where were they from to views and locations? were they from f"{prefix}_images.npy"?

Question

where were they from to views and locations? were they from f"{prefix}_images.npy"?

Closed this issue 2 years ago · 2 comments

(val_view, (views, locations)), labels = inputs

Answer 1 · 2022-12-14T01:22:28.000Z

https://github.com/facebookresearch/VICRegL/blob/main/transforms.py
the 333th line

Answer 2 · 2022-12-20T13:32:04.000Z

Hi, the views are the transformed images, loaded from ImageNet, and the locations are the coordinates maps of the crops which are generated by the _location_to_NxN_grid function from the crop parameters randomly sampled in the RandomResizedCropWithLocation class.

The train_images.npy and val_images.npy are just ImageNet stored in numpy tensor files.