How are standard data formats cocovid used for video instance segmentation?How to mark?

[int-gif]

[Seerkfang]

CocoVideoDataset is designed for coco-like data formats, but with video input. So plus to the original coco annotation, it has video-related annotation like video_id or frame_index. We've implemented YouTubeVISDataset, you can take it as an example.

[tctco]

Have you figured out how to create a customized CocoVideoDataset? Is there a tool to create such datasets, or are there any tools to create video instance segmentation datasets like youtube-vis?

[meikorol]

Have you gotten this？ I also have no idea on VID customizing dataset and training