Fix: Prevent in-place mutation of Detections input in ByteTrack

[Aaryan2304]

Summary

This PR fixes a side-effect in ByteTrackTracker.update() where the input sv.Detections object was modified in-place, overwriting existing tracker_id values with -1. This behavior caused data loss for pipelines that reuse the detections object after the tracking step.

Changes

• Removed redundant in-place initialization: detections.tracker_id = -np.ones(len(detections)) (line 111)
  • This was unnecessary because all outputs are built via deepcopy() calls in _update_detections(), the unmatched loop, and _spawn_new_trackers().
• Updated the empty tracks/detections edge case (line 103-104) to return a deepcopy() instead of mutating and returning the input directly.

Verification

• All ruff checks pass (linting and formatting)
• Type checking (mypy) passes with no errors
• Verified locally that the input detections object retains its original values after calling update()
• Confirmed that returned tracking results are correct and contain expected IDs

CLA Signing

I have read the CLA Document and I sign the CLA.

[CLAassistant]

All committers have signed the CLA.

[Copilot]

Pull request overview

This PR fixes a side-effect bug in ByteTrackTracker.update() where the input sv.Detections object was being modified in-place, causing data loss for downstream code that reuses the detections object. The fix ensures the input remains immutable by using deepcopy() for all return paths.

Changes:

• Removed redundant in-place mutation of input detections (detections.tracker_id = -np.ones(len(detections)))
• Fixed edge case handling to return a deep copy instead of mutating the input when both tracks and detections are empty

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[SkalskiP]

Hi @Aaryan2304 👋🏻 Thanks a lot for your interes in trackers and for submitting this PR. We are planning trackers-2.2.0 release for this week and I'll make sure we will include this fix.

[Aaryan2304]

Hi @Aaryan2304 👋🏻 Thanks a lot for your interes in trackers and for submitting this PR. We are planning trackers-2.2.0 release for this week and I'll make sure we will include this fix.

Thanks @SkalskiP ! Appreciate you including this in the next release.
Let me know if you need any changes to the PR before then.

[SkalskiP]

Hi @Aaryan2304 👋🏻

It looks like the same bug is also partially present in SORT.

trackers/core/sort/tracker.py lines 133-135

if len(self.trackers) == 0 and len(detections) == 0:
    detections.tracker_id = np.array([], dtype=int)
    return detections

trackers/core/bytetrack/tracker.py lines 112-114

if len(self.tracks) == 0 and len(detections) == 0:
    detections.tracker_id = np.array([], dtype=int)
    return detections

[AlexBodner]

@SkalskiP ran soccernet benchmark over this branch and gives same result + from my overview seems like an adequate change

[SkalskiP]

@Aaryan2304 I'm merging the PR. Thanks a lot for your contribution! We may need similar PR for SORT.

[Aaryan2304]

@Aaryan2304 I'm merging the PR. Thanks a lot for your contribution! We may need similar PR for SORT.

Ok, I'll make the necessary changes and raise a PR accordingly.