python之pytorch手写数字图片识别的实现方法

内容摘要
这篇文章主要为大家详细介绍了python之pytorch手写数字图片识别的实现方法，具有一定的参考价值，可以用来参考一下。
文章正文

这篇文章主要为大家详细介绍了python之pytorch手写数字图片识别的实现方法，具有一定的参考价值，可以用来参考一下。

数据集：MNIST数据集，代码中会自动下载，不用自己手动下载。数据集很小，不需要GPU设备，可以很好的体会到pytorch的魅力。模型+训练+预测程序：

代码如下：

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
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
<code>
import torch
from torch import nn
from torch.nn import functional as F
from torch import optim
import torchvision
from matplotlib import pyplot as plt
from utils import plot_image, plot_curve, one_hot
 
# step1  load dataset
batch_size = 512
train_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST('mnist_data', train=True, download=True,
                               transform=torchvision.transforms.Compose([
                                   torchvision.transforms.ToTensor(),
                                   torchvision.transforms.Normalize(
                                       (0.1307,), (0.3081,)
                                   )
                               ])),
    batch_size=batch_size, shuffle=True)
test_loader = torch.utils.data.DataLoader(
    torchvision.datasets.MNIST('mnist_data/', train=False, download=True,
                               transform=torchvision.transforms.Compose([
                                   torchvision.transforms.ToTensor(),
                                   torchvision.transforms.Normalize(
                                       (0.1307,), (0.3081,)
                                   )
                               ])),
    batch_size=batch_size, shuffle=False)
x , y = next(iter(train_loader))
print(x.shape, y.shape, x.min(), x.max())
plot_image(x, y, "image_sample")
 
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
 
        self.fc1 = nn.Linear(28*28, 256)
        self.fc2 = nn.Linear(256, 64)
        self.fc3 = nn.Linear(64, 10)
    def forward(self, x):
        # x: [b, 1, 28, 28]
        # h1 = relu(xw1 + b1)
        x = F.relu(self.fc1(x))
        # h2 = relu(h1w2 + b2)
        x = F.relu(self.fc2(x))
        # h3 = h2w3 + b3
        x = self.fc3(x)
 
        return x
net = Net()
optimizer = optim.SGD(net.parameters(), lr=0.01, momentum=0.9)
 
train_loss = []
for epoch in range(3):
    for batch_idx, (x, y) in enumerate(train_loader):
        #加载进来的图片是一个四维的tensor，x: [b, 1, 28, 28], y:[512]
        #但是我们网络的输入要是一个一维向量（也就是二维tensor)，所以要进行展平操作
        x = x.view(x.size(0), 28*28)
        #  [b, 10]
        out = net(x)
        y_onehot = one_hot(y)
        # loss = mse(out, y_onehot)
        loss = F.mse_loss(out, y_onehot)
 
        optimizer.zero_grad()
        loss.backward()
        # w' = w - lr*grad
        optimizer.step()
 
        train_loss.append(loss.item())
 
        if batch_idx % 10 == 0:
            print(epoch, batch_idx, loss.item())
 
plot_curve(train_loss)
    # we get optimal [w1, b1, w2, b2, w3, b3]
 
 
total_correct = 0
for x,y in test_loader:
    x = x.view(x.size(0), 28*28)
    out = net(x)
    # out: [b, 10]
    pred = out.argmax(dim=1)
    correct = pred.eq(y).sum().float().item()
    total_correct += correct
total_num = len(test_loader.dataset)
acc = total_correct/total_num
print("acc:", acc)
 
x, y = next(iter(test_loader))
out = net(x.view(x.size(0), 28*28))
pred = out.argmax(dim=1)
plot_image(x, pred, "test")</code>

pytorch实现手写数字图片识别

主程序中调用的函数（注意命名为utils)：

代码如下：

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
<code>
import  torch
from    matplotlib import pyplot as plt
 
 
def plot_curve(data):
    fig = plt.figure()
    plt.plot(range(len(data)), data, color='blue')
    plt.legend(['value'], loc='upper right')
    plt.xlabel('step')
    plt.ylabel('value')
    plt.show()
 
 
def plot_image(img, label, name):
 
    fig = plt.figure()
    for i in range(6):
        plt.subplot(2, 3, i + 1)
        plt.tight_layout()
        plt.imshow(img[i][0]*0.3081+0.1307, cmap='gray', interpolation='none')
        plt.title("{}: {}".format(name, label[i].item()))
        plt.xticks([])
        plt.yticks([])
    plt.show()
 
 
def one_hot(label, depth=10):
    out = torch.zeros(label.size(0), depth)
    idx = torch.LongTensor(label).view(-1, 1)
    out.scatter_(dim=1, index=idx, value=1)
    return out</code>

pytorch实现手写数字图片识别

打印出损失下降的曲线图：

训练3个epoch之后，在测试集上的精度就可以89%左右，可见模型的准确度还是很不错的。输出六张测试集的图片以及预测结果：

六张图片的预测全部正确。

以上就是本文的全部内容，希望对大家的学习有所帮助，也希望大家多多支持php教程。

注：关于python之pytorch手写数字图片识别的实现方法的内容就先介绍到这里，更多相关文章的可以留意
代码注释
上一篇Python之PyTorch 设置随机数种子使结果可复现的解决办法下一篇返回列表

作者：喵哥笔记

IDC笔记

学的不仅是技术，更是梦想！