@yocabon
Thank you for your outstanding work. I would like to inquire about the performance of DUST3R and MAST3R in dense view settings. In most scenarios, both methods appear to produce satisfactory 3D reconstruction results, with reasonable outcomes and minimal point cloud redundancy. Although you both use confidence-based pixel masking, I am still surprised by the view-consistent results of the feature points (where the same 3D point has correspondences in multiple RGB images) in dense view settings. Thanks！
@yocabon Thank you for your outstanding work. I would like to inquire about the performance of DUST3R and MAST3R in dense view settings. In most scenarios, both methods appear to produce satisfactory 3D reconstruction results, with reasonable outcomes and minimal point cloud redundancy. Although you both use confidence-based pixel masking, I am still surprised by the view-consistent results of the feature points (where the same 3D point has correspondences in multiple RGB images) in dense view settings. Thanks！