Does the model learn before or after transformations?

Hey I noticed my model learns a lot more when I’m passing big images although my transformations are the following:

train_trans = transforms.Compose([
    transforms.Resize(256),
    transforms.RandomCrop(_image_size),
    transforms.RandomHorizontalFlip(),
    transforms.ToTensor(),
    transforms.Normalize(_mean, _std),
])

When I feed 256x256 images directly, the model doesn’t learn much. This leads me to believe, something is learned prior to transformations. Can anyone confirm? Thanks

@alx
The learning happens on variables ion the model. If there are transformations that have learnable Variables then it would lead to some learning

No. That can’t happen. The model learns better if the you have ample data which is given by data augmentations.