Support for video segmentation demo without referring image

[mukulkhanna]

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]

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