What I explored
- Player detection + tracking from a single broadcast feed.
- Mapping 2D detections onto a 3D pitch for spatial analysis.
- Interactive analysis: select players, inspect formations, and replay sequences.
Case Study
Immersive Sports Entertainment
A longer-term exploration to combine live football broadcasts with computer vision and XR: detect and track players from a stream, render a miniature 3D pitch in front of the viewer, and enable analysis like selecting players and replaying plays.
Why it paused
Football broadcasts often cut to highlights and close-ups, and many players are off-frame. Without tracking data, estimating missing positions was unreliable compared to sports where all players stay visible.
Still interested
With recent ML advances, there’s new potential to infer player positions and build a more accurate real-time 3D view from broadcast-only inputs.
Demo