how to get the image of every question
I’m trying to understand how to retrieve the corresponding image for each question. Since the dataset only provides the scene name without specifying the frame ID, could you kindly clarify whether the input data is based on point clouds?
Thanks for your interest in our work. By "corresponding image for each question", I think you are referring to the visual inputs used in the MLLM setting, i.e. bird’s-eye view(BEV) image and eight uniformly sampled video frames annotated with object markers. These visual data follow GPT4Scene: BEV is generated using the point cloud while object markers are annotated using camera pose as well as object position & size provided within ScanNet, MultiScan, 3RScan and ARKitScenes. They are publicly available on Google Drive: https://drive.google.com/drive/folders/1TPHmH8B9iJB61hrhO2YBxCOGh0hQAX1y. We also mention this link on our code repo: https://github.com/anywhere-3d/Anywhere3D.
Best regards