src/dataset.py

import torch
from torchvision import datasets, transforms
from torch.utils.data import DataLoader

def create_dataloaders(data_dir, batch_size=8, num_workers=2):
    transform = transforms.Compose([
        transforms.Resize((512, 512)),  # keep 512p
        transforms.Grayscale(num_output_channels=3), # convert to 3-ch for pre-trained ViT
        transforms.ToTensor()  # handle scaling
        # Remove Normalize since doing reconstruction, Autoencoder -> target to be in same range as output (0,1)
        #transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) # Std ViT normalization (ImageNet)
    ])

    # Train only on 'normal' images
    train_dataset = datasets.ImageFolder(root=f"{data_dir}/train", transform=transform)
    test_dataset = datasets.ImageFolder(root=f"{data_dir}/test", transform=transform)

    train_loader = DataLoader(train_dataset, batch_size=batch_size, num_workers=num_workers, shuffle=True)
    test_loader = DataLoader(test_dataset, batch_size=batch_size, num_workers=num_workers, shuffle=False)

    return train_loader, test_loader