识别手写体

最新推荐文章于 2022-12-04 14:37:52 发布

CQQa

最新推荐文章于 2022-12-04 14:37:52 发布

阅读量136

点赞数

文章标签： python pytorch 神经网络

本文链接：https://blog.csdn.net/xchc00/article/details/120853235

版权

该博客通过飞桨(PaddlePaddle)框架搭建了一个简单的全连接网络，用于识别MNIST数据集的手写数字。首先展示了MNIST数据集中的一个样本图像，然后定义了网络结构，包括一个全连接层。接着进行模型训练，观察训练过程中损失(loss)的变化。最后，对本地的一张图像进行预处理和预测，结果显示预测的数字为3。

摘要由CSDN通过智能技术生成

#加载飞桨和相关类库
import paddle
from paddle.nn import Linear
import paddle.nn.functional as F
import os
import numpy as np
import matplotlib.pyplot as plt

# 设置数据读取器，API自动读取MNIST数据训练集
train_dataset = paddle.vision.datasets.MNIST(mode='train')

train_data0 = np.array(train_dataset[0][0])
train_label_0 = np.array(train_dataset[0][1])

# 显示第一batch的第一个图像
import matplotlib.pyplot as plt
plt.figure("Image") # 图像窗口名称
plt.figure(figsize=(2,2))
plt.imshow(train_data0, cmap=plt.cm.binary)
plt.axis('on') # 关掉坐标轴为 off
plt.title('image') # 图像题目
plt.show()

print("图像数据形状和对应数据为:", train_data0.shape)
print("图像标签形状和对应数据为:", train_label_0.shape, train_label_0)
print("\n打印第一个batch的第一个图像，对应标签数字为{}".format(train_label_0))

<Figure size 432x288 with 0 Axes>

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-0ehJeyEi-1634644142909)(output_2_1.png)]

图像数据形状和对应数据为: (28, 28)
图像标签形状和对应数据为: (1,) [5]

打印第一个batch的第一个图像，对应标签数字为[5]

# 定义mnist数据识别网络结构，同房价预测网络
class MNIST(paddle.nn.Layer):
    def __init__(self):
        super(MNIST, self).__init__()
        
        # 定义一层全连接层，输出维度是1
        self.fc = paddle.nn.Linear(in_features=784, out_features=1)
        
    # 定义网络结构的前向计算过程
    def forward(self, inputs):
        outputs = self.fc(inputs)
        return outputs

# 声明网络结构
model = MNIST()

def train(model):
    # 启动训练模式
    model.train()
    # 加载训练集 batch_size 设为 16
    train_loader = paddle.io.DataLoader(paddle.vision.datasets.MNIST(mode='train'), 
                                        batch_size=16, 
                                        shuffle=True)
    # 定义优化器，使用随机梯度下降SGD优化器，学习率设置为0.001
    opt = paddle.optimizer.SGD(learning_rate=0.001, parameters=model.parameters())

# 图像归一化函数，将数据范围为[0, 255]的图像归一化到[0, 1]
def norm_img(img):
    # 验证传入数据格式是否正确，img的shape为[batch_size, 28, 28]
    assert len(img.shape) == 3
    batch_size, img_h, img_w = img.shape[0], img.shape[1], img.shape[2]
    # 归一化图像数据
    img = img / 255
    # 将图像形式reshape为[batch_size, 784]
    img = paddle.reshape(img, [batch_size, img_h*img_w])
    
    return img

import paddle
# 确保从paddle.vision.datasets.MNIST中加载的图像数据是np.ndarray类型
paddle.vision.set_image_backend('cv2')

# 声明网络结构
model = MNIST()

def train(model):
    # 启动训练模式
    model.train()
    # 加载训练集 batch_size 设为 16
    train_loader = paddle.io.DataLoader(paddle.vision.datasets.MNIST(mode='train'), 
                                        batch_size=16, 
                                        shuffle=True)
    # 定义优化器，使用随机梯度下降SGD优化器，学习率设置为0.001
    opt = paddle.optimizer.SGD(learning_rate=0.001, parameters=model.parameters())
    EPOCH_NUM = 10
    for epoch in range(EPOCH_NUM):
        for batch_id, data in enumerate(train_loader()):
            images = norm_img(data[0]).astype('float32')
            labels = data[1].astype('float32')
            
            #前向计算的过程
            predicts = model(images)
            
            # 计算损失
            loss = F.square_error_cost(predicts, labels)
            avg_loss = paddle.mean(loss)
            
            #每训练了1000批次的数据，打印下当前Loss的情况
            if batch_id % 1000 == 0:
                print("epoch_id: {}, batch_id: {}, loss is: {}".format(epoch, batch_id, avg_loss.numpy()))
            
            #后向传播，更新参数的过程
            avg_loss.backward()
            opt.step()
            opt.clear_grad()
            
train(model)
paddle.save(model.state_dict(), './mnist.pdparams')

epoch_id: 0, batch_id: 0, loss is: [41.803543]
epoch_id: 0, batch_id: 1000, loss is: [7.3164177]
epoch_id: 0, batch_id: 2000, loss is: [6.408638]
epoch_id: 0, batch_id: 3000, loss is: [2.1390135]
epoch_id: 1, batch_id: 0, loss is: [1.8769809]
epoch_id: 1, batch_id: 1000, loss is: [2.778306]
epoch_id: 1, batch_id: 2000, loss is: [2.4463768]
epoch_id: 1, batch_id: 3000, loss is: [1.80804]
epoch_id: 2, batch_id: 0, loss is: [3.4746327]
epoch_id: 2, batch_id: 1000, loss is: [3.3961694]
epoch_id: 2, batch_id: 2000, loss is: [3.7104187]
epoch_id: 2, batch_id: 3000, loss is: [3.2753563]
epoch_id: 3, batch_id: 0, loss is: [3.4888704]
epoch_id: 3, batch_id: 1000, loss is: [4.4290376]
epoch_id: 3, batch_id: 2000, loss is: [4.4603386]
epoch_id: 3, batch_id: 3000, loss is: [1.787259]
epoch_id: 4, batch_id: 0, loss is: [3.2406425]
epoch_id: 4, batch_id: 1000, loss is: [2.8450933]
epoch_id: 4, batch_id: 2000, loss is: [4.1710615]
epoch_id: 4, batch_id: 3000, loss is: [2.5388634]
epoch_id: 5, batch_id: 0, loss is: [4.191299]
epoch_id: 5, batch_id: 1000, loss is: [1.888926]
epoch_id: 5, batch_id: 2000, loss is: [5.320978]
epoch_id: 5, batch_id: 3000, loss is: [3.8174102]
epoch_id: 6, batch_id: 0, loss is: [7.071102]
epoch_id: 6, batch_id: 1000, loss is: [1.8865409]
epoch_id: 6, batch_id: 2000, loss is: [1.2453774]
epoch_id: 6, batch_id: 3000, loss is: [4.2274723]
epoch_id: 7, batch_id: 0, loss is: [4.345709]
epoch_id: 7, batch_id: 1000, loss is: [1.8568087]
epoch_id: 7, batch_id: 2000, loss is: [1.2210071]
epoch_id: 7, batch_id: 3000, loss is: [4.8859897]
epoch_id: 8, batch_id: 0, loss is: [5.9070034]
epoch_id: 8, batch_id: 1000, loss is: [2.4232588]
epoch_id: 8, batch_id: 2000, loss is: [5.0883164]
epoch_id: 8, batch_id: 3000, loss is: [3.7737527]
epoch_id: 9, batch_id: 0, loss is: [7.1347823]
epoch_id: 9, batch_id: 1000, loss is: [2.5092049]
epoch_id: 9, batch_id: 2000, loss is: [2.7596505]
epoch_id: 9, batch_id: 3000, loss is: [2.2518513]

# 导入图像读取第三方库
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image

img_path = './work/3.jpg'
# 读取原始图像并显示
im = Image.open('./work/3.jpg')
plt.imshow(im)
plt.show()
# 将原始图像转为灰度图
im = im.convert('L')
print('原始图像shape: ', np.array(im).shape)
# 使用Image.ANTIALIAS方式采样原始图片
im = im.resize((28, 28), Image.ANTIALIAS)
plt.imshow(im)
plt.show()
print("采样后图片shape: ", np.array(im).shape)

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-YpjHp6R0-1634644142916)(output_7_0.png)]

原始图像shape:  (900, 900)

[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-odVffokD-1634644142919)(output_7_2.png)]

采样后图片shape:  (28, 28)

# 读取一张本地的样例图片，转变成模型输入的格式
def load_image(img_path):
    # 从img_path中读取图像，并转为灰度图
    im = Image.open(img_path).convert('L')
    # print(np.array(im))
    im = im.resize((28, 28), Image.ANTIALIAS)
    im = np.array(im).reshape(1, -1).astype(np.float32)
    # 图像归一化，保持和数据集的数据范围一致
    im = 1 - im / 255
    return im

# 定义预测过程
model = MNIST()
params_file_path = 'mnist.pdparams'
img_path = './work/3.jpg'
# 加载模型参数
param_dict = paddle.load(params_file_path)
model.load_dict(param_dict)
# 灌入数据
model.eval()
tensor_img = load_image(img_path)
result = model(paddle.to_tensor(tensor_img))
print('result',result)
#  预测输出取整，即为预测的数字，打印结果
_img))
print('result',result)
#  预测输出取整，即为预测的数字，打印结果
print("本次预测的数字是", result.numpy().astype('int32'))

result Tensor(shape=[1, 1], dtype=float32, place=CPUPlace, stop_gradient=False,
       [[3.18773866]])
本次预测的数字是 [[3]]

CQQa

关注

0
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
复制链接

分享到 QQ

分享到新浪微博

扫一扫