914
I0704 13:13:19.375723 6622 caffe.cpp:313] Batch 56, accuracy = 0.76
I0704 13:13:19.375792 6622 caffe.cpp:313] Batch 56, loss = 0.681724
I0704 13:13:19.974643 6622 caffe.cpp:313] Batch 57, accuracy = 0.86
I0704 13:13:19.974711 6622 caffe.cpp:313] Batch 57, loss = 0.460299
I0704 13:13:20.574430 6622 caffe.cpp:313] Batch 58, accuracy = 0.78
I0704 13:13:20.574501 6622 caffe.cpp:313] Batch 58, loss = 0.694127
I0704 13:13:21.174055 6622 caffe.cpp:313] Batch 59, accuracy = 0.74
I0704 13:13:21.174118 6622 caffe.cpp:313] Batch 59, loss = 0.748216
I0704 13:13:21.804786 6622 caffe.cpp:313] Batch 60, accuracy = 0.79
I0704 13:13:21.804847 6622 caffe.cpp:313] Batch 60, loss = 0.566229
I0704 13:13:22.410190 6622 caffe.cpp:313] Batch 61, accuracy = 0.8
I0704 13:13:22.410251 6622 caffe.cpp:313] Batch 61, loss = 0.53662
I0704 13:13:23.011648 6622 caffe.cpp:313] Batch 62, accuracy = 0.77
I0704 13:13:23.011713 6622 caffe.cpp:313] Batch 62, loss = 0.630629
I0704 13:13:23.618448 6622 caffe.cpp:313] Batch 63, accuracy = 0.8
I0704 13:13:23.618508 6622 caffe.cpp:313] Batch 63, loss = 0.553974
I0704 13:13:24.215498 6622 caffe.cpp:313] Batch 64, accuracy = 0.82
I0704 13:13:24.215561 6622 caffe.cpp:313] Batch 64, loss = 0.640087
I0704 13:13:24.812703 6622 caffe.cpp:313] Batch 65, accuracy = 0.78
I0704 13:13:24.812767 6622 caffe.cpp:313] Batch 65, loss = 0.720206
I0704 13:13:25.417578 6622 caffe.cpp:313] Batch 66, accuracy = 0.85
I0704 13:13:25.417637 6622 caffe.cpp:313] Batch 66, loss = 0.469227
I0704 13:13:26.015293 6622 caffe.cpp:313] Batch 67, accuracy = 0.74
I0704 13:13:26.015355 6622 caffe.cpp:313] Batch 67, loss = 0.662439
I0704 13:13:26.621019 6622 caffe.cpp:313] Batch 68, accuracy = 0.73
I0704 13:13:26.621078 6622 caffe.cpp:313] Batch 68, loss = 0.832033
I0704 13:13:27.226202 6622 caffe.cpp:313] Batch 69, accuracy = 0.74
I0704 13:13:27.226261 6622 caffe.cpp:313] Batch 69, loss = 0.7851
I0704 13:13:27.831529 6622 caffe.cpp:313] Batch 70, accuracy = 0.79
I0704 13:13:27.831593 6622 caffe.cpp:313] Batch 70, loss = 0.594752
I0704 13:13:28.437043 6622 caffe.cpp:313] Batch 71, accuracy = 0.81
I0704 13:13:28.437101 6622 caffe.cpp:313] Batch 71, loss = 0.600966
I0704 13:13:29.034024 6622 caffe.cpp:313] Batch 72, accuracy = 0.89
I0704 13:13:29.034085 6622 caffe.cpp:313] Batch 72, loss = 0.544593
I0704 13:13:29.639663 6622 caffe.cpp:313] Batch 73, accuracy = 0.81
I0704 13:13:29.639719 6622 caffe.cpp:313] Batch 73, loss = 0.578454
I0704 13:13:30.245527 6622 caffe.cpp:313] Batch 74, accuracy = 0.71
I0704 13:13:30.245589 6622 caffe.cpp:313] Batch 74, loss = 0.85487
I0704 13:13:30.851264 6622 caffe.cpp:313] Batch 75, accuracy = 0.75
I0704 13:13:30.851325 6622 caffe.cpp:313] Batch 75, loss = 0.703642
I0704 13:13:31.448464 6622 caffe.cpp:313] Batch 76, accuracy = 0.77
I0704 13:13:31.448526 6622 caffe.cpp:313] Batch 76, loss = 0.702493
I0704 13:13:32.053625 6622 caffe.cpp:313] Batch 77, accuracy = 0.77
I0704 13:13:32.053694 6622 caffe.cpp:313] Batch 77, loss = 0.616773
I0704 13:13:32.659436 6622 caffe.cpp:313] Batch 78, accuracy = 0.77
I0704 13:13:32.659543 6622 caffe.cpp:313] Batch 78, loss = 0.624216
I0704 13:13:33.256774 6622 caffe.cpp:313] Batch 79, accuracy = 0.76
I0704 13:13:33.256835 6622 caffe.cpp:313] Batch 79, loss = 0.702435
I0704 13:13:33.862572 6622 caffe.cpp:313] Batch 80, accuracy = 0.83
I0704 13:13:33.862634 6622 caffe.cpp:313] Batch 80, loss = 0.479035
I0704 13:13:34.460429 6622 caffe.cpp:313] Batch 81, accuracy = 0.77
I0704 13:13:34.460491 6622 caffe.cpp:313] Batch 81, loss = 0.679997
I0704 13:13:35.066623 6622 caffe.cpp:313] Batch 82, accuracy = 0.76
I0704 13:13:35.066686 6622 caffe.cpp:313] Batch 82, loss = 0.668821
I0704 13:13:35.672404 6622 caffe.cpp:313] Batch 83, accuracy = 0.79
I0704 13:13:35.672462 6622 caffe.cpp:313] Batch 83, loss = 0.59493
I0704 13:13:36.277426 6622 caffe.cpp:313] Batch 84, accuracy = 0.77
I0704 13:13:36.277488 6622 caffe.cpp:313] Batch 84, loss = 0.636169
I0704 13:13:36.883759 6622 caffe.cpp:313] Batch 85, accuracy = 0.79
I0704 13:13:36.883818 6622 caffe.cpp:313] Batch 85, loss = 0.655568
I0704 13:13:37.489516 6622 caffe.cpp:313] Batch 86, accuracy = 0.79
I0704 13:13:37.489574 6622 caffe.cpp:313] Batch 86, loss = 0.637793
I0704 13:13:38.086877 6622 caffe.cpp:313] Batch 87, accuracy = 0.84
I0704 13:13:38.086935 6622 caffe.cpp:313] Batch 87, loss = 0.604765
I0704 13:13:38.691869 6622 caffe.cpp:313] Batch 88, accuracy = 0.81
I0704 13:13:38.691931 6622 caffe.cpp:313] Batch 88, loss = 0.525659
I0704 13:13:39.297114 6622 caffe.cpp:313] Batch 89, accuracy = 0.76
I0704 13:13:39.297176 6622 caffe.cpp:313] Batch 89, loss = 0.657071
I0704 13:13:39.902731 6622 caffe.cpp:313] Batch 90, accuracy = 0.81
I0704 13:13:39.902787 6622 caffe.cpp:313] Batch 90, loss = 0.5901
I0704 13:13:40.499342 6622 caffe.cpp:313] Batch 91, accuracy = 0.85
I0704 13:13:40.499403 6622 caffe.cpp:313] Batch 91, loss = 0.433673
I0704 13:13:41.105139 6622 caffe.cpp:313] Batch 92, accuracy = 0.76
I0704 13:13:41.105197 6622 caffe.cpp:313] Batch 92, loss = 0.68505
I0704 13:13:41.711143 6622 caffe.cpp:313] Batch 93, accuracy = 0.84
I0704 13:13:41.711205 6622 caffe.cpp:313] Batch 93, loss = 0.526908
I0704 13:13:42.317083 6622 caffe.cpp:313] Batch 94, accuracy = 0.81
I0704 13:13:42.317149 6622 caffe.cpp:313] Batch 94, loss = 0.54016
I0704 13:13:42.923281 6622 caffe.cpp:313] Batch 95, accuracy = 0.81
I0704 13:13:42.923348 6622 caffe.cpp:313] Batch 95, loss = 0.568361
I0704 13:13:42.924571 6628 data_layer.cpp:73] Restarting data prefetching from start.
I0704 13:13:43.529124 6622 caffe.cpp:313] Batch 96, accuracy = 0.82
I0704 13:13:43.529189 6622 caffe.cpp:313] Batch 96, loss = 0.455237
I0704 13:13:44.135169 6622 caffe.cpp:313] Batch 97, accuracy = 0.78
I0704 13:13:44.135228 6622 caffe.cpp:313] Batch 97, loss = 0.753922
I0704 13:13:44.740659 6622 caffe.cpp:313] Batch 98, accuracy = 0.73
I0704 13:13:44.740911 6622 caffe.cpp:313] Batch 98, loss = 0.709454
I0704 13:13:45.338565 6622 caffe.cpp:313] Batch 99, accuracy = 0.78
I0704 13:13:45.338631 6622 caffe.cpp:313] Batch 99, loss = 0.737646
I0704 13:13:45.338644 6622 caffe.cpp:318] Loss: 0.634399
I0704 13:13:45.338677 6622 caffe.cpp:330] accuracy = 0.7859
I0704 13:13:45.338701 6622 caffe.cpp:330] loss = 0.634399 (* 1 = 0.634399 loss)
识别猫：
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$ classification examples/cifar10/cifar10_full.prototxt examples/cifar10/cifar10_full_iter_60000.caffemodel.h5 examples/cifar10/mean.binaryproto data/cifar10/synset_words.txt examples/images/cat.jpg
---------- Prediction for examples/images/cat.jpg ----------
0.7481 - " deer "
0.1352 - " bird "
0.0476 - " cat "
0.0162 - " frog "
0.0143 - " horse "
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$
灰色猫：
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$ classification examples/cifar10/cifar10_full.prototxt examples/cifar10/cifar10_full_iter_60000.caffemodel.h5 examples/cifar10/mean.binaryproto data/cifar10/synset_words.txt examples/images/cat_gray.jpg
---------- Prediction for examples/images/cat_gray.jpg ----------
0.2487 - " bird "
0.2476 - " horse "
0.1985 - " dog "
0.1277 - " cat "
0.0853 - " deer "
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$
人鱼自行车：
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$ classification examples/cifar10/cifar10_full.prototxt examples/cifar10/cifar10_full_iter_60000.caffemodel.h5 examples/cifar10/mean.binaryproto data/cifar10/synset_words.txt examples/images/fish-bike.jpg
---------- Prediction for examples/images/fish-bike.jpg ----------
0.6517 - " horse "
0.1291 - " truck "
0.0530 - " deer "
0.0441 - " cat "
0.0435 - " frog "
seag@seag-G41MT-S2PT:~/wsCaffe/caffe$
看起来效果很糟糕的。