Data Preprocessing for Tiny ImageNet

vision
1

Is there any good methods of data preprocessing for tiny imagenet? It seems the data augmentation methods for imagenet does not work well for tiny imagenet. Here is my code:

    normalize = transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
    train_dataset = datasets.ImageFolder(
        train_dir,
        transforms.Compose([
            transforms.RandomResizedCrop(224),
            transforms.RandomHorizontalFlip(),
            transforms.ToTensor(),
            normalize
        ]))
    train_loader = torch.utils.data.DataLoader(
        train_dataset,
        batch_size=args.batch_size, shuffle=True, **kwargs)
    valid_loader = torch.utils.data.DataLoader(
        datasets.ImageFolder(valid_dir, transforms.Compose([
            transforms.Resize(256),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            normalize
        ])),
        batch_size=args.test_batch_size, shuffle=False, **kwargs)
2

I was also wondering if there is an accepted standard data augmentation procedure for Tiny ImageNet?

@deJQK tiny ImageNet images are 64x 64 so taking crops of 224 pixels, or resizing to 256 is probably not such a great idea.