AI Analyzes Olympic Figure Skating to Unlock Quint Jumps
- •MIT researchers develop OOFSkate to analyze elite figure skating technical performance
- •Pose estimation AI overcomes depth challenges to track jump height and rotation speed
- •New study explores whether AI can replicate human aesthetic judgment in artistic sports
MIT Sports Lab is bringing high-tech precision to the ice with OOFSkate, an optical tracking system designed to help elite skaters master the elusive quintuple jump. Developed by researcher Jerry Lu and Professor Anette “Peko” Hosoi, the tool uses computer vision to dissect video footage, providing athletes with specific physical metrics like rotation speed and air time.
The system relies on pose estimation—the AI-driven process of identifying and tracking skeletal joints in motion—to provide feedback that the human eye might miss. While these models often struggle with depth in other sports, figure skating offers a unique environment where vertical height and angular momentum are the primary drivers of success, allowing the AI to function effectively with standard camera angles.
Beyond technical tracking, the team is investigating the "black box" of aesthetic judgment through a collaborative study. They aim to determine if AI models arrive at "beauty" scores through the same reasoning pathways as human experts or if they are simply mimicking patterns in existing data to parrot subjective human opinions.
This research could redefine how we evaluate sports that blend technical prowess with artistic flair. As skaters push the limits of human capability, AI provides the data-driven roadmap to help them rotate faster and jump higher, making the once-impossible five-rotation jump a looming reality in the near future.