IoU is not a great fit for small/fast moving objects

[colonelpanic8]

Search before asking

• I have searched the Yolov8 Tracking issues and found no similar enhancement requests.

Description

IoU is not a great fit for small/fast moving objects because there can easily be no overlap in cases where detections are actually quite close.

There are other metrics where the tolerance for discrepancy between the distance between detections does not depend on the size of the detected object like centroid distance.

Ideally, the Tracker class should abstract over the distance metric that is used. IoU could be the default, but in some cases centroid distance + perhaps a separate box size comparison might be more appropriate.

IOU is hard coded e.g. here:
https://github.com/mikel-brostrom/yolo_tracking/blob/05172962914805dab99acb3d2d8592797dbf38a7/boxmot/trackers/strongsort/sort/tracker.py#L147

Use case

No response

Are you willing to submit a PR?

• Yes I'd like to help by submitting a PR!

[mikel-brostrom]

I agree that the addition of centroid distance would make sense for fast moving objects. I have however not time to implement or experiment with this at the moment, but will leave this open 😄.

[github-actions]

👋 Hello, this issue has been automatically marked as stale because it has not had recent activity. Please note it will be closed if no further activity occurs.
Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

[mikel-brostrom]

#1246

[mikel-brostrom]

Centroid-based association added here: #1247