How can I use datumaro to merge the annotations.xml & video to get a yolo dataset that only has labelded images ?

[wenerme]

I'm new to ml world, I just found these tow

datum project import --format cvat -n cvat1 annotations.xml
datum project import --format video_frames -n vid1 video.mp4

I found #1251 this

datum project export -e '/item/annotation' --filter-mode 'i+a' -f --save-images < your_target_format > --

But I don't known how can I achivev this

• https://openvinotoolkit.github.io/datumaro/latest/docs/data-formats/formats/yolo_ultralytics.html

Why

• cvat is running in docker, slow than host (macOS)
• cvat export frame is slow, every export with image need to prepare all frames, even a little label change
• I hope I can use the annotations.xml+video/frames to get a faster export
• I hope I can use jpg instead of png - MP4 video, jpg is better, png is very large.