The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Deep Learning for Automatic Basketball Video Production with Julian Quiroga - #389
Jul 6, 2020
Julian Quiroga, Computer Vision Team Lead at Genius Sports, dives deep into the world of automated basketball video production. He discusses innovative techniques using Gaussian models for player dynamics and the integration of deep learning to enhance viewer experiences. Challenges like accurate player localization and adapting strategies for different sports are tackled head-on. Quiroga also shares insights into optimizing camera angles and real-time data usage, revolutionizing how basketball games are broadcasted.
AI Snips
Chapters
Transcript
Episode notes
Basketball Passion
- Julian Quiroga was offered a job at Genius Sports to work on basketball products.
- He accepted because he has played basketball since he was ten.
Automated Production Complexity
- Automating sports production depends on the number of camera views used.
- Single-camera systems create virtual cameras, while multi-camera systems require shot selection.
Democratizing Sports Broadcasting
- One goal is to enable schools and colleges without professional cameramen to record and stream games affordably.
- The automated production also allows for real-time data and analytics integration into broadcasts.