EventEgoHands++: Event-based Egocentric 3D Hand Mesh Reconstruction with Real Dataset

This paper proposes EventEgoHands++, a framework for 3D hand mesh reconstruction from an egocentric viewpoint using event cameras to address background noise and lack of instance-level information. The method incorporates a Hand Detector to estimate bounding boxes and masks for both left and right hands, combined with an Adaptive Attention mechanism to gate attention based on detection results. They also construct EEH-R, the largest real-world dataset comprising approximately 1M annotated frames.

EEH-R is the largest real-world event-based egocentric hand dataset to date, comprising approximately 1M annotated frames.