bhack
bhack
> If the reason for removing third_party/cub was that it was considered stable and unchanging, then this could be a good reasons to bring back third_party/cub as third_party/cccl then. I...
Or probably it is older coming from caffe2 https://github.com/facebookarchive/caffe2/pull/382
Relative challenges: https://taodataset.org/workshop/cvpr23/ https://codalab.lisn.upsaclay.fr/competitions/10703
Ok it is reasonable. Have you adopted any other migration about the motion blur problem?
Probably you could add some dataset like Youtube VIS long, like in that paper, to evaluate the robustness of you appearance based approach on fast moving objects.
Yes I have the same issue. Do you think we could introduce some safety check in the bbox prompt? I think fixing bbox it could be enough as the current...
It seems that currently the restore `restore_model` logic is available only on eval: https://github.com/SysCV/sam-hq/blob/322488826bda616798901c6280d13a9a90444ae7/train/train.py#L361-L373
@lkeab Can you share the sam-hq only network checkpoint so that we could eventually restore it for the train?
Resizing it is not working well with large images with small targets. https://github.com/SysCV/sam-hq/blob/ca530af6d6c8717e33704b07f4f3c6c8e778b0ff/segment_anything/predictor.py#LL31C26-L31C43 Skipping the resizing on ` ResizeLongestSide` it is working well instead but then the output mask is...
Not I still had that issue.