Watches the video
Point it at a live stream (HLS, RTMP), a video URL or an upload. It finds the players and the ball in every frame, tracks them, and maps them onto the pitch or court as the camera pans.
EyesOnPlay
Real-time sports video analysis from a single broadcast camera. Players, ball, events and a live TV-style mini view, for football and tennis. Open source.
What it does
Section titled “What it does”Calls the events
Football: possession, passes, shots, ball out, corners. Tennis: serves, hits, bounces in or out, faults and who won the point. Every event is JSON with a confidence.
Shows it live
The dashboard plays the video with detections drawn on it, next to an animated TV-style mini pitch or court, with a live event feed, all over WebSocket.
Feeds your systems
An API-key protected feed streams matches and events to betting, media or club platforms in real time.
The live dashboard
Section titled “The live dashboard”
How it works
Section titled “How it works”Video ─▶ FFmpeg ─▶ detection (YOLO, TrackNet) ─▶ ball + player tracking ─▶ pitch/court calibration ─▶ event engine ─▶ Redis ─▶ API + WebSocket ─▶ dashboard and integration feed
One camera, no stadium installation. Calibration comes from the lines on the pitch or court, so it works on ordinary broadcast footage. Nothing is invented: when the camera shows a close-up, positions are marked unknown instead of guessed. Read more.
Get going
Section titled “Get going”Try it in 2 minutesdocker compose up: a simulated match, no GPU needed.
Deploy to productionHTTPS, sign-in, GPU worker and backups on one server.
IntegrateRead matches and live events with an API key.
Commercial licenceClosed-source or SaaS use without the AGPL obligations.