This patch allows exporting YOLOv10 to ONNX file compatible with OpenCV

yolov10.patch

diff --git a/export_opencv.py b/export_opencv.py
new file mode 100644
index 00000000..15bfef90
--- /dev/null
+++ b/export_opencv.py
@@ -0,0 +1,23 @@
+from ultralytics import YOLOv10
+import argparse
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--weights", type=str,
+                        default="yolov10n.pt",
+                        help="model.pt path")
+    parser.add_argument("--imgsz", type=int, nargs=2,
+                        default=(640, 640),
+                        help="Image size for the model")
+    parser.add_argument("--half",
+                        action="store_true",
+                        help="FP16 half-precision export")
+    args = parser.parse_args()
+
+    model = YOLOv10(args.weights)
+
+    model.export(format='onnx',
+                 imgsz=args.imgsz,
+                 simplify=True,
+                 half=args.half)
diff --git a/ultralytics/nn/modules/head.py b/ultralytics/nn/modules/head.py
index a9c5d9ee..544ab8b7 100644
--- a/ultralytics/nn/modules/head.py
+++ b/ultralytics/nn/modules/head.py
@@ -79,7 +79,7 @@ class Detect(nn.Module):
     def forward(self, x):
         """Concatenates and returns predicted bounding boxes and class probabilities."""
         y = self.forward_feat(x, self.cv2, self.cv3)
-        
+
         if self.training:
             return y
 
@@ -507,7 +507,7 @@ class v10Detect(Detect):
 
         self.one2one_cv2 = copy.deepcopy(self.cv2)
         self.one2one_cv3 = copy.deepcopy(self.cv3)
-    
+
     def forward(self, x):
         one2one = self.forward_feat([xi.detach() for xi in x], self.one2one_cv2, self.one2one_cv3)
         if not self.export:
@@ -519,8 +519,7 @@ class v10Detect(Detect):
                 return {"one2many": one2many, "one2one": one2one}
             else:
                 assert(self.max_det != -1)
-                boxes, scores, labels = ops.v10postprocess(one2one.permute(0, 2, 1), self.max_det, self.nc)
-                return torch.cat([boxes, scores.unsqueeze(-1), labels.unsqueeze(-1).to(boxes.dtype)], dim=-1)
+                return one2one
         else:
             return {"one2many": one2many, "one2one": one2one}