YOLO-NAS from Deci.ai

Hello,

I’d like to use the YOLO NAS model on DeepStream 6.3. I noticed that NVIDIA acquired Deci.ai. They had a github super-gradients and it seems to be deprecated now.

Do you support this model?

Is it still being maintained? Would it be worth investing time in making it work, or does NVIDIA plan to add it to TAO?

Best regards.

1 Like

There is no plan to add it to TAO at present, but the repository you provided already provides a model in pt format. You only need to convert the model format to onnx and add the corresponding custom post-processing to make DeepStream work.

Here are some samples provided by community users. You can refer to this