Reconstruction with known camera poses
Hi, Thanks for the great work and releasing the code!!
Is it possible to incorporate known camera poses to mast3r pipeline similar to dust3r https://github.com/naver/dust3r/issues/30#issuecomment-1986012830
Best, Kevin
Hi, this is not implemented right now.
When will this be available?
Same thing here, I hope it will release soon!
Could I inquire about the preset camera pose in dust3r? I have posted a comment on the dust3r forum, but have not received a reply yet. My issue is that even when I preset the camera pose, it remains the same as when it is not set. Since my goal is to align the mesh and camera pose with the world coordinates, the predicted pose is on the opposite side of the ground truth.
Another issue is that the pose is close to the identity matrix. Consequently, projecting a 2D image point to 3D results in a completely different location.
Here is the link to the issue. It's just simple code with straightforward usage; I'm not sure why it can't set the camera pose..
https://github.com/naver/dust3r/issues/103