questions about obtaining camera pose

[a510721]

Thank you for sharing your excellent research. I am a student conducting research on robotics, and I would like to use your impressive algorithm. I plan to mount a monocular camera on a robot and execute VGGT while it moves.

(1) If the camera is calibrated and the internal matrix is known, is there a way to provide this as an input to the VGGT algorithm? If so, I would greatly appreciate it if you could provide an example code.

(2) If the camera pose matrix can be obtained through odometry on the robot, even if it is not accurate, is there a way to provide this as an input to the VGGT algorithm? If so, I would greatly appreciate it if you could provide an example code.

(3) The most important and meaningful question is whether, after generating a map or 3D point model like using VGGT, it is possible to obtain the pose of the camera at a specific location when an image from the camera is input. If so, I would greatly appreciate it if you could provide an example code.

Thank you.

[jytime]

Hi,

We do not support (1) or (2) by now. For (3), can you share more details? Is (3) relocalization?

[hiyyg]

Hi,

We do not support (1) or (2) by now. For (3), can you share more details? Is (3) relocalization?

Thanks for the info.

[yyypsycheguy]

Hi, I'm also in need for request 3). I would like to relocalize the coordinates when given one single picture from the series of pictures taken out of the video.

"(3) The most important and meaningful question is whether, after generating a map or 3D point model like using VGGT, it is possible to obtain the pose of the camera at a specific location when an image from the camera is input. If so, I would greatly appreciate it if you could provide an example code. "

Appreciated!