Generative Neural Radiance Field

Generative Neural Radiance Fields (NeRFs) are a class of neural networks designed to create realistic 3D scene representations from 2D images, enabling novel view synthesis and 3D manipulation. Current research focuses on improving efficiency, achieving zero-shot generation from single images, and enhancing control over attributes like pose and appearance through techniques such as latent-space encoding, multiplane representations, and compositional modeling. These advancements are significant for applications in computer graphics, robotics, and other fields requiring high-fidelity 3D scene understanding and manipulation from limited 2D data.

Papers