SAM

Label a 3D point cloud with any word you type. This complete guide walks the open-vocabulary 3D segmentation stack in Python: pretrained SAM cuts the scene, CLIP scores each region, and a scene graph lets an LLM answer questions about the space.