how to generate mask but video?

open-mmlab / mmtracking

OpenMMLab Video Perception Toolbox. It supports Video Object Detection (VID), Multiple Object Tracking (MOT), Single Object Tracking (SOT), Video Instance Segmentation (VIS) with a unified framework.

https://mmtracking.readthedocs.io/en/latest/

Apache License 2.0

3.54k stars 593 forks source link

how to generate mask but video? #593

Open jackleibest opened 2 years ago

jackleibest commented 2 years ago

seems mmtracking only generate videos, so how to generate the mask png images? like https://github.com/foolwood/SiamMask.

JingweiZhang12 commented 2 years ago

We support VIS tasks that can predict masks. Now, the SOT methods we support can not predict masks.

kartikgupta2607 commented 2 years ago

Can you please provide details on how to infer a video for the VIS task and output the instance_id segmented masks?