Currently, there are several approaches to 3D model editing, each with distinct focuses. One category utilizes diffusion editing methods, which combine specific loss functions to extend 2D image editing based on text prompts to 3D, thus enabling 3D model editing. Another approach integrates large language models (LLMs) to locate target areas and modify their color and texture.
3D Representations
Various 3D representations have been proposed to address corresponding tasks. The first and most groundbreaking work is NeRF, which uses volume rendering to complete reconstruction tasks with only 2D image supervision, fully preserving the information in the scene. However, due to this comprehensive data preservation,