Thank you very much for open-sourcing this powerful and interesting project, as well as providing many algorithm visualization tools. I would like to base my subsequent research on your work. However, I encountered an issue with the dataset loading. It seems that the image pairs in the data preprocessing provided by dust3r have problems. For example, the ScanNet++ dataset is listed as v1, but currently, only v2 of the ScanNet++ dataset is available for download. I noticed that the MUSt3R_512.pth you released uses the ScanNet++ v2 dataset compared to the CVPR version. I am wondering if the image pairs used in this dataset can be made public. If so, would it also be possible to share the dataset loading code used in the paper? Once again, thank you for your contribution to the 3D reconstruction community.
Thank you very much for open-sourcing this powerful and interesting project, as well as providing many algorithm visualization tools. I would like to base my subsequent research on your work. However, I encountered an issue with the dataset loading. It seems that the image pairs in the data preprocessing provided by dust3r have problems. For example, the ScanNet++ dataset is listed as v1, but currently, only v2 of the ScanNet++ dataset is available for download. I noticed that the MUSt3R_512.pth you released uses the ScanNet++ v2 dataset compared to the CVPR version. I am wondering if the image pairs used in this dataset can be made public. If so, would it also be possible to share the dataset loading code used in the paper? Once again, thank you for your contribution to the 3D reconstruction community.