1、准备
引包:torchvision,包含transforms、model(在这指的就是网络的模型)、datasets(对数据的处理)
注:.compose,表示代码的顺序执行
2、制作好数据源
对数据的操作(包含裁剪、旋转等)不会储存在本地
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])
]),
}
transform.resize:改变图像的大小。注意,卷积网络中要求每一张图像大小都是一样的,不同图像的大小会影响网络的迭代速度
transforms.Normalize:对数据集的标准化操作「R,G,B」
设置datasets
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
datasets.ImageFolder(路径,数据名):图像数据的文件夹
with open('cat_to_name.json', 'r') as f:
cat_to_name = json.load(f)
读取标签实际对应的名字
2、加载models中提供的模型,并且直接用训练的好权重当做初始化参数
在这里,引入一个新的名词backbone,表示特征提取器
引入一个新的概念,迁移学习,将公开模型已经训练好的参数作为待训练模型参数的初始化。假如,我们使用除输出层之外的所有层的参数,就将这些参数“冻住”,在模型训练的过程中,不更新这些我参数。
model_name = 'resnet' #可选的比较多 ['resnet', 'alexnet', 'vgg', 'squeezenet', 'densenet', 'inception']
#是否用人家训练好的特征来做
feature_extract = True #都用人家特征,咱先不更新
feature_extract = True ,表示参数“冻住”,使用过程中不更新
# 是否用GPU训练
train_on_gpu = torch.cuda.is_available()
if not train_on_gpu:
print('CUDA is not available. Training on CPU ...')
else:
print('CUDA is available! Training on GPU ...')
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
固定代码,确定训练所使用的硬件GPU/CPU
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
设置模型的输出层
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)
训练输出层
3、优化器的设置(包含学习率的衰减)
# 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()
4、训练模块
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
model,使用训练的网络模型
dataloaders,训练设置的数据集
criterion,损失函数
optimizer,优化器
num-epochs,迭代数据
filename,最优训练结果以及相关参数存储的文件名称
解冻,训练所有参数
for param in model_ft.parameters():
param.requires_grad = True
# 再继续训练所有的参数,学习率调小一点
optimizer = optim.Adam(model_ft.parameters(), lr=1e-3)
scheduler = optim.lr_scheduler.StepLR(optimizer_ft, step_size=7, gamma=0.1)
# 损失函数
criterion = nn.CrossEntropyLoss()
checkpoint = torch.load(filename)
best_acc = checkpoint['best_acc']
model_ft.load_state_dict(checkpoint['state_dict'])
best-acc 模型训练的准确率
model_ft.load_state_dict(checkpoint['state_dict']) 当前的权重参数
5、加载训练好的模型,存储训练的权重参数、偏执、训练的准确率、损失
model_ft, input_size = initialize_model(model_name, 102, feature_extract, use_pretrained=True)
# GPU模式
model_ft = model_ft.to(device)
# 保存文件的名字
filename='best.pt'
# 加载模型
checkpoint = torch.load(filename)
best_acc = checkpoint['best_acc']
model_ft.load_state_dict(checkpoint['state_dict'])
6、展示预测结果
def im_convert(tensor):
""" 展示数据"""
image = tensor.to("cpu").clone().detach()
image = image.numpy().squeeze()
image = image.transpose(1,2,0)
image = image * np.array((0.229, 0.224, 0.225)) + np.array((0.485, 0.456, 0.406))
image = image.clip(0, 1)
return image
格式的转化:
GPU处理之后的数据格式为tensor,mathplot(用作数据的可视化)需要的数据类型为numpy
torch的数据为C*H*W,分别占「0,1,2」号位,PIL所需要的数据为H*W*C,transpose为调整参数的位置
fig=plt.figure(figsize=(20, 20))
columns =4
rows = 2
for idx in range (columns*rows):
ax = fig.add_subplot(rows, columns, idx+1, xticks=[], yticks=[])
plt.imshow(im_convert(images[idx]))
ax.set_title("{} ({})".format(cat_to_name[str(preds[idx])], cat_to_name[str(labels[idx].item())]),
color=("green" if cat_to_name[str(preds[idx])]==cat_to_name[str(labels[idx].item())] else "red"))
plt.show()