Building an AI That Watches Rugby

a year ago

Gainline aims to enhance rugby fan experience by providing deeper match context beyond structured event feeds.
Current data lacks insights into referee decisions, player dominance in scrums, and other nuanced game aspects.
A prototype AI system was built to watch rugby games, extracting scores, game clock, and transcribing audio for richer data.
OpenAI's vision model successfully identified match state details but required cost optimization by cropping images to essential parts.
Alternative methods like image diffing and OCR were explored but proved less reliable than using LLMs for text extraction.
Whisper was used to transcribe referee and commentary audio, adding valuable context to the game's narrative.
The project highlights potential for AI in sports analytics but faces challenges in scaling, infrastructure, and ethical considerations.

Hasty Briefsbeta