Transforms Pipeline

[arbelhizmi]

Hi,

Regarding this pipeline:
transform = transforms.Compose( [ transforms.Resize(224), transforms.ToTensor(), transforms.Normalize(mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225)), ] ) - is it assuming that images are PIL in uint8 format? did you mean that images suppose to be in range [0, 1] before applying the transform.Normalize with the mean and std?

I'm using Xenium datasets and i want to extract features from pretrained UNIv2 model, but my images are already in size 224, already torch.float32 and already tensors (automatically converted to tensors by DataLoader), so i'm wondering what you originally intended that the pipeline should be? because from what i understand UNIv2 is partially trained on Xenium datasets as well, and Xenium datasets usually are not cropped as uint8 or PIL, unless you specifically crop them like this, so i'm not really sure what transformations should i use here..?

thanks