How to use pre-computed poses in this framework?

[lanlan96]

Hi, I want to see how Go-SLAM can improve the accuracy of camera poses derived from other system. I changed the self.video.append in

GO-SLAM/src/motion_filter.py

Line 79 in e954a85

self.video.append(timestamp, image[left_idx], None, None, depth,

and

GO-SLAM/src/motion_filter.py

Line 63 in e954a85

self.video.append(timestamp, image[left_idx], IdentityMat, 1.0, depth,

to replace the pose input None or IdentityMat with my poses (the shape is 7: [tx,ty,tz,qx,qy,qz,qw]). I test scene0169_00 in ScanNet. However, the optimization seems to collapse and not converge. The ATE error is more than 1m. I would appreciate it if you could give some suggestions. Thanks!