🤖 EgoForce: 3D Hand Reconstruction from Monocular Video
The EgoForce framework has been developed to reconstruct hand pose and forearm geometry from first-person camera video. The system utilizes a unified transformer capable of working with any lens type (fisheye, wide-angle, etc.) and improves accuracy on the HOT3D dataset by 28%.
🌍 This solution is critical for creating affordable AR/VR devices, as it allows for the replacement of expensive stereo cameras or LiDAR with simple monocular cameras.
👤 The technology transforms standard video from wearable cameras into precise 3D data, opening new possibilities for robotics, metaverses, and augmented reality systems.
Source 1: https://dfki-av.github.io/EgoForce/ Source 2: https://github.com/dfki-av/EgoForce/tree/main