Support for video segmentation demo without referring image

UX-Decoder / Segment-Everything-Everywhere-All-At-Once

[NeurIPS 2023] Official implementation of the paper "Segment Everything Everywhere All at Once"

Apache License 2.0

4.29k stars 378 forks source link

Support for video segmentation demo without referring image #121

Open mukulkhanna opened 7 months ago

mukulkhanna commented 7 months ago

Hello team,

Is there any way I can set up a demo for evaluating SEEM's video segmentation capabilities – without a referring image?

The current demo throws an error when I upload a video and click "SUBMIT" to do automatic video segmentation. It also throws an error if I provide a text-prompt with the video.

Please let me know. Thanks!

mukulkhanna commented 7 months ago

To be more specific, I am interested in doing multi-class video segmentation with a custom vocabulary of semantic classes.