Open
Description
Description
Currently, the batchers will aggregate detections on chunks, where each chunk will await all detections that are associated with it based on tracking of detectors associated with that chunk. However, there may be an edge case present if there were two chunks, if detections return for the second chunk but not for the first, that the tracking of the first "chunk" may be lost, since the batcher works on existent chunks with detections.
Discussion
The scenario hasn't presented yet but is based on knowledge of the flow. This may not be high priority to address until the scenario is reproduced concretely, but I thought it'd be better to note down since this has been noted in several discussions separately.
Metadata
Metadata
Assignees
Labels
No labels