FootFormer: Estimating Stability from Visual Input

This publication introduces FootFormer, a spatiotemporal transformer model designed to predict physical stability metrics from monocular video. The researchers utilized Moticon OpenGo Sensor Insoles as a critical component of the UnderPressure (UP) dataset used for model validation. Specifically, the Moticon insoles provided the ground truth for 16 plantar pressure sensors and vertical ground reaction forces (vGRFs), which were used to define binary foot contact labels for the heel and toe regions. By providing high-fidelity pressure and contact data, the Moticon sensor insoles enabled the training and evaluation of FootFormer's ability to accurately regress foot-ground interactions, which are essential for calculating classic kinesiology metrics like the Center of Pressure (CoP) and Base of Support (BoS).