How does AI for sports work?
summary
Gemini combines three concrete inputs: a live video feed for visual context, per-frame tracking from OpenCV tracking ingested into BigQuery, and a predefined prompt that sets tone and commentary rules. The pipeline maps ball and player coordinates to play events, generates timestamped narration and overlay data, and claims professional-level broadcast output; the demo presents an end-to-end system that turns camera plus tracking data into realtime sports commentary.
AI-written summary · about
related talks
01h 45mAI Engineering with the Google Gemini 2.5 Model Family
18mIntroduction to Gemini APIs and AI Studio
42mDeep Learning for Automatic Basketball Video Production
39mBuild the new generation of AI Agents
19mGemini 2.0: Beyond the Basics
36mApps That See 👀 Practical AI Vision for Developers
17mFrom recovering Pro Golfer to AI Entrepreneur
21mHow Google DeepMind is researching the next Frontier of AI for Gemini