手写数字识别模型

import paddle
from paddle.nn import Linear
import paddle.nn.functional as F
import os
import numpy as np
import matplotlib.pyplot as plt

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])

plt.figure("Image")
plt.figure(figsize=(2,2))
plt.imshow(train_data0,cmap=plt.cm.binary)
plt.axis('on')
plt.title('image')
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的第一个图象，对应标签数字为{}".format(train_label_0))

class MNIST(paddle.nn.Layer):
    def __init__(self):
        super(MNIST,self).__init__()
        self.fc=paddle.nn.Linear(in_features=784,out_features=1)

    def forward(self,inputs):
        outputs=self.fc(inputs)
        return outputs

model=MNIST()
def train(modle):
    model.train()
    train_loader=paddle.io.DataLoader(paddle.vision.datasets.MNIST(mode='train'),batch_size=16,shuffle=True)
    opt=paddle.optimizeer.SGD(learning_rate=0.001,parameters=model.paramenters())

def norm_img(img):
    assert len(img.shape)==3
    batch_size,img_h,img_w=img.shape[0],img.shape[1],img.shape[2]
    img=img/255
    img=paddle.reshape(img,[batch_size,img_h*img_w])
    return img

paddle.vision.set_image_backend('cv2')
model=MNIST()

def train(model):
    model.train()
    train_loader=paddle.io.DataLoader(paddle.vision.datasets.MNIST(mode='train'),batch_size=16,shuffle=True)
    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')
            predicts=model(images)
            loss=F.square_error_cost(predicts,labels)
            avg_loss=paddle.mean(loss)
            
            if batch_id % 1000==0:
                print("epoch_id:{},batch_id:{},loss is:{}".format(epoch,batch_id,avg_loss.numpy()))

            avg_loss.backward()
            opt.step()
            opt.clear_grad()
train(model)
paddle.save(model.state_dict(),'./mnist.pdparams') 

from PIL import Image

img_path='./work/00.png'
im=Image.open('./work/00.png')
plt.imshow(im)
plt.show()
im=im.convert('L')
print('原始图像shape: ',np.array(im).shape)
im=im.resize((28,28),Image.ANTIALIAS)
plt.imshow(im)
plt.show()
print("采样后图片shape: ",np.array(im).shape)

def load_image(img_path):
    im=Image.open(img_path).convert('L')
    im=im.resize((28,28),Image.ANTIALIAS)
    im=np.array(im).reshape(1,-1).astype(np.float32)
    im=1-im/255
    return im

model=MNIST()
params_file_path='mnist.pdparams'
img_path='./work/00.png'
param_dict=paddle.load(params_file_path)
model.load_dict(param_dict)
model.eval()
tensor_img=load_image(img_path)
result=model(paddle.to_tensor(tensor_img))
print('result',result)
ensor_img=load_image(img_path)
result=model(paddle.to_tensor(tensor_img))
print('result',result)
print("本次预测的数字是",result.numpy().astype('int32'))

<Figure size 432x288 with 0 Axes>

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的第一个图象，对应标签数字为[5]
epoch_id:0,batch_id:0,loss is:[28.449667]
epoch_id:0,batch_id:1000,loss is:[2.2133603]
epoch_id:0,batch_id:2000,loss is:[2.6192307]
epoch_id:0,batch_id:3000,loss is:[1.4816303]
epoch_id:1,batch_id:0,loss is:[4.873765]
epoch_id:1,batch_id:1000,loss is:[1.1245084]
epoch_id:1,batch_id:2000,loss is:[4.7486277]
epoch_id:1,batch_id:3000,loss is:[8.636472]
epoch_id:2,batch_id:0,loss is:[3.22654]
epoch_id:2,batch_id:1000,loss is:[4.1829042]
epoch_id:2,batch_id:2000,loss is:[3.8323183]
epoch_id:2,batch_id:3000,loss is:[3.3910174]
epoch_id:3,batch_id:0,loss is:[5.5484705]
epoch_id:3,batch_id:1000,loss is:[2.9698162]
epoch_id:3,batch_id:2000,loss is:[2.0920122]
epoch_id:3,batch_id:3000,loss is:[2.552484]
epoch_id:4,batch_id:0,loss is:[4.1760187]
epoch_id:4,batch_id:1000,loss is:[2.0923305]
epoch_id:4,batch_id:2000,loss is:[4.3818398]
epoch_id:4,batch_id:3000,loss is:[1.7267109]
epoch_id:5,batch_id:0,loss is:[2.777513]
epoch_id:5,batch_id:1000,loss is:[4.043735]
epoch_id:5,batch_id:2000,loss is:[4.275708]
epoch_id:5,batch_id:3000,loss is:[3.8894782]
epoch_id:6,batch_id:0,loss is:[3.5191324]
epoch_id:6,batch_id:1000,loss is:[2.433545]
epoch_id:6,batch_id:2000,loss is:[3.819005]
epoch_id:6,batch_id:3000,loss is:[2.326075]
epoch_id:7,batch_id:0,loss is:[2.5925946]
epoch_id:7,batch_id:1000,loss is:[3.17057]
epoch_id:7,batch_id:2000,loss is:[3.6248827]
epoch_id:7,batch_id:3000,loss is:[5.20137]
epoch_id:8,batch_id:0,loss is:[4.69654]
epoch_id:8,batch_id:1000,loss is:[2.5298047]
epoch_id:8,batch_id:2000,loss is:[4.6738186]
epoch_id:8,batch_id:3000,loss is:[3.891779]
epoch_id:9,batch_id:0,loss is:[4.864161]
epoch_id:9,batch_id:1000,loss is:[3.8327315]
epoch_id:9,batch_id:2000,loss is:[3.3783822]
epoch_id:9,batch_id:3000,loss is:[3.8774865]

原始图像shape:  (329, 329)

采样后图片shape:  (28, 28)
result Tensor(shape=[1, 1], dtype=float32, place=CPUPlace, stop_gradient=False,
       [[2.93318319]])
本次预测的数字是 [[2]]