Presence Data Company Nuoyiteng Robotics announced the public release of the high-precision hand motion capture data set HipHi-Hand at Hugging Face. The data set has been used to train Nvidia's 3D hand reconstruction model EgoHand. According to Nuoyiteng, EgoHand is built based on the Nvidia Isaac Capture framework. It only requires first-person view video taken with a monocular camera to restore the three-dimensional posture and shape of the hand. The model was released under the OpenMDW license, which allows free commercial use. According to Nvidia, the model uses 14 million images for training. The material comes from synthetic data, real scene video, and multi-camera motion capture true value data. In the public first-person perspective benchmark, zero sample accuracy reached the current SOTA level, and the single-frame inference delay was about 9 ms.

Presence Data Company Nuoyiteng Robotics announced the public release of the high-precision hand motion capture data set HipHi-Hand at Hugging Face. The data set has been used to train Nvidia's 3D hand reconstruction model EgoHand. According to Nuoyiteng, EgoHand is built based on the Nvidia Isaac Capture framework. It only requires first-person view video taken with a monocular camera to restore the three-dimensional posture and shape of the hand. The model was released under the OpenMDW license, which allows free commercial use. According to Nvidia, the model uses 14 million images for training. The material comes from synthetic data, real scene video, and multi-camera motion capture true value data. In the public first-person perspective benchmark, zero sample accuracy reached the current SOTA level, and the single-frame inference delay was about 9 ms.