Update

Author: panagiotamoraiti

supervision/metrics/detection.py

@@ -290,18 +290,26 @@ def evaluate_detection_batch(
 
         class_id_idx = 4
         true_classes = np.array(targets[:, class_id_idx], dtype=np.int16)
-        detection_classes = np.array(detection_batch_filtered[:, class_id_idx], dtype=np.int16)
+        detection_classes = np.array(
+            detection_batch_filtered[:, class_id_idx], dtype=np.int16
+        )
         true_boxes = targets[:, :class_id_idx]
         detection_boxes = detection_batch_filtered[:, :class_id_idx]
 
-        iou_batch = box_iou_batch(boxes_true=true_boxes, boxes_detection=detection_boxes)
+        iou_batch = box_iou_batch(
+            boxes_true=true_boxes, boxes_detection=detection_boxes
+        )
+        # matched_idx = np.asarray(iou_batch > iou_threshold).nonzero()
 
-        matched_gt_idx = set()
-        matched_det_idx = set()
+        # if matched_idx[0].shape[0]:
+        #     matches = np.stack(
+        #         (matched_idx[0], matched_idx[1], iou_batch[matched_idx]), axis=1
+        #     )
+        #     matches = ConfusionMatrix._drop_extra_matches(matches=matches)
+        # else:
+        #     matches = np.zeros((0, 3))
 
-        # For each GT, find best matching detection (highest IoU > threshold)
-        for gt_idx, gt_class in enumerate(true_classes):
-            candidate_det_idxs = np.where(iou_batch[gt_idx] > iou_threshold)[0]
+        matched_idx = np.asarray(iou_batch > iou_threshold).nonzero()
 
         if matched_idx[0].shape[0]:
             # Filter matches by class equality
@@ -312,18 +320,15 @@ def evaluate_detection_batch(
                 valid_true_idx = matched_idx[0][valid_matches_mask]
                 valid_pred_idx = matched_idx[1][valid_matches_mask]
 
-            best_det_idx = candidate_det_idxs[np.argmax(iou_batch[gt_idx, candidate_det_idxs])]
-            det_class = detection_classes[best_det_idx]
+                ious = iou_batch[valid_true_idx, valid_pred_idx]
+                matches = np.stack((valid_true_idx, valid_pred_idx, ious), axis=1)
 
-            if best_det_idx not in matched_det_idx:
-                # Count as matched regardless of class:
-                # same class → TP, different class → misclassification
-                result_matrix[gt_class, det_class] += 1
-                matched_gt_idx.add(gt_idx)
-                matched_det_idx.add(best_det_idx)
+                # Now drop extra matches with highest IoU per GT/pred
+                matches = ConfusionMatrix._drop_extra_matches(matches=matches)
             else:
-                # Detection already matched, GT is FN
-                result_matrix[gt_class, num_classes] += 1
+                matches = np.zeros((0, 3))
+        else:
+            matches = np.zeros((0, 3))
 
         matched_true_idx, matched_detection_idx, _ = matches.transpose().astype(
             np.int16
@@ -344,7 +349,6 @@ def evaluate_detection_batch(
 
         return result_matrix
 
-
     @staticmethod
     def _drop_extra_matches(matches: np.ndarray) -> np.ndarray:
         """