search

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
3.44k stars 580 forks source link

Question about multitask model #602

Open hcv1027 opened 2 years ago

hcv1027 commented 2 years ago

Is it possible to customize a multitask model like JDE and FairMOT by using mmtracking? Can you provide an example tutorial of how to customize multitask model? Is track_head designed for this purpose?

JingweiZhang12 commented 2 years ago

Sounds good! It's possible to customize a multitask model. You can design it by calling our existing track_head APIs. Specifically, maybe you need to unify the interface of initialization and forward of multiple heads.