Inquiry about 'model_uid' in the JSON file

[Jun-white]

Question: Regarding the paper's mention of -“In the retrieval process, our goal is to find and match the most suitable 3D object instance for each bounding box in the EmbodiedScan dataset from a pre-curated 3D asset library and place it in the corresponding location within the scene,” I did not see the relevant code on GitHub. I would like to ask whether this process was done manually? Because in the EmbodiedScan paper, only “id”, “category”, and “bbox” are provided.