• Hardware Platform (Jetson / GPU) - GPU
• DeepStream Version - 6.0
• JetPack Version (valid for Jetson only) - 4.6
• TensorRT Version - 8.0.1
• NVIDIA GPU Driver Version (valid for GPU only) - 470.129.06
• Issue Type( questions, new requirements, bugs) - questions
• Question - I’m trying to implement temporal 3d human pose estimation pipeline, which consists of 2d pose estimation → postprocessing → 3d pose estimation.
The “postprocessing” part consists of accumulating sequence (multiple frames’ data) of 2d pose estimation keypoints and passing as batch input to the 3d pose estimation model. Given the description of nvinfer plugin, it seems that I can use its Secondary mode or Preprocessed tensor input mode for this purpose, but there are no examples for the same. (I referred to deepstream-3d-action-recognition example pipeline, but ROI inputs are being passed to sgie there)
Moreover, for the preprocessed tensor input mode, it says that only full frame or ROI inputs are supported currently. Can you guide me if this mode can still be used with nvdspreprocess plugin?