3D Human Generation

3D human generation aims to create realistic, three-dimensional models of humans from various input sources, such as single images, videos, or text descriptions. Current research heavily utilizes diffusion models, often coupled with techniques like Gaussian splatting or neural radiance fields, to generate high-fidelity models with detailed textures and consistent multi-view rendering. These advancements are driven by the need for improved virtual try-on, animation, and gaming applications, impacting fields like computer vision, computer graphics, and virtual reality. The focus is on achieving greater realism, controllability (e.g., through text prompts or semantic editing), and efficiency in both training and rendering.