Virtual Italian Billiards Vision System

I need a complete Italian billiards / pool experience that serves both professional competition and serious training. The table will be surrounded by four fixed corner cameras; their live feeds must track the cue stick tip in real time, so the system can regenerate a perfect top-down and always-frontal view of both players while they shoot.

Key functions I expect:

• Facial-recognition login – as soon as a player approaches the table, the system confirms identity and loads personal stats. No emotion detection or gesture control is required; simple, reliable player identification is all I need.

• Automated scorekeeper – the program records every shot, updates the scoreboard automatically, and stores match history.

• Virtual video compositing – by using computer-vision image processing, the raw camera feeds are overlaid onto a clean 3-D or photographic background, producing a polished, TV-style scene that can be exposed as a virtual camera in OBS for live streaming.

Target platforms are PC (Windows) and Web (WebGL or comparable). The final build should run locally on a solid gaming PC; the web version must stream smoothly in modern browsers without plugins.

Deliverables I would like to see:
1. Calibrated multi-camera tracking module with reliable cue tip detection.
2. Facial-recognition module tied to a lightweight player database.
3. Real-time scoring engine compatible with Italian billiards and standard pool rule sets.
4. Virtual video output visible as a selectable source inside OBS.
5. Clear setup guide, source code, and short demo video.

Acceptance criteria: stable 60 fps output on a GTX-level GPU, cue tip tracking error under 5 mm, facial recognition accuracy above 95 % in normal club lighting, and seamless OBS integration.

If you have solid experience with OpenCV, Unity or Unreal, and OBS virtual devices, I’m ready to discuss milestones and move forward quickly.