Closed DevLinyan closed 6 months ago
Yes, the provided 2D masks can be from Grounded-SAM, where each mask has category information. Or you can use Grounding DINO to pick the output masks of SAM inside the 2D bounding boxes. The 3D-2D mapping relation can lift this category information to 3D masks directly.
Could you please provide information on whether the model is equipped to identify and classify semantic categories such as desk or table? If yes, where it is.