search

Zhongdao / UniTrack

[NeurIPS'21] Unified tracking framework with a single appearance model. It supports Single Object Tracking (SOT), Video Object Segmentation (VOS), Multi-Object Tracking (MOT), Multi-Object Tracking and Segmentation (MOTS), Pose Tracking, Video Instance Segmentation (VIS), and class-agnostic MOT (e.g. TAO dataset).
MIT License
338 stars 34 forks source link

SOT on custom videos #15

Open ooza opened 3 years ago

ooza commented 3 years ago

Hi Zhongdao, Thanks a lot for this great work! I'm wondering how can we apply one of your SOT models on a custom video. I couldn't find any details in your readme files on how to do that and if I understood correctly the test_sot_xx.py scripts are dedicated to the OTB2015 dataset.

Zhongdao commented 3 years ago

Hi @ooza,

You are right the current repo does not contain such a script. I planed to write one before but didn't manage to. Please stay tuned and wait for me to finish the SOT demo!

Zhongdao commented 2 years ago

Hi @ooza, I just wrote an SOT demo, see here. This demo is based on Cross Correlation (SiamFC). DCF would be better, but I encountered some issues on inconsistency between pytorch versions for complex number calculation. I will leave this issue open to see if anyone has suggestions.