LERF: Language Embedded Radiance Fields

Summary: Researchers have proposed a method called Language Embedded Radiance Fields (LERFs) which can ground language embeddings from off-the-shelf models into NeRF, enabling open-ended language queries in 3D. LERF can extract 3D relevancy maps for a range of language prompts interactively in real-time, with potential use cases in robotics, vision-language models, and interacting with 3D scenes.