Action recognition preprocessing

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

• Hardware: NVIDIA RTX A6000
• Network Type: ActionRecognitionNet
• Tao version: 5.3.0

Hello,
I trained action recognition net using the notebook (tao_tutorials/notebooks/tao_launcher_starter_kit/action_recognition_net/actionrecognitionnet.ipynb at main · NVIDIA/tao_tutorials · GitHub) and now I want to deploy it to my edge machine as onnx. The problem is that even though the model says it uses rgb the input shape of the model is 9x224x224. I guess that is because it uses some kind of greyscaling. But I am unable to find any specific information about what preprocessing is used. I tried to look in open source TAO at github (tao_tensorflow1_backend/nvidia_tao_tf1/cv at main · NVIDIA/tao_tensorflow1_backend · GitHub) because I found there a solution for similar question for different model, but I do not see the action recognition model there.
Could you please help me where to find the scripts for action recognition preprocessing?

Thank you in advance
Adam Karásek

3

You can find the source code in NVIDIA Corporation · GitHub → GitHub - NVIDIA/tao_pytorch_backend: TAO Toolkit deep learning networks with PyTorch backend.

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.