AI Event Bib Detection Upgrade

I need an experienced computer-vision developer to step into an existing codebase and finish a real-time bib detection system for mass-participation sporting events. The current prototype already recognises bib numbers; your task is to stabilise it and expand the feature set so it can handle marathons, cycling races, triathlons—essentially any event where athletes wear start numbers.

Core goals
• Strengthen and optimise real-time detection accuracy under changing light, motion blur and camera angles.
• Automate the creation of short clips that isolate each participant’s key moments.
• Overlay a configurable sponsor logo on every generated clip.
• Add seamless WhatsApp integration so the system can push the personalised clips straight to athletes or organisers.
• Clean up, document and hand over the code so the next developer—or I—can maintain it easily.

Tech context
The previous developer used Python with OpenCV and a YOLO-style model; feel free to refactor or swap frameworks as long as inference remains lightning fast. Expect video streams up to 4K, multiple concurrent cameras and fairly busy backgrounds.

Success criteria
• ≥95 % accurate bib recognition measured on my test footage.
• Clip creation and logo overlay processed within 30 s of finish-line crossing.
• One-click WhatsApp delivery functioning from both desktop and mobile.
• Well-commented code, README and deployment script.

If you have proven experience in real-time computer vision, OCR, and messaging API integration, let’s talk.