Merge pull request #1165 from fengyuentau:gold_yolo

[BugFix] dnn (ONNX): Foce dropping constant inputs in parseClip if they are shared #25319

Resolves #25278
Merge with opencv/opencv_extra#1165

In Gold-YOLO ,`Div` has a constant input `B=6` which is then parsed into a `Const` layer in the ONNX importer, but `Clip` also has the shared constant input `max=6` which is already a `Const` layer and then connected to `Elementwise` layer. This should not happen because in the `forward()` of `Elementwise` layer, the legacy code goes through and apply activation to each input. More details on #25278 (comment).

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake

Author: fengyuentau

modules/dnn/src/onnx/onnx_importer.cpp

@@ -91,7 +91,8 @@ class ONNXImporter
 
     void addConstant(const std::string& name, const Mat& blob);
     void addLayer(LayerParams& layerParams,
-                  const opencv_onnx::NodeProto& node_proto);
+                  const opencv_onnx::NodeProto& node_proto,
+                  int num_inputs = std::numeric_limits<int>::max());
     void setParamsDtype(LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto);
 
     void lstm_extractConsts(LayerParams& layerParams, const opencv_onnx::NodeProto& lstm_proto, size_t idx, int* blobShape_, int size);
@@ -617,7 +618,8 @@ ONNXImporter::TensorInfo ONNXImporter::getBlobExtraInfo(const std::string& input
 }
 
 void ONNXImporter::addLayer(LayerParams& layerParams,
-                            const opencv_onnx::NodeProto& node_proto)
+                            const opencv_onnx::NodeProto& node_proto,
+                            int num_inputs)
 {
     int depth = layerParams.get<int>("depth", CV_32F);
     int id = dstNet.addLayer(layerParams.name, layerParams.type, depth, layerParams);
@@ -632,7 +634,8 @@ void ONNXImporter::addLayer(LayerParams& layerParams,
 
     std::vector<MatShape> layerInpShapes, layerOutShapes, layerInternalShapes;
     int inpNum = 0;
-    for (int j = 0; j < node_proto.input_size(); j++)
+    num_inputs = std::min(node_proto.input_size(), num_inputs);
+    for (int j = 0; j < num_inputs; j++)
     {
         const std::string& input_name = node_proto.input(j);
         IterLayerId_t layerId = layer_id.find(input_name);
@@ -1799,7 +1802,7 @@ void ONNXImporter::parseClip(LayerParams& layerParams, const opencv_onnx::NodePr
 
     layerParams.set("min_value", layerParams.get<float>("min", min_value));
     layerParams.set("max_value", layerParams.get<float>("max", max_value));
-    addLayer(layerParams, node_proto);
+    addLayer(layerParams, node_proto, 1);
 }
 
 void ONNXImporter::parseLeakyRelu(LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto)

modules/dnn/test/test_onnx_importer.cpp

@@ -3096,6 +3096,10 @@ TEST_P(Test_ONNX_layers, MatMulAddFusion) {
     testONNXModels("biased_matmul", npy, l1, lInf);
 }
 
+TEST_P(Test_ONNX_layers, ClipDivSharedConstant) {
+    testONNXModels("clip_div_shared_constant");
+}
+
 INSTANTIATE_TEST_CASE_P(/**/, Test_ONNX_nets, dnnBackendsAndTargets());
 
 }} // namespace