图像识别模型与训练策略

图像预处理

1.需要将图像Resize到相同大小输入到卷积网络中
2.翻转、裁剪、色彩偏移等操作
3.转化为Tensor数据格式
4.对RGB三种颜色通道进行标准化

data_transforms = {
    'train': 
        transforms.Compose([
        transforms.Resize([96, 96]),
        transforms.RandomRotation(45),#随机旋转，-45到45度之间随机选
        transforms.CenterCrop(64),#从中心开始裁剪
        transforms.RandomHorizontalFlip(p=0.5),#随机水平翻转 选择一个概率概率
        transforms.RandomVerticalFlip(p=0.5),#随机垂直翻转
        transforms.ColorJitter(brightness=0.2, contrast=0.1, saturation=0.1, hue=0.1),#参数1为亮度，参数2为对比度，参数3为饱和度，参数4为色相
        transforms.RandomGrayscale(p=0.025),#概率转换成灰度率，3通道就是R=G=B
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])#均值，标准差
    ]),
    'valid': 
        transforms.Compose([
        transforms.Resize([64, 64]),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
}

读取数据

将训练集中各个类别文件夹中的数据经过Transforms增强后进行统一读取封装

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

batch_size = 128

image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x), data_transforms[x]) for x in ['train', 'valid']}
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=batch_size, shuffle=True) for x in ['train', 'valid']}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'valid']}
class_names = image_datasets['train'].classes

迁移学习

使用官方发布的模型和参数，将参数冻住不更新

def set_parameter_requires_grad(model, feature_extracting):
    if feature_extracting:
        for param in model.parameters():
            param.requires_grad = False

model_ft = models.resnet18()#18层的能快点，条件好点的也可以选152
model_ft

修改输出层

def initialize_model(model_name, num_classes, feature_extract, use_pretrained=True):
    
    model_ft = models.resnet18(pretrained=use_pretrained)
    set_parameter_requires_grad(model_ft, feature_extract)
    
    num_ftrs = model_ft.fc.in_features
    model_ft.fc = nn.Linear(num_ftrs, 102)#类别数自己根据自己任务来
                            
    input_size = 64#输入大小根据自己配置来

    return model_ft, input_size

更新输出层参数

model_ft, input_size = initialize_model(model_name, 102, feature_extract, use_pretrained=True)

#GPU还是CPU计算
model_ft = model_ft.to(device)

# 模型保存，名字自己起
filename='checkpoint.pth'

# 是否训练所有层
params_to_update = model_ft.parameters()
print("Params to learn:")
if feature_extract:
    params_to_update = []
    for name,param in model_ft.named_parameters():
        if param.requires_grad == True:
            params_to_update.append(param)
            print("\t",name)
else:
    for name,param in model_ft.named_parameters():
        if param.requires_grad == True:
            print("\t",name)

优化器设置

optimizer_ft = optim.Adam(params_to_update, lr=1e-2)#要训练啥参数，你来定
scheduler = optim.lr_scheduler.StepLR(optimizer_ft, step_size=10, gamma=0.1)#学习率每7个epoch衰减成原来的1/10
criterion = nn.CrossEntropyLoss()

训练策略

def train_model(model, dataloaders, criterion, optimizer, num_epochs=25,filename='best.pt'):
    #咱们要算时间的
    since = time.time()
    #也要记录最好的那一次
    best_acc = 0
    #模型也得放到你的CPU或者GPU
    model.to(device)
    #训练过程中打印一堆损失和指标
    val_acc_history = []
    train_acc_history = []
    train_losses = []
    valid_losses = []
    #学习率
    LRs = [optimizer.param_groups[0]['lr']]
    #最好的那次模型，后续会变的，先初始化
    best_model_wts = copy.deepcopy(model.state_dict())
    #一个个epoch来遍历
    for epoch in range(num_epochs):
        print('Epoch {}/{}'.format(epoch, num_epochs - 1))
        print('-' * 10)

        # 训练和验证
        for phase in ['train', 'valid']:
            if phase == 'train':
                model.train()  # 训练
            else:
                model.eval()   # 验证

            running_loss = 0.0
            running_corrects = 0

            # 把数据都取个遍
            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)#放到你的CPU或GPU
                labels = labels.to(device)

                # 清零
                optimizer.zero_grad()
                # 只有训练的时候计算和更新梯度
                outputs = model(inputs)
                loss = criterion(outputs, labels)
                _, preds = torch.max(outputs, 1)
                # 训练阶段更新权重
                if phase == 'train':
                    loss.backward()
                    optimizer.step()

                # 计算损失
                running_loss += loss.item() * inputs.size(0)#0表示batch那个维度
                running_corrects += torch.sum(preds == labels.data)#预测结果最大的和真实值是否一致
                
            
            
            epoch_loss = running_loss / len(dataloaders[phase].dataset)#算平均
            epoch_acc = running_corrects.double() / len(dataloaders[phase].dataset)
            
            time_elapsed = time.time() - since#一个epoch我浪费了多少时间
            print('Time elapsed {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
            print('{} Loss: {:.4f} Acc: {:.4f}'.format(phase, epoch_loss, epoch_acc))
            

            # 得到最好那次的模型
            if phase == 'valid' and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())
                state = {
                  'state_dict': model.state_dict(),#字典里key就是各层的名字，值就是训练好的权重
                  'best_acc': best_acc,
                  'optimizer' : optimizer.state_dict(),
                }
                torch.save(state, filename)
            if phase == 'valid':
                val_acc_history.append(epoch_acc)
                valid_losses.append(epoch_loss)
                #scheduler.step(epoch_loss)#学习率衰减
            if phase == 'train':
                train_acc_history.append(epoch_acc)
                train_losses.append(epoch_loss)
        
        print('Optimizer learning rate : {:.7f}'.format(optimizer.param_groups[0]['lr']))
        LRs.append(optimizer.param_groups[0]['lr'])
        print()
        scheduler.step()#学习率衰减

    time_elapsed = time.time() - since
    print('Training complete in {:.0f}m {:.0f}s'.format(time_elapsed // 60, time_elapsed % 60))
    print('Best val Acc: {:4f}'.format(best_acc))

    # 训练完后用最好的一次当做模型最终的结果,等着一会测试
    model.load_state_dict(best_model_wts)
    return model, val_acc_history, train_acc_history, valid_losses, train_losses, LRs