A research team from Zhejiang University recently launched an innovative AI framework called Human101, which is capable of generating high-fidelity 3D digital humans in an astonishing 100-second training time using only single-view videos. Human101 uses 3D Gaussian scatter technology to achieve a real-time rendering speed of more than 100 FPS, and is open source on GitHub to facilitate developers to use and improve. This breakthrough technology brings new possibilities to fields such as virtual reality and the metaverse.
The research team of Zhejiang University launched Human101, a new AI framework that can generate high-fidelity 3D digital humans in only one-view video with 100 seconds of training time. Using innovative 3D Gaussian scatter technology, it achieves a real-time rendering speed of 100+ FPS and is an open source project on GitHub. By introducing standardized human body initialization and forward Gaussian animation methods, Human101 successfully solves the contradiction between modeling and real-time rendering, providing a new solution for virtual reality.The open source release of Human101 will greatly promote the development of 3D digital human technology and provide researchers and developers with a powerful tool and new research direction. Its efficient training speed and real-time rendering capabilities bring huge application potential to fields such as virtual reality and game development. It is worth looking forward to more innovative applications in the future.