Avatar Modeling

Avatar modeling focuses on creating realistic and animatable 3D human representations from various input data, such as images and videos, aiming for high-fidelity rendering and efficient processing. Current research emphasizes the development of hybrid 3D models, combining explicit representations like meshes with implicit methods such as neural radiance fields (NeRFs) and Gaussian splatting, to capture both detailed geometry and appearance. These advancements improve the speed and quality of avatar generation, enabling applications in virtual and augmented reality, gaming, and digital entertainment. The field is also exploring techniques for disentangling different avatar components (body, clothing, hair) to enhance control and facilitate tasks like virtual try-on and animation.