Please provide complete information as applicable to your setup.
• Hardware Platform (Jetson / GPU)
• DeepStream Version
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only)
• Issue Type( questions, new requirements, bugs)
• How to reproduce the issue ? (This is for bugs. Including which sample app is using, the configuration files content, the command line used and other details for reproducing)
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description)
I am trying to using a onnx model instead of the raw config files of yolov4 tiny.
But when TRT tries to build the engine file using the onnx, it gives different output names, due to which the inference fails.
1 Engine file build using the onnx
0:00:00.141057408 16914 0x5c65c8b7ff40 WARN nvinfer gstnvinfer.cpp:679:gst_nvinfer_logger:<person_version1-person_model> NvDsInferContext[UID 7]: Warning from NvDsInferContextImpl::initialize() <nvdsinfer_context_impl.cpp:1244> [UID = 7]: Warning, OpenCV has been deprecated. Using NMS for clustering instead of cv::groupRectangles with topK = 20 and NMS Threshold = 0.5
0:00:04.570073425 16914 0x5c65c8b7ff40 INFO nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<person_version1-person_model> NvDsInferContext[UID 7]: Info from NvDsInferContextImpl::deserializeEngineAndBackend() <nvdsinfer_context_impl.cpp:2095> [UID = 7]: deserialized trt engine from :/opt/nvidia/deepstream/deepstream-7.0/nvodin24/model-personn.engine
WARNING: [TRT]: The getMaxBatchSize() function should not be used with an engine built from a network created with NetworkDefinitionCreationFlag::kEXPLICIT_BATCH flag. This function will always return 1.
INFO: [Implicit Engine Info]: layers num: 3
0 INPUT kFLOAT input 3x512x512
1 OUTPUT kFLOAT boxes 3840x1x4
2 OUTPUT kFLOAT confs 3840x1
2 Engine file build using the yolov4 cfg
0:00:00.133819034 17776 0x615f316521c0 WARN nvinfer gstnvinfer.cpp:679:gst_nvinfer_logger:<person_version1-person_model> NvDsInferContext[UID 70]: Warning from NvDsInferContextImpl::initialize() <nvdsinfer_context_impl.cpp:1244> [UID = 70]: Warning, OpenCV has been deprecated. Using NMS for clustering instead of cv::groupRectangles with topK = 20 and NMS Threshold = 0.5
Deserialize yoloLayer plugin: yolo
0:00:04.497218264 17776 0x615f316521c0 INFO nvinfer gstnvinfer.cpp:682:gst_nvinfer_logger:<person_version1-person_model> NvDsInferContext[UID 70]: Info from NvDsInferContextImpl::deserializeEngineAndBackend() <nvdsinfer_context_impl.cpp:2095> [UID = 70]: deserialized trt engine from :/opt/nvidia/deepstream/deepstream-7.0/engine/person_version1-person_model.engine
WARNING: [TRT]: The getMaxBatchSize() function should not be used with an engine built from a network created with NetworkDefinitionCreationFlag::kEXPLICIT_BATCH flag. This function will always return 1.
INFO: [Implicit Engine Info]: layers num: 4
0 INPUT kFLOAT input 3x512x512
1 OUTPUT kFLOAT boxes 3840x4
2 OUTPUT kFLOAT scores 3840x1
3 OUTPUT kFLOAT classes 3840x1
The layers are different, and what is missing while onn → engine conversion is Deserialize yoloLayer plugin: yolo
which is resulting in different outpu names.
How do I solve this?
Thank you