🐛 [Fix] wrong filter_duplicates argument

The filter_duplicates() function should receive the topk_targets instead of the iou_mat as its first argument.

yolo/utils/bounding_box_utils.py

@@ -278,7 +278,7 @@ def __call__(self, target: Tensor, predict: Tuple[Tensor]) -> Tuple[Tensor, Tens
         topk_targets, topk_mask = self.filter_topk(target_matrix, topk=self.topk)
 
         # delete one anchor pred assign to mutliple gts
-        unique_indices, valid_mask, topk_mask = self.filter_duplicates(iou_mat, topk_mask)
+        unique_indices, valid_mask, topk_mask = self.filter_duplicates(topk_targets, topk_mask)
 
         align_bbox = torch.gather(target_bbox, 1, unique_indices.repeat(1, 1, 4))
         align_cls = torch.gather(target_cls, 1, unique_indices).squeeze(-1)