pytorch猫狗大战项目代码

                               pytorch猫狗大战-项目代码

各位看官老爷，小白我知道pytorch的这个猫狗大战的代码真是漫天都是，这基本上也就是一个hello world的程序。我这个代码也是从《pytorch 机器学习从入门到实践》一书中摘出来了。对于初学者来说（像小白我）这本书还是不错的，难能可贵的是，它良心的在github上给出了代码。如果各位看官也有这本书的话，一定不要直接敲书中的代码，而是要结合着github上的代码再敲。因为这本书中的印刷的代码有缩进错误和一些疏漏的地方。有时照着这个书中的代码敲是会报错的。书中的代码只是一个核心代码的参考。我这也是敲过之后的感悟，现在就直接贴出这个hello word代码，本人亲测是可以运行的通的。感兴趣的大神就运行一下吧。

0. 前提：

已经下载了kaggle中的猫狗数据集，这个注册是需要翻墙的，但是好在国内的网盘中有很多已经下好的数据集可以直接使用。

1. 数据预处理代码

将下载好的数据集中的trainzip处理成一个小的数据集，并且区分出train和test比例为0.95和0.05。并将其中的猫狗图像进行归一化处理

import os, shutil
import numpy as np
import pdbrandom_state = 42
np.random.seed(random_state)original_dataset_dir = &＃39;./data/CAT-DOG/train&＃39; #这是我的train的路径可以换成你的
total_num = int(len(os.listdir(original_dataset_dir)) / 2)
random_idx = np.array(range(total_num))
np.random.shuffle(random_idx)
base_dir = &＃39;cats_and_dogs_small&＃39;
if not os.path.exists(base_dir):os.mkdir(base_dir)sub_dirs = [&＃39;train&＃39;, &＃39;test&＃39;]
animals = [&＃39;cats&＃39;, &＃39;dogs&＃39;]
train_idx = random_idx[: int(total_num * 0.9)]
test_idx = random_idx[int(total_num * 0.9) :]
numbers = [train_idx, test_idx]
for idx, sub_dir in enumerate(sub_dirs):dir = os.path.join(base_dir, sub_dir)if not os.path.exists(dir):os.mkdir(dir)for animal in animals:animal_dir = os.path.join(dir, animal)if not os.path.exists(animal_dir):os.mkdir(animal_dir)fnames = [animal[: -1]+ &＃39;.{}.jpg&＃39;.format(i) for i in numbers[idx]]for fname in fnames:src = os.path.join(original_dataset_dir, fname)dst = os.path.join(animal_dir,fname)shutil.copyfile(src, dst)print(dir + &＃39;total images : %d&＃39;%(len(os.listdir(animal_dir))))

 2. 自己设计的网络中的训练和测试代码

这部分代码在书中是存在部分问题的，但是我这是修改后的能够运行的代码

from __future__ import print_function, division
import shutil
import torch
import os
import torch.nn as nn
import torch.nn.functional as F
from torch.autograd import Variable
import numpy as np
from torch.utils.data import Dataset, DataLoader
from torchvision import transforms, datasets, utils
from torch.utils.data import DataLoader
import torch.optim as optimtorch.manual_seed(1)
epochs = 10
batch_size = 4
num_workers = 4
use_gpu = torch.cuda.is_available()data_transform = transforms.Compose([transforms.Scale(256), transforms.CenterCrop(224), transforms.ToTensor(), transforms.Normalize(mean = [0.485, 0.456, 0.406], std = [0.229, 0.224, 0.225])])train_dataset = datasets.ImageFolder(root = &＃39;cats_and_dogs_small/train/&＃39;, transform = data_transform)
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size = batch_size, shuffle = True, num_workers = num_workers)
test_dataset = datasets.ImageFolder(root = &＃39;cats_and_dogs_small/test/&＃39;, transform=data_transform)
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size = batch_size, shuffle=True, num_workers = num_workers)class Net(nn.Module):def __init__(self):super(Net, self).__init__()self.conv1 = nn.Conv2d(3, 6, 5)self.maxpool = nn.MaxPool2d(2, 2)self.conv2 = nn.Conv2d(6, 16, 5)self.fc1 = nn.Linear(16 * 53 * 53, 1024)self.fc2 = nn.Linear(1024, 512)self.fc3 = nn.Linear(512, 2)def forward(self, x):x = self.maxpool(F.relu(self.conv1(x)))x = self.maxpool(F.relu(self.conv2(x)))x = x.view(-1, 16 * 53 * 53)x = F.relu(self.fc1(x))x = F.relu(self.fc2(x))x = self.fc3(x)return x
if use_gpu:net = Net().cuda()
else:net = Net()
print(net)cirterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.0001, momentum=0.9)net.train()
for epoch in range(epochs):running_loss = 0.0train_correct = 0train_total = 0for i, data in enumerate(train_loader, 0):inputs, train_labels = dataif use_gpu:inputs, labels = Variable(inputs.cuda()), Variable(train_labels.cuda())else:inputs, labels = Variable(inputs), Variable(train_labels)optimizer.zero_grad()outputs = net(inputs)_, train_predicted = torch.max(outputs.data, 1)train_correct += (train_predicted == labels.data).sum()loss = cirterion(outputs, labels)loss.backward()optimizer.step()running_loss += loss.item()train_total += train_labels.size(0)print(&＃39;train %d epoch loss: %.3f acc: %.3f &＃39;% (epoch + 1, running_loss / train_total, 100 * train_correct / train_total))correct = 0test_loss = 0.0test_total = 0net.eval()for data in test_loader:images, labels = dataif use_gpu:images, labels = Variable(images.cuda()), Variable(labels.cuda())else:images, labels = Variable(images), Variable(labels)outputs = net(images)_, predicted = torch.max(outputs.data, 1)loss = cirterion(outputs, labels)test_loss += loss.item()test_total += labels.size(0)correct += (predicted == labels.data).sum()print(&＃39;test %d epoch loss: %.3f acc: %.3f&＃39; % (epoch + 1, test_loss / test_total, 100 * correct / test_total))

2. 迁移学习用resnet网络进行训练和预测的代码

# coding=utf-8from __future__ import print_function, division
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.autograd import Variable
from torch.utils.data import Dataset
from torchvision import transforms, datasets, models# 配置参数
random_state = 1
torch.manual_seed(random_state) # 设置随机数种子，确保结果可重复
torch.cuda.manual_seed(random_state)
torch.cuda.manual_seed_all(random_state)
np.random.seed(random_state)
# random.seed(random_state)epochs = 10 # 训练次数
batch_size = 4 # 批处理大小
num_workers = 4 # 多线程的数目
use_gpu = torch.cuda.is_available()
# 对加载的图像作归一化处理， 并裁剪为[224x224x3]大小的图像
data_transform = transforms.Compose([transforms.Scale(256),transforms.CenterCrop(224),transforms.ToTensor(),transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])# 数据的批处理，尺寸大小为batch_size,
# 在训练集中，shuffle 必须设置为True, 表示次序是随机的
train_dataset = datasets.ImageFolder(root=&＃39;cats_and_dogs_small/train/&＃39;, transform=data_transform)
train_loader = torch.utils.data.DataLoader(train_dataset, batch_size=batch_size, shuffle=True, num_workers=num_workers)test_dataset = datasets.ImageFolder(root=&＃39;cats_and_dogs_small/test/&＃39;, transform=data_transform)
test_loader = torch.utils.data.DataLoader(test_dataset, batch_size=batch_size, shuffle=True, num_workers=num_workers)# 创建模型class Net(nn.Module):def __init__(self):super(Net, self).__init__()self.conv1 = nn.Conv2d(3, 6, 5)self.maxpool = nn.MaxPool2d(2, 2)self.conv2 = nn.Conv2d(6, 16, 5)self.fc1 = nn.Linear(16 * 53 * 53, 1024)self.fc2 = nn.Linear(1024, 512)self.fc3 = nn.Linear(512, 2)def forward(self, x):x = self.maxpool(F.relu(self.conv1(x)))x = self.maxpool(F.relu(self.conv2(x)))x = x.view(-1, 16 * 53 * 53)x = F.relu(self.fc1(x))x = F.relu(self.fc2(x))x = self.fc3(x)return x# 加载resnet18 模型，
net = models.resnet18(pretrained=False)
num_ftrs = net.fc.in_features
net.fc = nn.Linear(num_ftrs, 2) # 更新resnet18模型的fc模型，if use_gpu:net = net.cuda()
print(net)
# 定义loss和optimizer
cirterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(), lr=0.0001, momentum=0.9)# 开始训练
net.train()
for epoch in range(epochs):running_loss = 0.0train_correct = 0train_total = 0for i, data in enumerate(train_loader, 0):inputs, train_labels = dataif use_gpu:inputs, labels = Variable(inputs.cuda()), Variable(train_labels.cuda())else:inputs, labels = Variable(inputs), Variable(train_labels)# inputs, labels = Variable(inputs), Variable(train_labels)optimizer.zero_grad()outputs = net(inputs)_, train_predicted = torch.max(outputs.data, 1)# import pdb# pdb.set_trace()train_correct += (train_predicted == labels.data).sum()loss = cirterion(outputs, labels)loss.backward()optimizer.step()running_loss += loss.item()train_total += train_labels.size(0)print(&＃39;train %d epoch loss: %.3f acc: %.3f &＃39; % (epoch + 1, running_loss / train_total, 100 * train_correct / train_total))# 模型测试correct = 0test_loss = 0.0test_total = 0test_total = 0net.eval()for data in test_loader:images, labels = dataif use_gpu:images, labels = Variable(images.cuda()), Variable(labels.cuda())else:images, labels = Variable(images), Variable(labels)outputs = net(images)_, predicted = torch.max(outputs.data, 1)loss = cirterion(outputs, labels)test_loss += loss.item()test_total += labels.size(0)correct += (predicted == labels.data).sum()print(&＃39;test %d epoch loss: %.3f acc: %.3f &＃39; % (epoch + 1, test_loss / test_total, 100 * correct / test_total))

 希望各位看官敲代码愉快！