`ultralytics 8.2.85` YOLOv10 `max_det` arg fix (#15917)

Co-authored-by: Laughing-q <1185102784@qq.com>
Co-authored-by: Laughing <61612323+Laughing-q@users.noreply.github.com>
Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>

ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics YOLO 🚀, AGPL-3.0 license
 
-__version__ = "8.2.84"
+__version__ = "8.2.85"
 
 import os
 

ultralytics/engine/exporter.py

@@ -249,6 +249,7 @@ class Exporter:
                 m.dynamic = self.args.dynamic
                 m.export = True
                 m.format = self.args.format
+                m.max_det = self.args.max_det
             elif isinstance(m, C2f) and not is_tf_format:
                 # EdgeTPU does not support FlexSplitV while split provides cleaner ONNX graph
                 m.forward = m.forward_split

ultralytics/nn/modules/head.py

@@ -144,12 +144,12 @@ class Detect(nn.Module):
             (torch.Tensor): Processed predictions with shape (batch_size, min(max_det, num_anchors), 6) and last
                 dimension format [x, y, w, h, max_class_prob, class_index].
         """
-        batch_size, anchors, predictions = preds.shape  # i.e. shape(16,8400,84)
+        batch_size, anchors, _ = preds.shape  # i.e. shape(16,8400,84)
         boxes, scores = preds.split([4, nc], dim=-1)
         index = scores.amax(dim=-1).topk(min(max_det, anchors))[1].unsqueeze(-1)
         boxes = boxes.gather(dim=1, index=index.repeat(1, 1, 4))
         scores = scores.gather(dim=1, index=index.repeat(1, 1, nc))
-        scores, index = scores.flatten(1).topk(max_det)
+        scores, index = scores.flatten(1).topk(min(max_det, anchors))
         i = torch.arange(batch_size)[..., None]  # batch indices
         return torch.cat([boxes[i, index // nc], scores[..., None], (index % nc)[..., None].float()], dim=-1)
 

ultralytics/utils/ops.py

@@ -218,7 +218,7 @@ def non_max_suppression(
         classes = torch.tensor(classes, device=prediction.device)
 
     if prediction.shape[-1] == 6:  # end-to-end model (BNC, i.e. 1,300,6)
-        output = [pred[pred[:, 4] > conf_thres] for pred in prediction]
+        output = [pred[pred[:, 4] > conf_thres][:max_det] for pred in prediction]
         if classes is not None:
             output = [pred[(pred[:, 5:6] == classes).any(1)] for pred in output]
         return output