Nvinferserver inference with onnx returns wrong predictions

[ypge]

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version 7.0
• JetPack Version (valid for Jetson only)
• TensorRT Version
• NVIDIA GPU Driver Version (valid for GPU only) 535.161.08
• Issue Type( questions, new requirements, bugs) Questions/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)

Hi, I am trying to implement DS with Triton inference server on Yolo V5 in the multiarch docker mentioned in DS 7.0 Quick Start page, but I have encountered a problem that the inference result directly obtained from onnx in triton (throught gRPC protocal) is incorrect while through onnx runtime the model is working well, check the wrong_output.jpg and correct_output.jpg for example.
for reproducing, here are the files I am using for testing:

1. Python binding: deepstream_test1_rtsp_in_rtsp_out.py
2. official Github repository for postprocessing: yolo deepstream modified to adapt to yolov5 output format (the original YoloV4 decode/parse function) nvdsparsebbox_Yolo.cpp
3. Yolo V5s onnx file exported from official pt file (put in my github repository): model.zip (trained with coco128) , labels in labels.txt in github (run in triton without config.pbtxt)
4. deepstream config file dstest1_yolov5s.txt in github
5. command line used is python3 deepstream_test1_rtsp_in_rtsp_out.py -i {source} -g nvinferserver

I guess the problem lies in preprocess part, where the input data is not well formatted for onnx model (NCWH), but I don’t have any idea on why this is happening cuz the input is supposed to be automatically processes with configuration TENSOR_ORDER_LINEAR?

Hoping to get some suggestions or help, thanks.

[fanzh]

please refer to this topic for yolov5 triton sample.

[ypge]

It turns out that there needs some more settings in either config.pbtxt or preprocessing config in DS. The repository DeepStream-Yolo seems not to show such configurations and will lead to similar issue in my case.
Happily the problem is solved and I appreciate your help.