Long video input for visual odometry

[Bing1002]

Hi @jytime Great work. I am wondering if you ever tried long sequence video input as in visual odometry. Also if the number of input frames is limited by resources, we need to do alignment for each small batch input due to the relative scale . I appreciate your any comments.

[jytime]

Please check:

https://github.com/facebookresearch/vggt/issues/37