Congratulations to the SAM-3D team for winning the CVPR Best Paper Honorable Mention! This is truly an outstanding work, and I’ve been closely following your research recently.
However, I’ve run into difficulties constructing the ground-truth poses, which has become a major roadblock for me. Could you please share a robust dataloader to help build training datasets that pair geometric data with object poses?
Best regards
Congratulations to the SAM-3D team for winning the CVPR Best Paper Honorable Mention! This is truly an outstanding work, and I’ve been closely following your research recently.
However, I’ve run into difficulties constructing the ground-truth poses, which has become a major roadblock for me. Could you please share a robust dataloader to help build training datasets that pair geometric data with object poses?
Best regards