Fostering the Metaverse Immersion: Unraveling Personalized Dynamic Human Avatars
Sirisha Talapuru, Ram Dantu, Shakila Zaman, Vinh Quach, Apurba Pokharel · 2024
The Metaverse, utilizing cutting-edge semiconductor and internet technology, offers captivating online experiences by integrating blockchain, AI, and extended reality. This facilitates social interactions through virtual representations labeled as human avatars. However, the absence of lifelike natural avatars hinders full engagement and immersion within the Metaverse. Current avatar development methods rely heavily on expensive hardware like 3D scanners, limiting the widespread usage. This paper introduces GenesisImaging (GI), a novel approach utilizing high-quality 2D images to generate highly realistic human avatars. GI, along with Shape from Silhouette (SFS), employs advanced computer vision and graphics techniques for lifelike movement and mobility of the avatars. By leveraging SFS, highly accurate 3D models are generated without 3D scanning. These models are enhanced with motion using Inverse Kinematics (IK), resulting in realistic behavior. Experiments demonstrate that GI, SFS, and dynamic mapping produce remarkably realistic avatars with natural movement and behavior.