Problem deploying trtpose on deepstream

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

• Hardware Platform (Jetson / GPU)
Orin NX 16Gb - Seeed A608 carrier

• DeepStream Version
6.2

• JetPack Version (valid for Jetson only)
JP5.1.1

• TensorRT Version
8.5.2-1+cuda11.4

• Issue Type( questions, new requirements, bugs)
Can not convert correctly pth to onnx model, the obtained model is not usable in deep stream

• 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)
I am trying to integrate trt_pose to a deepstream pipeline so I am following the instructions from this tutorial: https://developer.nvidia.com/blog/creating-a-human-pose-estimation-application-with-deepstream-sdk/?ncid=so-link-52952-vt24&sfdcid=EM08#cid=em08_so-link_en-us
and this repository: GitHub - NVIDIA-AI-IOT/deepstream_pose_estimation: This is a DeepStream application to demonstrate a human pose estimation pipeline.

When I try to convert the pth file to onnx format using the script export_for_isaac.py as explained here: https://developer.nvidia.com/blog/creating-a-human-pose-estimation-application-with-deepstream-sdk/?ncid=so-link-52952-vt24&sfdcid=EM08#cid=em08_so-link_en-us
I obtain an onnx model then I include this in my deepstream pipeline and it does not work as expected (the joints are randomly drawn).

In the other hand if I use the onnx file generated by the author of the tutorial which is here: deepstream_pose_estimation/pose_estimation.onnx at master · NVIDIA-AI-IOT/deepstream_pose_estimation · GitHub it runs correctly in my pipeline (body pose seems correct)

In both cases the deepstream pipeline is the same the only difference is that in the first case I use the onnx file generated by export_for_isaac script and in the second case I take the onnx stored in the github repository

As the repository is old I am wondering if this problem is related to my tensorrt or torch version

Could you please help me with this?
Are you able to reproduce this issue please?

Regards.

PS: This is a copy from here and I am creating this new topic as recommended.

3

The sample is not maintained any more. NVIDIA-AI-IOT/deepstream_pose_estimation: This is a DeepStream application to demonstrate a human pose estimation pipeline. (github.com)

Please use this sample instead. deepstream_reference_apps/deepstream-bodypose-3d at master · NVIDIA-AI-IOT/deepstream_reference_apps (github.com)

OR this sample can be referred too. deepstream_tao_apps/apps/tao_others/deepstream-bodypose2d-app at master · NVIDIA-AI-IOT/deepstream_tao_apps (github.com), you can follow the instructions here to run the sample deepstream_tao_apps/apps/tao_others at master · NVIDIA-AI-IOT/deepstream_tao_apps (github.com)

4

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

5

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.