- deployed on NVIDIA Jetson
- detecting people
- tracking detected people
- counting in/out of people
- using YOLOv4-tiny as object detection model
- model is trained in Darknet framework and converted to TensorRT inference engine with 608 x 608 input size
- a cam or a video with people to replace
crowd.mp4
inmain.py
- modify the
main.py
: replace 'crowd.mp4' by your video - modify the
main.py
: point_1_y and point_2_y to define a line to represent the boundary for in/out counting - download the engine file: 'yolov4_tiny_608.engine'
- simply type:
python3 main.py