End-to-end video instance segmentation with transformers

[jamesp]

For those of us interested in segmentation, particularly with correlated sequential images (video), this paper seems fascinating and I wanted to mention it here since I haven’t seen it discussed.