Antoine Simoulin

Hey @hyperkai, thank you for opening this issue. After reviewing the details, I suggest we consolidate the discussion with #9065, as it appears to be closely related.

Hey @jiangxingxian thank you for posting, and apologies for the delayed response. Could you please provide a minimal reproducible example so that I can replicate the issue on my end?...

Hey @abionics, Thanks for reporting the issue, and apologies for the delayed response. After investigating, I suspect this behavior is due to how GIF animations are optimized. Many GIFs store...

Just merged the fix in #9241. Thanks @sg3-141-592 for submitting the PR and @abionics for opening and discussing the issue!

@justincdavis could you complete the missing Contributor License Agreement (c.f. earlier comment from the [meta-cla](https://github.com/apps/meta-cla) bot)?

@vedantdalimkar thanks for posting! Can you give more detail about the bug regarding the `ignore_index` indices in the labels and propose a short code snippet on the current state and...

@vedantdalimkar, after going through the codebase it seems [`MixUp`](https://github.com/pytorch/vision/blob/main/torchvision/transforms/v2/_augment.py#L222C7-L222C12) and [`CutMix`](https://github.com/pytorch/vision/blob/main/torchvision/transforms/v2/_augment.py#L270) transformations only take `tv_tensors.Image`, `tv_tensors.Video`, and entry marked as "labels" as input. If those transformations are indeed used for...

Hi @rumpg, Thanks for posting. It seems to me the fact that returns `{'image_id': image_id}` is returned for empty targets is probably due to this [line](https://github.com/pytorch/vision/blob/main/torchvision/tv_tensors/_dataset_wrapper.py#L377). However, in your case,...

Hey @siemdejong thanks for your detailed post. I understand the limitations. At the same times, I suspect changing structuring assumptions such as the number of channels per images may have...

Hey @siemdejong, after internal discussion, we believe that the hypothesis regarding the number of channels is a fundamental aspect of torchvision’s design. Allowing images with an arbitrary number of channels...