Player tracking through occlusion in the application of sports

Beginners
1

Was wondering how it would be possible to consistently track and keep the same id of players(specifically 2 in this scenario) through heavy occlusion. I have tried integrating deepsort, but I am unable to create a decent algorithm that can track players through occlusion.
Specifically, I am getting problems in id switching, in which one player takes another’s while they cross paths.
Has anyone solved this problem?

If anyone has more info about how to solve this problem, I’d gladly send over my code and some examples. Really am stumped on this!

3 Likes
4

If you’re looking for an entire library to perform this task, suggest you check out GitHub - tryolabs/norfair: Lightweight Python library for adding real-time multi-object tracking to any detector.. It pairs nicely with object detection, so you can optimize “re-identification” with it: norfair/demos/reid at master · tryolabs/norfair · GitHub.

There is also this interesting library https://cotracker3.github.io/ which looks like was built for tracking through occlusions.

1 Like