We use our hands constantly. We communicate through gestures and manipulate objects in our surroundings: We grab, carry, and touch in many different ways. Capturing hands in action is a challenging computer vision problem and the 4D data data we collect in this facility helps us train neural networks to understand how we interact with the physical world.

In the Hand Scanner, built for us by 3dMD (Atlanta, GA), we capture the shape and texture of the hands in high detail with up to 60 frames per second. The system contains 48 machine vision cameras, combined in 16 Modular Camera Units (MCUs), synchronized with speckle pattern projectors and LED lights.

Depending on the research goals, we can record a single hand or both hands. A glass table allows us to capture the interaction of a person with objects.