Search before asking
Description
#1562 discusses the OBB support for Metrics in detail, and #1593 added OBB support for mAP, Precision, Recall, and F1 Score. I see that ConfusionMatrix still lacks this support.
Why is it important?
Currently, anyone using ConfusionMatrix with OBB may assume that IoU is computed for the OBBs, but actually, IoU is being computed for the derived axis-aligned bounding boxes.
|
iou_batch = box_iou_batch( |
Use case
It will help anyone working with OBBs to compute ConfusionMatrix correctly.
Additional
No response
Are you willing to submit a PR?
Search before asking
Description
#1562 discusses the OBB support for Metrics in detail, and #1593 added OBB support for mAP, Precision, Recall, and F1 Score. I see that
ConfusionMatrixstill lacks this support.Why is it important?
Currently, anyone using
ConfusionMatrixwith OBB may assume that IoU is computed for the OBBs, but actually, IoU is being computed for the derived axis-aligned bounding boxes.supervision/supervision/metrics/detection.py
Line 299 in 77ae682
Use case
It will help anyone working with OBBs to compute ConfusionMatrix correctly.
Additional
No response
Are you willing to submit a PR?