BruceFan's Blog

Stay hungry, stay foolish

0%

卷积神经网络

发表于 Edited on

关于卷积神经网络的原理这篇文章先不做介绍，推荐机器视角：长文揭秘图像处理和卷积神经网络架构和卷积：如何成为一个很厉害的神经网络这两篇文章，这里记录一下PyTorch对卷积神经网络的实现。

代码清单 model.py

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
import torch.nn as nn

# 卷积神经网络(两个卷积层)
class ConvNet(nn.Module):
    def __init__(self, num_classes=10):
        super(ConvNet, self).__init__()
        self.layer1 = nn.Sequential(
            nn.Conv2d(1, 16, kernel_size=5, stride=1, padding=2),
            nn.BatchNorm2d(16),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2))
        self.layer2 = nn.Sequential(
            nn.Conv2d(16, 32, kernel_size=5, stride=1, padding=2),
            nn.BatchNorm2d(32),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2))
        self.fc = nn.Linear(7*7*32, num_classes) # full connection，即输出层

    def forward(self, x):
        out = self.layer1(x)
        out = self.layer2(out)
        out = out.reshape(out.size(0), -1)
        out = self.fc(out)
        return out

这里首先构建了一个两层的卷积神经网络，torch.nn.Squential是一个顺序的容器，模块按照构造函数顺序加入。
torch.nn.Conv2d是将2D卷积（卷积核是2D的）应用到输入，其参数依次为：

  • in_channels 输入图像的通道数
  • out_channels 卷积产生的通道数
  • kernel_size 卷积核大小
  • stride 卷积的步长
  • padding 在输入的各边0填充

输入为一个图像的Tensor形式，其中包含$N$词袋大小，$C$是通道数，$H$是输入的高度，$W$是输入的宽度，单位是pixel。下面的demo.py中可以看到一个图像的Tensor如何构造。
torch.nn.BatchNorm2dBatch Normalization对每个隐藏层的输入进行标准化，有加速收敛等好处。
torch.nn.ReLU是一个非线性激活函数，$ReLU(x) = max(0, x)$正数不变，负数都转化为0。
torch.nn.MaxPool2d是对输入进行一个2D最大池化。
torch.nn.Linear对输入的数据进行一个线性转变$y = Ax + b$，参数依次为：

  • in_features 输入样本的大小
  • out_features 输出样本的大小，这里是10个数字
  • bias

代码清单 train.py

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
from model import ConvNet

# Device configuration
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')

# Hyper parameters
num_epochs = 5
num_classes = 10
batch_size = 100
learning_rate = 0.001

# MNIST dataset
train_dataset = torchvision.datasets.MNIST(root='./data/',
                                           train=True,
                                           transform=transforms.ToTensor(),
                                           download=True)

test_dataset = torchvision.datasets.MNIST(root='./data/',
                                          train=False,
                                          transform=transforms.ToTensor())

# Data loader
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
                                           batch_size=batch_size,
                                           shuffle=True)

test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
                                          batch_size=batch_size,
                                          shuffle=False)

model = ConvNet(num_classes).to(device)

# Loss and optimizer
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

# Train the model
total_step = len(train_loader)
for epoch in range(num_epochs):
    for i, (images, labels) in enumerate(train_loader):
        images = images.to(device)
        labels = labels.to(device)

        # Forward pass
        outputs = model(images)
        loss = criterion(outputs, labels)

        # Backward and optimize
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if (i+1) % 100 == 0:
            print ('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}'
                   .format(epoch+1, num_epochs, i+1, total_step, loss.item()))

# Test the model
model.eval()  # eval mode (batchnorm uses moving mean/variance instead of mini-batch mean/variance)
with torch.no_grad():
    correct = 0
    total = 0
    for images, labels in test_loader:
        images = images.to(device)
        labels = labels.to(device)
        outputs = model(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()

    print('Test Accuracy of the model on the 10000 test images: {} %'.format(100 * correct / total))

# Save the model checkpoint
torch.save(model.state_dict(), 'model.ckpt')
1
2
3
4
5
6
7
8
$ python train.py
Epoch [1/5], Step [100/600], Loss: 0.1743
Epoch [1/5], Step [200/600], Loss: 0.1452
Epoch [1/5], Step [300/600], Loss: 0.1029
Epoch [1/5], Step [400/600], Loss: 0.0549
Epoch [1/5], Step [500/600], Loss: 0.0608
...
Test Accuracy of the model on the 10000 test images: 99 %

这里首先介绍一下训练数据，PyTorch支持多种数据集，上述代码用到的是MNIST数据集，一般在训练和测试时都是直接使用官网提供的二进制数据集，但是作为初学者，不太理解这个二进制到底是什么东西，所以先用下面的代码将二进制文件解析为原始的图片和对应的标签：
代码清单 resolve.py

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
from PIL import Image
import struct

def read_image(filename):
  f = open(filename, 'rb')
  index = 0
  buf = f.read()
  f.close()
  magic, images, rows, columns = struct.unpack_from('>IIII' , buf , index)
  index += struct.calcsize('>IIII')
  for i in xrange(images):
    image = Image.new('L', (columns, rows))
    for x in xrange(rows):
      for y in xrange(columns):
        image.putpixel((y, x), int(struct.unpack_from('>B', buf, index)[0]))
        index += struct.calcsize('>B')

    print 'save ' + str(i) + 'image'
    image.save('test/' + str(i) + '.png')

def read_label(filename, saveFilename):
  f = open(filename, 'rb')
  index = 0
  buf = f.read()

  f.close()

  magic, labels = struct.unpack_from('>II' , buf , index)
  index += struct.calcsize('>II')

  labelArr = [0] * labels

  for x in xrange(labels):
    labelArr[x] = int(struct.unpack_from('>B', buf, index)[0])
    index += struct.calcsize('>B')

  save = open(saveFilename, 'w')

  save.write(','.join(map(lambda x: str(x), labelArr)))
  save.write('\n')

  save.close()
  print 'save labels success'

if __name__ == '__main__':
  read_image('data/raw/t10k-images-idx3-ubyte')
  read_label('data/raw/t10k-labels-idx1-ubyte', 'test/label.txt')

解析出的图片
解析出数据集中的图片后，我们可以随意选取图片来测试训练的模型model.ckpt：
代码清单 demo.py

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
import numpy as np
import torch
import torchvision.transforms as transforms
from model import ConvNet
from PIL import Image


model_path = "./model.ckpt"
img_path = "./12.png"

model = ConvNet()
print "load pretrained model from %s" % model_path
model.load_state_dict(torch.load(model_path))

transformer = transforms.ToTensor()
# 将图像转换为灰度模式，即单通道
image = Image.open(img_path).convert('L')
#image.resize((28, 28), Image.BILINEAR)
# 将图像转换为Tensor
image = transformer(image)
# 为Tensor添加一维，表示batch
image = image.view(1, *image.size())

model.eval()
output = model(image)

preds = torch.max(output, 1)[1]

print preds.item()
1
2
3
$ python demo.py
load pretrained model from ./model.ckpt
9

可以看到用训练的模型可以正确识别出图片中的数字9。
reference
https://github.com/yunjey/pytorch-tutorial/blob/master/tutorials/02-intermediate/convolutional_neural_network/main.py