Video Object Segmentation

[swap-10]

Is your feature request related to a problem? Please describe. With video data and applications on the rise, a video object segmentation interface would be really helpful.

Describe the solution you'd like Polygonal annotation for video frames. Something like VideoPolygon in addition to the currently available VideoRectangle. VideoRectangle enables VOT (Video Object Tracking) VideoPolygon will enable the other major task, VOS (Video Object Segmentation).

Describe alternatives you've considered Tried using custom templates with Video Object tag and PolygonLabels control tag, but it doesn't work. Interface shoes error that PolygonLabels can only control image, not video.

Additional context

[heidi-humansignal]

Hi there!

Thank you for your feature request! We greatly appreciate your feedback and the opportunity to consider your suggestion. Your request will be evaluated and ranked alongside other roadmap items. We understand that video object segmentation (VOS) is an important task for many applications, and having the ability to create polygonal annotations on video frames would be highly beneficial. This feature has been requested by several users in the community, and we are actively considering it for future development. While we cannot provide a specific timeline, please know that our development team is aware of this need.

Thanks,

Tyler Conlee Head of Support HumanSignal

Comment by Tyler Conlee Workflow Run

[michaelmyc]

Would +1 for this feature, especially with integration of SAM2.

[michaelmyc]

I'd also add that mask and polygons are both important for video object segmentation. Maybe mask being even more important/relevant considering integration with SAM2 would hugely accelerate the annotation process.