- Multi-threading yolov5 inference with Nvidia jetson on-board camera (RaspberryPi camera mudule ver.2 is used).
- Save detection results to video file.
This work is derivated form yolov5-multi-video-inference. So you can use that git repo as reference.
./yolov5-csi-camera -csi [engine]
To interrup program, press "Esc" and you can then access the saved video files.
-
wang-xinyu/tensorrt/yolov5 for yolov5 tensorrt implementation.
-
AlexeyAB/darknet for data structure to pass objects between threads.