Problem with tensor meta of sgie

Accelerated Computing Intelligent Video Analytics DeepStream SDK
1

• Hardware Platform GPU (Quadro P4000) •
• Container DeepStream 5.1 triton •

I’m building a face recognition system using DeepStream with multiple input and output multiple RTSP, everything work just fine from the detection. In the recognition part, I want to add tensor data from recognition model to the pipeline for custom post processing.

  • I enable PROCESS_MODE_CLIP_OBJECTS, operate_on_gie_id = 1, operate_on_class_ids = 0 in recognition model config. I do not use tracker in my pipeline.
  • I try to access obj_user_meta_list in object_meta from sgie_probe (I try to add probe to both sink_pad and src_pad), however, I alway get None value for obj_user_meta_list.

I also read other related post for debug my case, but no solution founded. Please help me find a solution, thanks!

3

Could you check deepstream_tao_apps/apps/tlt_others/deepstream-faciallandmark-app at release/tlt3.0 · NVIDIA-AI-IOT/deepstream_tao_apps · GitHub to see if any help?

4

Thanks for your reply! I solve my problem by comment operate_on_gie_id = 1. However I do not really understand why I need to do that because my pgie id is 1.