When trying to implement, one of the steps is to resize each image to (448, 448). But even with the transform applied the Dataloader throw an exception about difference of size in the dataset.

The exact error message: “RuntimeError: each element in list of batch should be of equal size”

from torchvision import datasets
from torchvision.transforms import v2, ToTensor

from torch.utils.data import DataLoader

validation_data = datasets.voc.VOCDetection(
    root='.DATA/',
    download=False,
    image_set="val",
    transform=v2.Compose([ v2.Resize(size=(448, 448)), ToTensor() ])
)

batch_size = 64

validation_dataloader = DataLoader(validation_data, batch_size=batch_size)

for X, y in validation_dataloader:
    print(f"Shape of X [N, C, H, W]: {X.shape}")
    break