Marathon Multimedia Management System Creation

I’m looking for a complete end-to-end solution that can automatically detect marathon bib numbers from video footage, generate personalized runner clips, and deliver them directly via WhatsApp (or a similar messaging API).

The system will use two fixed cameras capturing race footage and must intelligently process, segment, and distribute clips with minimal manual intervention.

Core Requirements

1. Bib Detection (OCR / Computer Vision)

High-accuracy detection of marathon bib numbers
Must perform reliably under varying lighting, motion, and crowd conditions
Prefer models trained or fine-tuned specifically for race bibs

2. Dual Camera Processing

Ingest synchronized video streams from two fixed HD cameras
Automatically select the best angle (or combine views) for each runner

3. Processing Speed

Near real-time processing preferred
Clips should be available within minutes after the runner passes

4. Automatic Clip Generation

Each clip should:
Start a few seconds before the bib becomes visible
End a few seconds after the runner exits the frame

5. Branding Overlay

Add a preset sponsor logo to each clip (corner overlay)
Ability to easily update/replace the logo

6. Database Integration

Connect to a database (SQL, Firebase, or similar)
Match detected bib numbers to phone numbers

7. Automated Messaging

Send compressed, share-ready clips via WhatsApp or similar API
Include a customizable message template
System should be future-proof for alternative messaging platforms

8. Simple Admin Dashboard (Non-Technical Users)
A browser-based interface that allows a non-technical volunteer to:

Start/stop recording
Monitor detection performance
Resend or delete clips
Upload/edit bib-to-phone mappings (CSV or spreadsheet)
Change the overlay logo
Deliverables
Fully functional prototype (cloud VM or local server) ready for race-day use
Complete source code with clear documentation and installation steps
Admin guide (PDF or short video) covering setup, usage, and troubleshooting
One remote handover session for live Q&A
Additional Notes
Open to your preferred tech stack (e.g., OpenCV, PyTorch, TensorFlow, FFmpeg, or commercial APIs)
Licensing must be suitable for event/commercial use
The system must be extremely simple to operate for non-technical users
When Applying, Please Include
Proposed architecture
Expected detection accuracy
Recommended hardware specifications (for smooth race-day performance)

Important: This system must be ready-to-use and designed for non-IT operators.