Vid2Avatar is a technology that can create high-fidelity 3D avatars from real-world videos. This innovation in avatar technology has the potential to transform numerous applications in VR, robotics, video games, cinema, sports, and other fields. It does not require any external segmentation modules, ground truth supervision, or priors generated from huge datasets. Vid2Avatar is a 3D avatar generator that separates the person from the background in a scene and models it as a neural field using two independent neural fields. It also employs a single, temporally consistent representation of human shape and texture in canonical space, as well as an optimisation algorithm to adjust multiple parameters related to the background, the human subject, and their poses to best fit the available data from a sequence of image or video frames.
Where Industry Meet Trends
Menu
English