pytorch實現個人手寫數字識別

網上的大多數例子都是基於Mnist數據集進行測試的，今天實現一個自己手寫數字的識別。

 

首先訓練模型，使用Mnist數據集，網絡的backbone採用LeNet。

1. 導入需要的模塊並添加GPU設備

import torch
import torchvision as tv
import torchvision.transforms as transforms
import torch.nn as nn
import torch.optim as optim
import cv2

# 定義是否使用GPU
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

2. 定義網絡結構

class LeNet(nn.Module):
    def __init__(self):
        super(LeNet, self).__init__()
        self.conv1 = nn.Sequential(  # input_size=(1*28*28)
            nn.Conv2d(1, 6, 5, 1, 2),  # padding=2保證輸入輸出尺寸相同
            nn.ReLU(),  # input_size=(6*28*28)
            nn.MaxPool2d(kernel_size=2, stride=2),  # output_size=(6*14*14)
        )
        self.conv2 = nn.Sequential(
            nn.Conv2d(6, 16, 5),
            nn.ReLU(),  # input_size=(16*10*10)
            nn.MaxPool2d(2, 2)  # output_size=(16*5*5)
        )
        self.fc1 = nn.Sequential(
            nn.Linear(16 * 5 * 5, 120),
            nn.ReLU()
        )
        self.fc2 = nn.Sequential(
            nn.Linear(120, 84),
            nn.ReLU()
        )
        self.fc3 = nn.Linear(84, 10)

    # 定義前向傳播過程，輸入爲x
    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        # nn.Linear()的輸入輸出都是維度爲一的值，所以要把多維度的tensor展平成一維（一行）
        x = x.view(x.size()[0], -1)
        x = self.fc1(x)
        x = self.fc2(x)
        x = self.fc3(x)
        return x

3. 設置超參數和定義訓練和測試數據提取器

# 超參數設置
EPOCH = 10 # 遍歷數據集次數
BATCH_SIZE = 256  # 批處理尺寸(batch_size)
LR = 0.001  # 學習率

# 定義數據預處理方式
transform = transforms.ToTensor()

# 定義訓練數據集
trainset = tv.datasets.MNIST(
    root='./data/',
    train=True,
    download=False,
    transform=transform)

# 定義訓練批處理數據
trainloader = torch.utils.data.DataLoader(
    trainset,
    batch_size=BATCH_SIZE,
    shuffle=True,
)

# 定義測試數據集
testset = tv.datasets.MNIST(
    root='./data/',
    train=False,
    download=False,
    transform=transform)

4. 定義訓練函數

def train():
    # 定義損失函數loss function 和優化方式（採用SGD）
    net = LeNet().to(device)
    criterion = nn.CrossEntropyLoss()  # 交叉熵損失函數，通常用於多分類問題上
    optimizer = optim.SGD(net.parameters(), lr=LR, momentum=0.9)
    for epoch in range(EPOCH):
        sum_loss = 0.0
        # 數據讀取
        for i, data in enumerate(trainloader):
            inputs, labels = data
            inputs, labels = inputs.to(device), labels.to(device)

            # 梯度清零
            optimizer.zero_grad()

            # forward + backward
            outputs = net(inputs)
            loss = criterion(outputs, labels)
            loss.backward()
            optimizer.step()

            # 每訓練100個batch打印一次平均loss
            sum_loss += loss.item()
            if i % 100 == 99:
                print('[%d, %d] loss: %.03f'
                      % (epoch + 1, i + 1, sum_loss / 100))
                sum_loss = 0.0
        # 每跑完一次epoch測試一下準確率
        with torch.no_grad():
            correct = 0
            total = 0
            for data in testloader:
                images, labels = data
                images, labels = images.to(device), labels.to(device)
                outputs = net(images)
                # 取得分最高的那個類
                _, predicted = torch.max(outputs.data, 1)
                total += labels.size(0)
                correct += (predicted == labels).sum()
            print('第%d個epoch的識別準確率爲：%d%%' % (epoch + 1, (100 * correct / total)))
        # 保存模型參數
        torch.save(net.state_dict(), './params.pth')

5. 先進行訓練，訓練結果會保存在params.pth中。

if __name__ == "__main__":
    train()

6. 訓練完成後註釋掉訓練函數，讀取訓練好的模型參數並進行測試。

# 讀取訓練好的網絡參數
net = LeNet().to(device)
a = torch.load('./params.pth')
net.load_state_dict(torch.load('./params.pth'))


if __name__ == "__main__":
    # train()
    img = cv2.imread('./2.png', cv2.IMREAD_GRAYSCALE) #讀取圖片
    img = cv2.resize(img,(28, 28))  # 調整圖片爲28*28
    img = torch.from_numpy(img).float()
    img = img.view(1, 1, 28, 28)
    img = img.to(device)
    outputs = net(img)
    _, predicted = torch.max(outputs.data, 1)
    print(predicted.to('cpu').numpy().squeeze())

測試圖片使用windows軟件畫圖繪製，如下：

輸出結果如下：

完整代碼如下：

import torch
import torchvision as tv
import torchvision.transforms as transforms
import torch.nn as nn
import torch.optim as optim
import cv2

# 定義是否使用GPU
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")


# 定義網絡結構
class LeNet(nn.Module):
    def __init__(self):
        super(LeNet, self).__init__()
        self.conv1 = nn.Sequential(  # input_size=(1*28*28)
            nn.Conv2d(1, 6, 5, 1, 2),  # padding=2保證輸入輸出尺寸相同
            nn.ReLU(),  # input_size=(6*28*28)
            nn.MaxPool2d(kernel_size=2, stride=2),  # output_size=(6*14*14)
        )
        self.conv2 = nn.Sequential(
            nn.Conv2d(6, 16, 5),
            nn.ReLU(),  # input_size=(16*10*10)
            nn.MaxPool2d(2, 2)  # output_size=(16*5*5)
        )
        self.fc1 = nn.Sequential(
            nn.Linear(16 * 5 * 5, 120),
            nn.ReLU()
        )
        self.fc2 = nn.Sequential(
            nn.Linear(120, 84),
            nn.ReLU()
        )
        self.fc3 = nn.Linear(84, 10)

    # 定義前向傳播過程，輸入爲x
    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)
        # nn.Linear()的輸入輸出都是維度爲一的值，所以要把多維度的tensor展平成一維（一行）
        x = x.view(x.size()[0], -1)
        x = self.fc1(x)
        x = self.fc2(x)
        x = self.fc3(x)
        return x




# 使得我們能夠手動輸入命令行參數，就是讓風格變得和Linux命令行差不多
# parser = argparse.ArgumentParser()
# parser.add_argument('--outf', default='./model/', help='folder to output images and model checkpoints')  # 模型保存路徑
# parser.add_argument('--net', default='./model/net.pth', help="path to netG (to continue training)")  # 模型加載路徑
# opt = parser.parse_args()

# 超參數設置
EPOCH = 10 # 遍歷數據集次數
BATCH_SIZE = 256  # 批處理尺寸(batch_size)
LR = 0.001  # 學習率

# 定義數據預處理方式
transform = transforms.ToTensor()

# 定義訓練數據集
trainset = tv.datasets.MNIST(
    root='./data/',
    train=True,
    download=False,
    transform=transform)

# 定義訓練批處理數據
trainloader = torch.utils.data.DataLoader(
    trainset,
    batch_size=BATCH_SIZE,
    shuffle=True,
)

# 定義測試數據集
testset = tv.datasets.MNIST(
    root='./data/',
    train=False,
    download=False,
    transform=transform)

# 定義測試批處理數據
testloader = torch.utils.data.DataLoader(
    testset,
    batch_size=BATCH_SIZE,
    shuffle=False,
)

# 定義損失函數loss function 和優化方式（採用SGD）
net = LeNet().to(device)
a = torch.load('./params.pth')
net.load_state_dict(torch.load('./params.pth'))
criterion = nn.CrossEntropyLoss()  # 交叉熵損失函數，通常用於多分類問題上
optimizer = optim.SGD(net.parameters(), lr=LR, momentum=0.9)

# 訓練並保存模型參數
def train():

    for epoch in range(EPOCH):
        sum_loss = 0.0
        # 數據讀取
        for i, data in enumerate(trainloader):
            inputs, labels = data
            inputs, labels = inputs.to(device), labels.to(device)

            # 梯度清零
            optimizer.zero_grad()

            # forward + backward
            outputs = net(inputs)
            loss = criterion(outputs, labels)
            loss.backward()
            optimizer.step()

            # 每訓練100個batch打印一次平均loss
            sum_loss += loss.item()
            if i % 100 == 99:
                print('[%d, %d] loss: %.03f'
                      % (epoch + 1, i + 1, sum_loss / 100))
                sum_loss = 0.0
        # 每跑完一次epoch測試一下準確率
        with torch.no_grad():
            correct = 0
            total = 0
            for data in testloader:
                images, labels = data
                images, labels = images.to(device), labels.to(device)
                outputs = net(images)
                # 取得分最高的那個類
                _, predicted = torch.max(outputs.data, 1)
                total += labels.size(0)
                correct += (predicted == labels).sum()
            print('第%d個epoch的識別準確率爲：%d%%' % (epoch + 1, (100 * correct / total)))
        # 保存模型參數
        torch.save(net.state_dict(), './params.pth')

if __name__ == "__main__":
    # train()
    img = cv2.imread('./2.png', cv2.IMREAD_GRAYSCALE)
    img = cv2.resize(img,(28, 28))
    img = torch.from_numpy(img).float()
    img = img.view(1, 1, 28, 28)
    img = img.to(device)
    outputs = net(img)
    _, predicted = torch.max(outputs.data, 1)
    print(predicted.to('cpu').numpy().squeeze())