Torchvision/ImageFolder меняет количество каналов при чтении изображений ⇐ Python
-
Гость
Torchvision/ImageFolder меняет количество каналов при чтении изображений
I have my data (MNIST Images) for training and testing in separate folders and then each class in a different folder. Now I am trying to load the data using the ImgageFolder method. But when I try the Dataloader, the shape of each element in each batch is (3, 28, 28) instead of (28, 28). All images are of standard size (28, 28). Would someone explain why it is changing the depth from 1 to 3?
train_transform = transforms.Compose([transforms.ToTensor()]) valid_transform = transforms.Compose([transforms.ToTensor()]) train_dataset = torchvision.datasets.ImageFolder(root="MNISTImages/training", transform=train_transform) valid_dataset = torchvision.datasets.ImageFolder(root="MNISTImages/testing", transform=valid_transform) train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True) next(iter(train_loader))[1].shape
Источник: https://stackoverflow.com/questions/780 ... ing-images
I have my data (MNIST Images) for training and testing in separate folders and then each class in a different folder. Now I am trying to load the data using the ImgageFolder method. But when I try the Dataloader, the shape of each element in each batch is (3, 28, 28) instead of (28, 28). All images are of standard size (28, 28). Would someone explain why it is changing the depth from 1 to 3?
train_transform = transforms.Compose([transforms.ToTensor()]) valid_transform = transforms.Compose([transforms.ToTensor()]) train_dataset = torchvision.datasets.ImageFolder(root="MNISTImages/training", transform=train_transform) valid_dataset = torchvision.datasets.ImageFolder(root="MNISTImages/testing", transform=valid_transform) train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True) next(iter(train_loader))[1].shape
Источник: https://stackoverflow.com/questions/780 ... ing-images
