Automated Bib-Recognition Video System

I need an end-to-end solution that can automatically detect marathon bib numbers from footage captured by two fixed cameras, cut the raw video into individual participant clips, add a preset sponsor logo overlay, and then push a low-resolution copy of each clip directly to the runner via WhatsApp (or a future-proof messaging API) by matching the detected number with a phone record in my database.

Core requirements • Accurate OCR or vision model tuned specifically for marathon race bib numbers, even under varied lighting and motion. • Synchronized ingest from two fixed HD cameras; the system must decide which angle to use (or combine) for the cleanest view of each runner. • Real-time or near-real-time processing so clips are available minutes after the athlete crosses the filming zone. • Automatic clip generation: each video should start a few seconds before the bib is first visible and end a few seconds after the runner leaves the frame. • Optional logo overlay on a corner of every clip. • Database connection (SQL or Firebase preferred, open to alternatives) that links detected bibs to phone numbers. • Automated WhatsApp dispatch of a compressed, share-ready file together with a customizable text message. • Simple, browser-based admin dashboard that a non-technical volunteer can use to:

• start/stop recording
• monitor detection accuracy
• resend or delete clips
• upload/edit the bib-to-phone spreadsheet
• swap the overlay logo.

Deliverables

1. Fully working prototype deployed on a cloud VM or local server, ready for race-day use.
2. Source code with clear documentation and installation script.
3. Admin guide (PDF or short video) showing setup, operation, and troubleshooting.
4. One remote handover session for live Q&A.

I’m open to your preferred stack—OpenCV, TensorFlow, PyTorch, FFmpeg, or a commercial vision API—so long as licensing fits an event production environment and the admin UI remains dead-simple. Please outline your proposed architecture, expected accuracy, and any hardware specs I should budget for when you bid.

READY TO USE SYSTEM NEEDED FOR NON IT SKILLED PERSONS. ADMIN is NON IT SKILLED. Budget: EUR 30–250 Skills: Python, C++ Programming, OCR, Video Production, Video Editing, Video Processing, Computer Vision, Database Management