cifar數據集訓練與測試
CIFAR數據集下載
https://download.csdn.net/download/wangxiaobei2017/12474160
# !/usr/bin/env python
# -*- coding: utf-8 -*-
import torch
import torch.nn as nn
import torch.nn.functional as F
import torchvision
import torchvision.transforms as transforms
import torch.optim as optim
import cv2
from PIL import Image
import numpy as np
class LeNet(nn.Module):
# 一般在__init__中定義網絡需要的操作算子，比如卷積、全連接算子等等
def __init__(self):
super(LeNet, self).__init__()
# Conv2d的第一個參數是輸入的channel數量，第二個是輸出的channel數量，第三個是kernel size
self.conv1 = nn.Conv2d(3, 6, 5)
self.conv2 = nn.Conv2d(6, 16, 5)
# 由於上一層有16個channel輸出，每個feature map大小為5*5，所以全連接層的輸入是16*5*5
self.fc1 = nn.Linear(16*5*5, 120)
self.fc2 = nn.Linear(120, 84)
# 最終有10類，所以最后一個全連接層輸出數量是10
self.fc3 = nn.Linear(84, 10)
self.pool = nn.MaxPool2d(2, 2)
# forward這個函數定義了前向傳播的運算，只需要像寫普通的python算數運算那樣就可以了
def forward(self, x):
x = F.relu(self.conv1(x))
x = self.pool(x)
x = F.relu(self.conv2(x))
x = self.pool(x)
# 下面這步把二維特征圖變為一維，這樣全連接層才能處理
x = x.view(-1, 16*5*5)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
# 在構建數據集的時候指定transform，就會應用我們定義好的transform
# root是存儲數據的文件夾，download=True指定如果數據不存在先下載數據
class ImageClassify(object):
def __init__(self):
self.cifar_train = 0
self.cifar_test = 0
self.trainloader =0
self.testloader =0
self.criterion = 0
self.optimizer = 0
self.net = LeNet()
def GetImgData(self):
# cifar-10官方提供的數據集是用numpy array存儲的
# 下面這個transform會把numpy array變成torch tensor，然后把rgb值歸一到[0, 1]這個區間
print('獲取數據集....')
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
self.cifar_train = torchvision.datasets.CIFAR10(root='E:\\2019PythonProject\\Pytorch_step1\\data', train=True,download=False, transform=transform)
self.cifar_test = torchvision.datasets.CIFAR10(root='E:\\2019PythonProject\\Pytorch_step1\\data', train=False,transform=transform)
print(self.cifar_train)
print(self.cifar_test)
def LoadData(self):
# 加載數據集
print("加載數據集...")
batch_size = 32
self.trainloader = torch.utils.data.DataLoader(self.cifar_train, batch_size=batch_size, shuffle=True)
self.testloader = torch.utils.data.DataLoader(self.cifar_test, batch_size=batch_size, shuffle=True)
def LossFun(self):
print("lossing...")
# CrossEntropyLoss就是我們需要的損失函數
self.criterion = nn.CrossEntropyLoss()
self.optimizer = optim.SGD(self.net.parameters(), lr=0.001, momentum=0.9)
def TrainingData(self):
self.LoadData()
self.LossFun()
print("Start Training...")
for epoch in range(30):
# 我們用一個變量來記錄每100個batch的平均loss
loss100 = 0.0
# 我們的dataloader派上了用場
for i, data in enumerate(self.trainloader):
inputs, labels = data
# inputs, labels = inputs.to(device), labels.to(device) # 注意需要復制到GPU
self.optimizer.zero_grad()
# print(len(inputs))
outputs = self.net.forward(inputs)
# print(len(outputs))
loss = self.criterion(outputs, labels)
loss.backward()
self.optimizer.step()
loss100 += loss.item()
if i % 2 == 0:
print('[Epoch %d, Batch %5d] loss: %.3f' %
(epoch + 1, i + 1, loss100 / 2))
loss100 = 0.0
# 保存網絡模型 保存整個模型
torch.save(self.net, 'model_shanbu_128.pkl')
print("Done Training!")
def TestingData(self):
model_net = torch.load('model_shanbu_128.pkl')
self.LoadData()
# 構造測試的dataloader
dataiter = iter(self.testloader)
# 預測正確的數量和總數量
correct = 0
total = 0
# 使用torch.no_grad的話在前向傳播中不記錄梯度，節省內存
# cv2.namedWindow('predictPic', cv2.WINDOW_NORMAL)
to_pil_image = transforms.ToPILImage()
with torch.no_grad():
for images, labels in dataiter:
# images, labels = data
# print(images)
print(len(images.data))
# images, labels = images.to(device), labels.to(device)
# 預測
# outputs = self.net(images)
outputs = model_net(images)
# 我們的網絡輸出的實際上是個概率分布，去最大概率的哪一項作為預測分類
_, predicted = torch.max(outputs.data, 1)
total += labels.size(0)
correct += (predicted == labels).sum().item()
# print(images.data[0])
# print(len(images.data[0]))
# input_flag = input()
# if input_flag == 'p':
# break
# elif input_flag == 'c':
# continue
# cv2.imshow('predictPic', images)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
print('Accuracy of the self.network on the 10000 test images: %d %%' % (
100 * correct / total))
def main():
ImgCla = ImageClassify()
ImgCla.GetImgData()
ImgCla.TrainingData()
ImgCla.TestingData()
pass
if __name__ == '__main__':
main()
下載數據后更改配置，將下載路徑改成自己的路徑
