第四次作业：猫狗大战挑战赛

说明：在谷歌 Colab 上完成了猫狗大战VGG模型的迁移学习，在该代码的基础上，下载AI研习社“猫狗大战”比赛的测试集，利用fine-tune的VGG模型进行测试，按照比赛规定的格式输出，上传结果在线评测。

1.导包

import numpy as np
import matplotlib.pyplot as plt
import os
import torch
import torch.nn as nn
import torchvision
from torchvision import models,transforms,datasets
import time
import json


# 判断是否存在GPU设备
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print('Using gpu: %s ' % torch.cuda.is_available())

2.下载并处理数据集

datasets 是 torchvision 中的一个包，可以用来加载图像数据，torchvision 支持对输入数据进行一些复杂的预处理/变换

! wget http://fenggao-image.stor.sinaapp.com/dogscats.zip
! unzip dogscats.zip
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])

vgg_format = transforms.Compose([
                transforms.CenterCrop(224),
                transforms.ToTensor(),
                normalize,
            ])

data_dir = './dogscats'

dsets = {x: datasets.ImageFolder(os.path.join(data_dir, x), vgg_format)
         for x in ['train', 'valid']}

dset_sizes = {x: len(dsets[x]) for x in ['train', 'valid']}
dset_classes = dsets['train'].classes
loader_train = torch.utils.data.DataLoader(dsets['train'], batch_size=64, shuffle=True, num_workers=6)
loader_valid = torch.utils.data.DataLoader(dsets['valid'], batch_size=5, shuffle=False, num_workers=6)

3.导入VGG模型，修改并进行训练

目标是使用预训练好的模型，因此，需要把最后的Linear 层由1000类，替换为2类。为了在训练中冻结前面层的参数，需要设置 required_grad=False。这样，反向传播训练梯度时，前面层的权重就不会自动更新了。训练中，只会更新最后一层的参数。

model_vgg = models.vgg16(pretrained=True)
model_vgg_new = model_vgg;

for param in model_vgg_new.parameters():
    param.requires_grad = False
model_vgg_new.classifier._modules['6'] = nn.Linear(4096, 2)
model_vgg_new.classifier._modules['7'] = torch.nn.LogSoftmax(dim = 1)

model_vgg_new = model_vgg_new.to(device)
'''
第一步：创建损失函数和优化器

损失函数 NLLLoss() 的 输入 是一个对数概率向量和一个目标标签. 
它不会为我们计算对数概率，适合最后一层是log_softmax()的网络. 
'''
criterion = nn.NLLLoss()

# 学习率
lr = 0.001

# 随机梯度下降
optimizer_vgg = torch.optim.SGD(model_vgg_new.classifier[6].parameters(),lr = lr)

'''
第二步：训练模型
'''

def train_model(model,dataloader,size,epochs=1,optimizer=None):
    model.train()
    
    for epoch in range(epochs):
        running_loss = 0.0
        running_corrects = 0
        count = 0
        for inputs,classes in dataloader:
            inputs = inputs.to(device)
            classes = classes.to(device)
            outputs = model(inputs)
            loss = criterion(outputs,classes)           
            optimizer = optimizer
            optimizer.zero_grad()
            loss.backward()
            optimizer.step()
            _,preds = torch.max(outputs.data,1)
            # statistics
            running_loss += loss.data.item()
            running_corrects += torch.sum(preds == classes.data)
            count += len(inputs)
            print('Training: No. ', count, ' process ... total: ', size)
        epoch_loss = running_loss / size
        epoch_acc = running_corrects.data.item() / size
        print('Loss: {:.4f} Acc: {:.4f}'.format(
                     epoch_loss, epoch_acc))
        path = './models/model'+ '.pth'
        torch.save(model, path)
        
        
# 模型训练
train_model(model_vgg_new,loader_train,size=dset_sizes['train'], epochs=1, 
            optimizer=optimizer_vgg)

训练结果如下：

4.测试模型

def test_model(model,dataloader,size):
    model.eval()
    predictions = np.zeros(size)
    all_classes = np.zeros(size)
    all_proba = np.zeros((size,2))
    i = 0
    running_loss = 0.0
    running_corrects = 0
    for inputs,classes in dataloader:
        inputs = inputs.to(device)
        classes = classes.to(device)
        outputs = model(inputs)
        loss = criterion(outputs,classes)           
        _,preds = torch.max(outputs.data,1)
        # statistics
        running_loss += loss.data.item()
        running_corrects += torch.sum(preds == classes.data)
        predictions[i:i+len(classes)] = preds.to('cpu').numpy()
        all_classes[i:i+len(classes)] = classes.to('cpu').numpy()
        all_proba[i:i+len(classes),:] = outputs.data.to('cpu').numpy()
        i += len(classes)
        print('Testing: No. ', i, ' process ... total: ', size)        
    epoch_loss = running_loss / size
    epoch_acc = running_corrects.data.item() / size
    print('Loss: {:.4f} Acc: {:.4f}'.format(
                     epoch_loss, epoch_acc))
    return predictions, all_proba, all_classes
  
predictions, all_proba, all_classes = test_model(model_vgg_new,loader_valid,size=dset_sizes['valid'])

测试结果如下：

5.下载AI研习社上的测试集进行测试

! unzip './drive/My Drive/test.zip'
data_dir1 = './test'

dsets1 =datasets.ImageFolder(os.path.join(data_dir1), vgg_format)
         
dset_sizes1 =len(dsets1)
loader_test = torch.utils.data.DataLoader(dsets1, batch_size=1, shuffle=False, num_workers=6)
model_vgg_new = torch.load('./models/model.pth')
dic = {}
def test_model(model,dataloader,size):
    model.eval()
    i = 0
    for inputs,_ in dataloader:
        inputs = inputs.to(device)
        outputs = model(inputs)           
        _,preds = torch.max(outputs.data,1)
        key = dsets1.imgs[i][0].split("/")[-1].split('.')[0]
        dic[key] = preds[0].cpu().numpy()
        i += 1
        print('Testing: No. ', i, ' process ... total: ', size)        
  
test_model(model_vgg_new,loader_test,size=dset_sizes1)
with open("./csv/predictions.csv",'a+') as f:
    for i in range(2000):
        f.write("{},{}\n".format(i,dic[str(i)]))

提交结果

6.改进

将优化器改为Adam可以提高准确率



进一步改进可以改用Resnet的预训练模型，感兴趣的可以尝试一下。