TouchSight Predicts Dense Tactile Force from Egocentric Monocular Video
September 18, 2026
TouchSight uses generative visual augmentation to bridge the gap between gloved training data and bare-hand real-world scenarios. The framework leverages the TwinTouch-20H dataset to predict full-hand contact force from standard egocentric video.
HOW THIS AFFECTS YOU
●
builderYou can implement contact force prediction in robotics using only monocular visual input.
●
researcherYou can use generative video re-rendering to solve domain gaps in tactile sensing datasets.