项目一手写数字识别

做个小项目的目的是熟悉paddle基本框架以及如何导入数据
所涉及到的paddle api

1、模型架构

Linear层的导入

from paddle.nn import Linear

class Regressor(paddle.nn.Layer):
	def __init__(self):
		super(Regressor,self).__init__()
		self.fc=Linear(in_features=13,out_features=1)
	def forward(self,inputs):
		x=self.fc(inputs)
		return x

Linear层的说明：
in_features (int): The number of input units.
out_features (int): The number of output units.
# Define the linear layer.
weight_attr = paddle.ParamAttr(
name=“weight”,
initializer=paddle.nn.initializer.Constant(value=0.5))
bias_attr = paddle.ParamAttr(
name=“bias”,
initializer=paddle.nn.initializer.Constant(value=1.0))
linear = paddle.nn.Linear(2, 4, weight_attr=weight_attr, bias_attr=bias_attr)
# linear.weight: [[0.5 0.5 0.5 0.5]
# [0.5 0.5 0.5 0.5]]
# linear.bias: [1. 1. 1. 1.]

2、数据导入

dataloader

paddle.vision.set_image_backend('cv2') # 必须加
train_loader=paddle.io.DataLoader(paddle.vision.datasets.MNIST(mode='train'),
	batch_size=16,shuffle=True)

其中，该loader接收PIL等数据类型的

3、训练策略

优化器

opt=paddle.optimizer.SGD(learning_rate=1e-3,parameters=model.parameters())

4、损失函数

import paddle.nn.functional as F
loss=F.square_error_cost(predict,label)

5、模型权重保存与加载

paddle.save(model.state_dict(),'xxx.pdparams')

params_file_path='xxx.pdparams'
param_dict=paddle.load(params_file_path)
model=MODEL()
model.load_dict(params_dict)

6、其余边边角角

（1）to tensor

tensor_result=paddle.to_tensor(tensor_img)

(2）reshape

img=paddle.reshape(img,[batch_size,img_h*img_W])

(3) 求平均

avg_loss=paddle.mean(loss)

最后，综合上述所有的API，我们搭建一套简易版的项目代码

import paddle
import paddle.nn.functional as F
import os
import numpy as np

# 数据处理
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

class MNIST_module(paddle.nn.Layer):
	def __init__(self):
		super(MNIST_module,self).__init__()
		self.fc=paddle.nn.Linear(in_features=784,out_features=1)
	def forward(self,inputs):
		outptus=self.fc(inputs)
		return outputs

def train(model):
	model.train()
	paddle.vision.set_image_backend('cv2')
	train_loader=paddle.io.DataLoader(paddle.vision.datasets.MNIST(mode='train'),batch_size=16,shuffle=True)
	opt=paddle.optimizer.SGD(learning_rate=1e-3,parameters=model.parameters())
	EPOCH_NUM=10
	for epoch in range(EPOCH_NUM):
		for batch_id,data in enumerate(train_loader):
			images=norm_img(data)
			labels=data[1].astype('float32')
			predicts=model(images)
			loss=F.square_error_cost(predicts,labels)
			avg_loss=paddle.mean(loss)
			# 训练1000批次打印
            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()
model=MNIST_module()
train(model)
paddle.save(model.state_dict(),'mnist.pdparams')

测试

import matplotlib.pyplot as plt
import numpy as np
from PIL import Image
import paddle
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/256
	return im

img_path='./1.png'
tensor_img=load_image(img_path)
params_file_path='mnist.pdparams'
params_dict=paddle.load(params_file_path)
model.load_dict(params_dict)
model.eval()
result=model(paddle.to_tensor(tensor_img))
print("本次预测的数字是",result.numpy().astype('int32'))