we are
Metric
Hive
An advanced implementation of Apple's CoMotion AI tested on high-intensity NFL footage. Unlike traditional frame-by-frame detection, this system continuously tracks player identities through heavy occlusions, camera zooms, and pile-ups.
Temporal Memory
Standard systems lose track when a player is hidden. CoMotion assigns a "memory state" to each player, remembering their movement patterns to bridge gaps during occlusions.
We utilized Cross-Attention mechanisms, allowing the AI to focus on subtle cues (like feet visible under a pile-up) to maintain the tracking lock when the body is obscured.
Sports Tech
Computer Vision
Concurrent 3D Motion
The system tracks detailed 3D poses of multiple people simultaneously from monocular video. Even with aggressive camera panning, the spatial stability outperforms traditional "detection-then-match" paradigms.
This capability is critical for the next generation of injury prevention analytics and immersive fan experiences, where maintaining identity during a tackle is the hardest technical hurdle.
(Project)
