Implement custom yolo or any custom model in Deepstream using Python

[navyashivads]

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) -NA
• TensorRT Version -
• NVIDIA GPU Driver Version (valid for GPU only) -
• Issue Type( questions, new requirements, bugs) - questions

** Issue** - Our team is trying to build the Deepstream application with Yolo model. We implemented the GitHub - marcoslucianops/DeepStream-Yolo: NVIDIA DeepStream SDK 6.1.1 / 6.1 / 6.0.1 / 6.0 configuration for YOLO models and built YOLOv5 engine. Could anyone please let me know if I can just use the yolo.cfg and yolo.wts without including the libnvdsinfer_custom_impl_Yolo.so file (custom-lib-path=/opt/nvidia/deepstream/deepstream-6.1/sources/yolo/nvdsinfer_custom_impl_Yolo/libnvdsinfer_custom_impl_Yolo.so).

If not, what is the use of ibnvdsinfer_custom_impl_Yolo.so in this context?

I come from a Python background so, understanding CPP files or terms associated with it is becoming tough to maneuver through NVIDIA Deepstream.

Any resources/ answers are highly appreciated. Thank you so much :)

[fanzh]

please refer to How to add custom post process after infer in deepstream python app - #8 by 328541716

[navyashivads]

Thanks for your reply. I have modified the config file in a below manner:

I changed the Network-type = 100, and commented the custom-lib-path. However, when I tested it did not give me the output. Could you please correct me where I am doing it wrong?

I did not quite understand what you meant by, “you can access output data NvDsInferLayerInfo”. Can you please elaborate on this? Could you also please explain me what does post-processing in Python language mean?

Thank you :)

[fanzh]

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

please refer to deepstream sample deepstream-infer-tensor-meta-test, here is the key setting:

0=Detector, 1=Classifier, 2=Segmentation, 100=Other

network-type=100

Enable tensor metadata output

output-tensor-meta=1
then inference results NvDsInferLayerInfo can be got in probe functions, you need to convert the postprocess code the python.