This is a pipeline to create an instance map of a scene from RGB+ Depth Data and camera pose information. We use standard models like RAM, DINO, and SAM to create an instance level map. The data association for each object is done using CLIP embeddings and 3D position data.