Deploying tao developed Yolov4 model in jetson device

[Riddhi]

I am going to run Tao Toolkit generated Yolov4 model on Jetson Xavier Nx. There are some Deepstream_tao_apps examples here: deepstream_tao_apps/pgie_yolov4_tao_config.txt at master · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub

However, I am after an example where there is Pgie, and 3 sgie models run together like this: deepstream_python_apps/apps/deepstream-test2 at master · NVIDIA-AI-IOT/deepstream_python_apps · GitHub Is there any such example for the Deepstream_Tao?

Also, all the examples in deepstream_python_apps include caffe model. Any option to run a Tao-generated model in the same manner i.e.: to fully integrate the model in deepstream just like the caffe models in those exmaples?

[Morganh]

@Riddhi

Is there any such example for the Deepstream_Tao?

Yes, there are some examples in DS.
For example,
/opt/nvidia/deepstream/deepstream-6.0/samples/configs/tao_pretrained_models/deepstream_app_source1_dashcamnet_vehiclemakenet_vehicletypenet.txt

See its README for more details.
Above example will run object detection network(dashcamnet) and then two classification networks.

[Riddhi]

Hi Morganh,

Thank you for the above, it worked perfectly for me. However, are there any python code examples available for it?

[Morganh]

You can have a look at GitHub - NVIDIA-AI-IOT/tao-toolkit-triton-apps: Sample app code for deploying TAO Toolkit trained models to Triton.

[Riddhi]

Thanks, it was helpful.