CVPR 2023 - NeuralEditor: Editing Neural Radiance Fields via Manipulating Point Clouds

In this episode we discuss NeuralEditor: Editing Neural Radiance Fields via Manipulating Point Clouds by Authors: - Jun-Kun Chen - Jipeng Lyu - Yu-Xiong Wang Affiliations: - Jun-Kun Chen and Yu-Xiong Wang: University of Illinois at Urbana-Champaign - Jipeng Lyu: Peking University. The paper introduces NeuralEditor, a new system that allows for easy shape editing of neural radiance fields (NeRFs), which are typically difficult to edit. The system uses the point cloud representation of the scene to build NeRFs and introduces a rendering scheme based on deterministic integration within density-adaptive voxels. The system enables precise point cloud reconstruction and achieves state-of-the-art performance in shape deformation and scene morphing tasks. Code, benchmark, and demo video are available.