Deepstream-3d-action-recognition preprocessing on detected ROIs

Hi, my question is related to DeepStream 3D Action Recognition App — DeepStream 6.1.1 Release documentation
I want to use the preprocessing + action recognition as sgie, and add a person detector as the pgie. However, the preprocessing is done on the input frame and I can’t find any resources on how to modify it to accept ROIs from the detection results instead of fixed user-defined ROIs.
I’d appreciate any help, suggestions, orrelated resources.
Thank you.

• Hardware Platform (Jetson / GPU) NVIDIA GeForce RTX 3070
• DeepStream Version 6.1
• TensorRT Version 8.4.3
• NVIDIA GPU Driver Version (valid for GPU only) 11.3
• Issue Type( questions, new requirements, bugs) questions

Currently the nvdspreprocess for SGIE is under development. Please wait for future release.

Thank you for your response. If I want to modify the codes to use the 3d action recognition engine as sgie, where should I start?
Is it just the preprocessing that I need to modify?

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.

Both nvdspreprocessing and nvinfer should be modified. They are open source.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.