people_counter/README.md

People Counter with YOLOv8, ByteTrack, and Prometheus

This project implements a real-time people counting system based on object detection and tracking using YOLOv8 and ByteTrack. It supports RTSP input, in/out counting using a vertical line, and exposes Prometheus metrics for integration with monitoring systems.

🚀 Features

• 🎥 RTSP camera stream input
• 🧠 YOLOv8 person detection
• 🔁 ByteTrack object tracking
• ➕➖ In/Out counting via virtual vertical line
• 📊 Prometheus metrics at http://localhost:9100/
• 🌐 Streamlit UI for live video and statistics

Start mediamtx

mediamtx

Stream webcam

ffmpeg \
    -vaapi_device /dev/dri/renderD128 \
    -y \
    -f v4l2 \
    -framerate 30 \
    -video_size 1280x720 \
    -input_format mjpeg \
    -i /dev/video0 \
    -vf 'format=nv12,hwupload' \
    -c:v h264_vaapi \
    -f rtsp \
    rtsp://localhost:8554/cam

Run application

streamlit run streamlit_app.py