OC_Sort Vs. ByteTracker Inference Speed

[Rm1n90]

Hello @noahcao,

Thanks for sharing the code. I have a question regard to speed of OC_Sort.
Im using YoloV5 as a object detector for both OC_Sort and ByteTracker tracking algorithm with the same setting. However, all the process for one frame (Detection and tracking) took almost ~14ms for ByteTracker and ~45ms for OC_Sort. I would like to ask you how can I speed up the OC_Sort (Based on your paper, OC_Sort should be faster than ByteTrack)?

Thanks!

[noahcao]

Hi @Rm1n90,

Yes, in my raw implementation, OC-SORT should be slightly faster than ByteTrack. In this repo, the released version is based on a naive and less efficienct implementation. The key should be in the FOR-loops in oc-sort. I am not updating this repo in the coming days as my focus is to support a higher-efficient version in mmtracking (see PR in mmtracking), for which we would make it faster.

Apologize for the confusion and inconvenience I make here.

[mohaghighat]

Does this repo has the efficient code yet, or should we only use the version in the mmtracking?

[noahcao]

@mohaghighat

For the time efficiency, they are almost the same. This version is cleaner if you only want to use OC-SORT. The mmtracking version is more flexible if you'd like to try more features, such as adaptive detector, backbones, more engineering tricks etc.