Hello,
Do you plan to train and release newer feedforward 3D reconstruction models like VGGT, StreamVGGT, CUT3R, etc with Puzzles augmentation? The improvements mentioned in the priprint are impressive. If similar improvements transfer to these newer models, it would be immensely useful for the community.
Thanks.