stack expects each tensor to be equal size, but got [3, 224, 224] at entry 0 and [1,224,224] at entr

首先可以确定地是在数据读取的以及处理发生的问题。目前可以归纳为一下两种：

1：使用Image.open（）没有添加.convert('RGB')

错误代码例子：

class Mydataset(data.Dataset):
    def __init__(self,root,labels,transform):
        super(Mydataset,self).__init__()
        self.imgs_path = root
        self.labels = labels
        self.transform = transform
    def __getitem__(self,index):
        ig_path = self.imgs_path[index]
        label=self.labels[index]
        ######################################
        pil_image = Image.open(ig_path)
        #########################################
        data = self.transform(pil_image)
        return data,label
    def __len__(self):
        return len(self.imgs_path)

正确代码例子：

class Mydataset(data.Dataset):
    def __init__(self,root,labels,transform):
        super(Mydataset,self).__init__()
        self.imgs_path = root
        self.labels = labels
        self.transform = transform
    def __getitem__(self,index):
        ig_path = self.imgs_path[index]
        label=self.labels[index]
        ######################################
        pil_image = Image.open(ig_path).convert('RGB')
        #########################################
        data = self.transform(pil_image)
        return data,label
    def __len__(self):
        return len(self.imgs_path)

2：transform过程中错误使用Resize功能

错误示范：

transform = transforms.Compose([
    transforms.Resize((224)),
    transforms.ToTensor()
])

正确代码：

train_dataset = datasets.ImageFolder(
    train_data,
    transforms.Compose([
        transforms.Resize((224，224))

 验证方法：

path_load = glob.glob(r'D:\BaiduNetdiskDownload\pytorch_learning\dataset\dataset2\*.jpg')
#图片路径
transform = transforms.Compose([
    transforms.Resize((224,224)),
    transforms.ToTensor()
])

all_labels=[]
species = ['cloudy', 'rain', 'shine', 'sunrise']
#图片类别
for img in path_load:
    for i,c in enumerate(species):
        if c in img:
            all_labels.append(i)
#图片标签            
species_to_idx = dict((c, i) for i, c in enumerate(species))          
label_to_class =  dict((v,k) for k,v in species_to_idx.items())           
class Mydataset(data.Dataset):
    def __init__(self,root,labels,transform):
        super(Mydataset,self).__init__()
        self.imgs_path = root
        self.labels = labels
        self.transform = transform
    def __getitem__(self,index):
        ig_path = self.imgs_path[index]
        label=self.labels[index]
        pil_image = Image.open(ig_path).convert('RGB')
        data = self.transform(pil_image)
        return data,label
    def __len__(self):
        return len(self.imgs_path)


wheather_dataset = Mydataset(path_load,all_labels,transform)
wheather_dl = data.DataLoader(wheather_dataset,
                              batch_size=16,
                             shuffle=True,
                             drop_last=True)



plt.figure(figsize=(12,8))
imgs_batch,labels_batch=next(iter(wheather_dl))
for i,(img,label) in enumerate(zip(imgs_batch,labels_batch)):
        img = img.permute(1,2,0).numpy()
        plt.subplot(4,4,i+1)
        plt.title(label_to_class.get(label.item()))
        plt.imshow(img)

result：