We would like to understand the possible usage of HW accelerators (supported by Jetson Nano) instead of SW libraries in the typical application of receiving the RTSP stream to face recognition and uploading into the cloud.
Currently utilizing the opencv, face recognition and dlib libraries.
- Dlib -> compiled for CUDA library
- Gstreamer -> using pipelined decoders instead of ffmpeg
- Ethernet to Gstreamer - ??? Not sure
How can I verify whether the CPU is utilized or GPU and other accelerators are used ? Any tools or commands provided by NVIDIA ?