Invited speakers
Maaike Van Roy, Postdoc at KU Leuven
Talk: Analyzing Soccer Actions and Tactics by Learning and Reasoning
Abstract: This talk presents several approaches to reasoning about models learned from in-game event data in soccer. It highlights practical insights that such reasoning can enable. In particular, the talk focuses on two common tasks in soccer analytics for which learned models are commonly used: action valuation and tactical analysis. First, it demonstrates how reasoning techniques can be used to contextualize the predictions of action-value models and how this can be used to debug the data provided to the model. Second, it explores how reasoning can provide deeper insights into the offensive and defensive tactics of teams.
Atom Scott, Founder of Playbox Inc.
Talk: When Sports CV Leaves the Lab: Lessons on Bridging Academia and Industry
Abstract: What happens when sports computer vision moves out of the lab and into a company? This talk is a field report from that journey, going from a college-level sports analyst, to research on tracking in sports (including publishing at CVsports), to building a product and founding Playbox. The talk covers how I ended up here, what broke when the technology left the lab, a few lessons I did not expect, and where I think the value in sports CV can be created/captured today.
Christina Chase, Co-founder and Managing Director at MIT Sports Lab
Talk: From Tracking to Understanding: The Next Decade of Computer Vision in Sports
Abstract: Computer vision has transformed sports through advances in detection, tracking, pose estimation, and 3D reconstruction. Today, we can measure athlete movements, ball trajectories, and game events with remarkable precision. Yet some of the most important decisions in sport—including officiating and judging—cannot be resolved through perception alone. They require context, reasoning, and an understanding of what actions mean within the game.
This talk explores the evolution of sports computer vision from tracking what happened to understanding why it happened. Drawing on examples from officiating in soccer and basketball, as well as judging in performance-based sports, I will discuss how advances in multiview geometry, temporal modeling, and multimodal learning are enabling systems to move beyond localization and event detection toward context-aware decision support. I will highlight emerging challenges including intent, uncertainty, explainability, fairness, and human-AI collaboration, and argue that the next decade of sports computer vision will be defined not by what we can track, but by what we can understand. Ultimately, I will suggest that officiating and judging represent some of the most compelling grand challenges in sports computer vision, requiring systems that can not only perceive the game, but reason about it.
Johsan Billingham, Senior research manager at FIFA
Talk: FIFA Research – Case Studies and Research Highlights