Customized Keypoint detection model for Orin Nano

[erence]

Please provide complete information as applicable to your setup.

**• Hardware Platform: Jetson Orin Nano **
**• DeepStream Version 6.3 **
• JetPack Version (5.12)
• TensorRT Version
• Issue Type( question)

We would like to create a custom keypoint recognition model to be used live with our basler monochrome camera.

Which keypoint recognition model would Nvidia suggest us, yolo, mobilenet or another pose model?

Can we convert our 1920-1200-1 grayscale yolov8-pose model to deepstream compatible engine file? If yes what is the easiest way to convert it?

Could we reach >10 fps with Orin Nano Deepstream int8 engine model?

Which annotation tool can you suggest for us?

Best Regards
Eren

[Fiona.Chen]

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

@erence

Which keypoint recognition model would Nvidia suggest us, yolo, mobilenet or another pose model?

It depends on your resources and requirements. There is no limitation from DeepStream point of view.

@erence

Can we convert our 1920-1200-1 grayscale yolov8-pose model to deepstream compatible engine file? If yes what is the easiest way to convert it?

Yes. We will suggest you to use ONNX model.

@erence

Could we reach >10 fps with Orin Nano Deepstream int8 engine model?

It depends on your model and scenario.

@erence

Which annotation tool can you suggest for us?

It depends on your resources and requirements. There is no limitation from Nvidia.