Interactive Sports Analytics
summary
A system uses Stats Perform tracking archives and learned similarity metrics to retrieve visually similar plays (example set: 40,000 three-point shots) and returns a ranked list in less than a second. The pipeline combines multi-agent trajectory models, an interactive chalkboard-style interface for sketch or visual query, and pose extraction from broadcast video (body-pose from broadcast video). The demo shows adjustable expected points estimates when moving or swapping players and rapid neighborhood search of past plays.
AI-written summary · about
chapters (6)
- 0:00Start
- 1:54How To Find Similar Plays?
- 2:09Introducing "Chalkboarding"
- 3:03Manipulating Exemplars
- 3:16Drawing Plays
- 3:29Analytics on the Query
related talks
41mThe evolution of basketball with data science
51mVideo Analytics for Football games
29mAI in Sports - Basketball Rule Violation Detection
23mSport analysis with Python
41mEnhancing Event Analysis at Scale: Leveraging Tracking Data in Sports
31mAI reshaping sports tech
27mAnalysis of sports scenes using computer vision models
32mIdentifying playstyles in football through spatial networks