mseitzer / pytorch-fid

Compute FID scores with PyTorch.

Repository from Github https://github.commseitzer/pytorch-fidRepository from Github https://github.commseitzer/pytorch-fid

tifffile for .tiff support

cr458 opened this issue · comments

PIL has some issues with reading TIF files as it automatically converts to uint8. What do you think about adding something like:

if path.suffix == ".tif":
    img = tifffile.imread(path).astype(np.float32)
else:
    img = Image.open(path).convert('RGB') 

here. FWIW I've tested this modification locally and it's not a breaking change.

I would rather not add an additional dependency. It looks to me as if PIL supports TIF. Is it possible to avoid the conversion to uint8 by skipping .convert('RGB')? What format do your images have?

I think there's only support for single-band floating point image formats in PIL. My images are multi-band floating point images (remote sensing imagery).

From python-pillow/Pillow#1888, it looks like you are correct. I will look into supporting multi-band float images then (probably using imageio).