【邱希鹏】nndl-chap5-数字识别(pytorch)

1. 导入包

import os
import torch
import torch.nn as nn
from torch.autograd import Variable
import torch.utils.data as Data
import torchvision
import torch.nn.functional as F
import numpy as np
from torchvision import datasets, transforms
learning_rate = 1e-4
keep_prob_rate = 0.7 #
max_epoch = 3
BATCH_SIZE = 50

DOWNLOAD_MNIST = False
if not(os.path.exists('MNIST')) or not os.listdir('MNIST'):
    # not mnist dir or mnist is empyt dir
    DOWNLOAD_MNIST = True

2. 导入数据

train_data = torchvision.datasets.MNIST(root='./', train = True, download=DOWNLOAD_MNIST,
                                        transform = torchvision.transforms.Compose([
                                            transforms.ToTensor(),
                                        ]))
train_loader = Data.DataLoader(dataset = train_data ,batch_size= BATCH_SIZE ,shuffle= True)

test_data = torchvision.datasets.MNIST(root = './', train = False,
                                      transform = transforms.Compose([
                                          transforms.ToTensor(),
                                      ]))
test_loader = Data.DataLoader(dataset = test_data, batch_size = BATCH_SIZE, shuffle = True)

test_x = Variable(torch.unsqueeze(test_data.test_data,dim  = 1),volatile = True).type(torch.FloatTensor)[:500]/255.
test_y = test_data.test_labels[:500].numpy()
print(test_x.shape)
print(test_y.shape)

# torch.Size([500, 1, 28, 28])
# (500,)

3. CNN模型

class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        self.conv1 = nn.Sequential(
            nn.Conv2d( # ???
                # patch 7 * 7 ; 1  in channels ; 32 out channels ; ; stride is 1
                # padding style is same(that means the convolution opration's input and output have the same size)
                in_channels = 1     ,  
                out_channels = 32   ,
                kernel_size = 7     ,
                stride = 1          ,
                padding = 0    ,
            ),
            nn.ReLU(),        # activation function
            nn.MaxPool2d(2),  # pooling operation
        )
        self.conv2 = nn.Sequential( # ???
            # line 1 : convolution function, patch 5*5 , 32 in channels ;64 out channels; padding style is same; stride is 1
            # line 2 : choosing your activation funciont
            # line 3 : pooling operation function.
            nn.Conv2d(
                in_channels = 32,
                out_channels = 64,
                kernel_size = 5,
                stride = 1,
                padding = 0,
            ),
            nn.ReLU(),
            nn.MaxPool2d(1),
        )
        self.out1 = nn.Linear(7*7*64 , 1024 , bias= True)   # full connection layer one

        self.dropout = nn.Dropout(keep_prob_rate)
        self.out2 = nn.Linear(1024, 10, bias=True)


    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        x = x.view(-1, 64*7*7)  # flatten the output of coonv2 to (batch_size ,64 * 7 * 7)    # ???
        out1 = self.out1(x)
        out1 = F.relu(out1)
        out1 = self.dropout(out1)
        out2 = self.out2(out1)
        output = F.softmax(out2)
        return output

4. 训练与测试

def test(cnn):
    global prediction
    y_pre = cnn(test_x)
    _,pre_index= torch.max(y_pre,1)
    pre_index= pre_index.view(-1)
    prediction = pre_index.data.numpy()
    correct  = np.sum(prediction == test_y)
    return correct / 500.0


def train(cnn):
    optimizer = torch.optim.Adam(cnn.parameters(), lr=learning_rate )
    loss_func = nn.CrossEntropyLoss()
    for epoch in range(max_epoch):
        for step, (x_, y_) in enumerate(train_loader):
            x ,y= Variable(x_),Variable(y_)
            output = cnn(x)  
            loss = loss_func(output, y)   # 标量
            
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()
            
            
            if step != 0 and step % 20 ==0:
                print("=" * 10,step,"="*5,"="*5, "test accuracy is ",test(cnn) ,"=" * 10 )

4.1 训练

cnn = CNN()
train(cnn)

========== 20 ===== ===== test accuracy is  0.25 ==========
========== 40 ===== ===== test accuracy is  0.458 ==========
========== 60 ===== ===== test accuracy is  0.572 ==========
========== 80 ===== ===== test accuracy is  0.624 ==========
========== 100 ===== ===== test accuracy is  0.638 ==========
========== 120 ===== ===== test accuracy is  0.718 ==========
========== 140 ===== ===== test accuracy is  0.744 ==========
========== 160 ===== ===== test accuracy is  0.796 ==========
========== 180 ===== ===== test accuracy is  0.806 ==========
========== 200 ===== ===== test accuracy is  0.808 ==========
========== 220 ===== ===== test accuracy is  0.844 ==========
========== 240 ===== ===== test accuracy is  0.84 ==========
========== 260 ===== ===== test accuracy is  0.864 ==========
========== 280 ===== ===== test accuracy is  0.86 ==========
========== 300 ===== ===== test accuracy is  0.878 ==========
========== 320 ===== ===== test accuracy is  0.868 ==========
========== 340 ===== ===== test accuracy is  0.876 ==========
========== 360 ===== ===== test accuracy is  0.862 ==========
========== 380 ===== ===== test accuracy is  0.86 ==========
========== 400 ===== ===== test accuracy is  0.892 ==========
========== 420 ===== ===== test accuracy is  0.87 ==========
========== 440 ===== ===== test accuracy is  0.882 ==========
========== 460 ===== ===== test accuracy is  0.898 ==========
========== 480 ===== ===== test accuracy is  0.892 ==========
========== 500 ===== ===== test accuracy is  0.884 ==========
========== 520 ===== ===== test accuracy is  0.892 ==========
========== 540 ===== ===== test accuracy is  0.892 ==========
========== 560 ===== ===== test accuracy is  0.902 ==========
========== 580 ===== ===== test accuracy is  0.902 ==========
========== 600 ===== ===== test accuracy is  0.904 ==========
========== 620 ===== ===== test accuracy is  0.902 ==========
========== 640 ===== ===== test accuracy is  0.904 ==========
========== 660 ===== ===== test accuracy is  0.906 ==========
========== 680 ===== ===== test accuracy is  0.908 ==========
========== 700 ===== ===== test accuracy is  0.922 ==========
========== 720 ===== ===== test accuracy is  0.916 ==========
========== 740 ===== ===== test accuracy is  0.918 ==========
========== 760 ===== ===== test accuracy is  0.906 ==========
========== 780 ===== ===== test accuracy is  0.924 ==========
========== 800 ===== ===== test accuracy is  0.928 ==========
========== 820 ===== ===== test accuracy is  0.918 ==========
========== 840 ===== ===== test accuracy is  0.922 ==========
========== 860 ===== ===== test accuracy is  0.918 ==========
========== 880 ===== ===== test accuracy is  0.93 ==========
========== 900 ===== ===== test accuracy is  0.924 ==========
========== 920 ===== ===== test accuracy is  0.932 ==========
========== 940 ===== ===== test accuracy is  0.934 ==========
========== 960 ===== ===== test accuracy is  0.926 ==========
========== 980 ===== ===== test accuracy is  0.932 ==========
========== 1000 ===== ===== test accuracy is  0.934 ==========
========== 1020 ===== ===== test accuracy is  0.926 ==========
========== 1040 ===== ===== test accuracy is  0.924 ==========
========== 1060 ===== ===== test accuracy is  0.934 ==========
========== 1080 ===== ===== test accuracy is  0.932 ==========
========== 1100 ===== ===== test accuracy is  0.934 ==========
========== 1120 ===== ===== test accuracy is  0.936 ==========
========== 1140 ===== ===== test accuracy is  0.936 ==========
========== 1160 ===== ===== test accuracy is  0.93 ==========
========== 1180 ===== ===== test accuracy is  0.932 ==========
========== 20 ===== ===== test accuracy is  0.934 ==========
========== 40 ===== ===== test accuracy is  0.946 ==========
========== 60 ===== ===== test accuracy is  0.94 ==========
========== 80 ===== ===== test accuracy is  0.946 ==========
========== 100 ===== ===== test accuracy is  0.946 ==========
========== 120 ===== ===== test accuracy is  0.944 ==========
========== 140 ===== ===== test accuracy is  0.946 ==========
========== 160 ===== ===== test accuracy is  0.956 ==========
========== 180 ===== ===== test accuracy is  0.936 ==========
========== 200 ===== ===== test accuracy is  0.95 ==========
========== 220 ===== ===== test accuracy is  0.956 ==========
========== 240 ===== ===== test accuracy is  0.946 ==========
========== 260 ===== ===== test accuracy is  0.944 ==========
========== 280 ===== ===== test accuracy is  0.944 ==========
========== 300 ===== ===== test accuracy is  0.954 ==========
========== 320 ===== ===== test accuracy is  0.964 ==========
========== 340 ===== ===== test accuracy is  0.95 ==========
========== 360 ===== ===== test accuracy is  0.962 ==========
========== 380 ===== ===== test accuracy is  0.948 ==========
========== 400 ===== ===== test accuracy is  0.96 ==========
========== 420 ===== ===== test accuracy is  0.946 ==========
========== 440 ===== ===== test accuracy is  0.96 ==========
========== 460 ===== ===== test accuracy is  0.948 ==========
========== 480 ===== ===== test accuracy is  0.95 ==========
========== 500 ===== ===== test accuracy is  0.958 ==========
========== 520 ===== ===== test accuracy is  0.954 ==========
========== 540 ===== ===== test accuracy is  0.948 ==========
========== 560 ===== ===== test accuracy is  0.958 ==========
========== 580 ===== ===== test accuracy is  0.948 ==========
========== 600 ===== ===== test accuracy is  0.96 ==========
========== 620 ===== ===== test accuracy is  0.96 ==========
========== 640 ===== ===== test accuracy is  0.96 ==========
========== 660 ===== ===== test accuracy is  0.95 ==========
========== 680 ===== ===== test accuracy is  0.962 ==========
========== 700 ===== ===== test accuracy is  0.964 ==========
========== 720 ===== ===== test accuracy is  0.962 ==========
========== 740 ===== ===== test accuracy is  0.96 ==========
========== 760 ===== ===== test accuracy is  0.954 ==========
========== 780 ===== ===== test accuracy is  0.956 ==========
========== 800 ===== ===== test accuracy is  0.962 ==========
========== 820 ===== ===== test accuracy is  0.962 ==========
========== 840 ===== ===== test accuracy is  0.968 ==========
========== 860 ===== ===== test accuracy is  0.962 ==========
========== 880 ===== ===== test accuracy is  0.972 ==========
========== 900 ===== ===== test accuracy is  0.96 ==========
========== 920 ===== ===== test accuracy is  0.958 ==========
========== 940 ===== ===== test accuracy is  0.966 ==========
========== 960 ===== ===== test accuracy is  0.972 ==========
========== 980 ===== ===== test accuracy is  0.964 ==========
========== 1000 ===== ===== test accuracy is  0.968 ==========
========== 1020 ===== ===== test accuracy is  0.968 ==========
========== 1040 ===== ===== test accuracy is  0.956 ==========
========== 1060 ===== ===== test accuracy is  0.96 ==========
========== 1080 ===== ===== test accuracy is  0.97 ==========
========== 1100 ===== ===== test accuracy is  0.968 ==========
========== 1120 ===== ===== test accuracy is  0.964 ==========
========== 1140 ===== ===== test accuracy is  0.97 ==========
========== 1160 ===== ===== test accuracy is  0.97 ==========
========== 1180 ===== ===== test accuracy is  0.96 ==========
========== 20 ===== ===== test accuracy is  0.96 ==========
========== 40 ===== ===== test accuracy is  0.962 ==========
========== 60 ===== ===== test accuracy is  0.97 ==========
========== 80 ===== ===== test accuracy is  0.958 ==========
========== 100 ===== ===== test accuracy is  0.966 ==========
========== 120 ===== ===== test accuracy is  0.962 ==========
========== 140 ===== ===== test accuracy is  0.968 ==========
========== 160 ===== ===== test accuracy is  0.972 ==========
========== 180 ===== ===== test accuracy is  0.972 ==========
========== 200 ===== ===== test accuracy is  0.978 ==========
========== 220 ===== ===== test accuracy is  0.968 ==========
========== 240 ===== ===== test accuracy is  0.956 ==========
========== 260 ===== ===== test accuracy is  0.97 ==========
========== 280 ===== ===== test accuracy is  0.964 ==========
========== 300 ===== ===== test accuracy is  0.97 ==========
========== 320 ===== ===== test accuracy is  0.972 ==========
========== 340 ===== ===== test accuracy is  0.976 ==========
========== 360 ===== ===== test accuracy is  0.968 ==========
========== 380 ===== ===== test accuracy is  0.97 ==========
========== 400 ===== ===== test accuracy is  0.974 ==========
========== 420 ===== ===== test accuracy is  0.974 ==========
========== 440 ===== ===== test accuracy is  0.968 ==========
========== 460 ===== ===== test accuracy is  0.976 ==========
========== 480 ===== ===== test accuracy is  0.97 ==========
========== 500 ===== ===== test accuracy is  0.96 ==========
========== 520 ===== ===== test accuracy is  0.966 ==========
========== 540 ===== ===== test accuracy is  0.974 ==========
========== 560 ===== ===== test accuracy is  0.974 ==========
========== 580 ===== ===== test accuracy is  0.972 ==========
========== 600 ===== ===== test accuracy is  0.974 ==========
========== 620 ===== ===== test accuracy is  0.97 ==========
========== 640 ===== ===== test accuracy is  0.974 ==========
========== 660 ===== ===== test accuracy is  0.976 ==========
========== 680 ===== ===== test accuracy is  0.97 ==========
========== 700 ===== ===== test accuracy is  0.974 ==========
========== 720 ===== ===== test accuracy is  0.962 ==========
========== 740 ===== ===== test accuracy is  0.98 ==========
========== 760 ===== ===== test accuracy is  0.976 ==========
========== 780 ===== ===== test accuracy is  0.966 ==========
========== 800 ===== ===== test accuracy is  0.968 ==========
========== 820 ===== ===== test accuracy is  0.974 ==========
========== 840 ===== ===== test accuracy is  0.964 ==========
========== 860 ===== ===== test accuracy is  0.974 ==========
========== 880 ===== ===== test accuracy is  0.974 ==========
========== 900 ===== ===== test accuracy is  0.982 ==========
========== 920 ===== ===== test accuracy is  0.972 ==========
========== 940 ===== ===== test accuracy is  0.974 ==========
========== 960 ===== ===== test accuracy is  0.976 ==========
========== 980 ===== ===== test accuracy is  0.976 ==========
========== 1000 ===== ===== test accuracy is  0.984 ==========
========== 1020 ===== ===== test accuracy is  0.976 ==========
========== 1040 ===== ===== test accuracy is  0.976 ==========
========== 1060 ===== ===== test accuracy is  0.982 ==========
========== 1080 ===== ===== test accuracy is  0.974 ==========
========== 1100 ===== ===== test accuracy is  0.976 ==========
========== 1120 ===== ===== test accuracy is  0.974 ==========
========== 1140 ===== ===== test accuracy is  0.98 ==========
========== 1160 ===== ===== test accuracy is  0.98 ==========
========== 1180 ===== ===== test accuracy is  0.978 ==========

4.2 测试

def predict(test_x, idx):
    y_pre = cnn(test_x)
    print(y_pre.shape)
    _, pre_index = torch.max(y_pre, 1)
    prediction = pre_index.data.numpy()
    print(prediction)
    print("img: ", test_y[idx].data.numpy())

import matplotlib.pylab as plt

def showTorchImage(image):
    mode = transforms.ToPILImage()(image)
    plt.imshow(mode)
    plt.show()
    
idx = 18
showTorchImage(test_x[idx, :, :, :])
predict(test_x, idx)

torch.Size([50, 10])
[7 8 8 3 1 5 1 6 9 4 3 5 8 1 7 1 6 9 2 7 6 2 3 9 5 1 4 7 0 0 5 0 9 2 9 2 6
3 6 1 9 2 5 7 2 0 5 6 2 6]
img: 2