Is it included in the engine when running nms with cluster-mode=2?

[yeongjae1]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 6.1.1
• 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)

Hello, I put yolo.onnx in deepstream config and created an engine file using the cudaEngine function. I can think of onnx as {0:inputs, 1:outputs} format that does not include nms.

My question is, when cluster-mode=2, is it connecting by inserting a plugin such as batchedNMSDynamicPlugin, or is it returning to C++ from the host memory every time the model runs?

When I looked up the code, it seems to work in the latter way, but I’m curious

[yeongjae1]

NvDsInferParseYolo -> clusterAndFillDetectionOutputNMS

It seems that the function moves in the above order, is that correct?

[Fiona.Chen]

If you are talking about our sample /opt/nvidia/deepstream/deepstream/sources/objectDetector_Yolo, the answer is yes.

[yeongjae1]

Then, in the case of nms, you mean that it is processed in real time at the cpu (host) level, not included in the engine??

[Fiona.Chen]

Yes. NMS is working on CPU for the sample /opt/nvidia/deepstream/deepstream/sources/objectDetector_Yolo. The tensor parsing can be GPU accelerated. Please refer to /opt/nvidia/deepstream/deepstream/sources/objectDetector_Yolo/nvdsinfer_custom_impl_Yolo/NvDsInferParseCustomYolov3_cuda.cu