"How to utilize COLMAP's trajectory for training."

[liweishuo-bigai]

When training on the Replica and ScanNet datasets, I found that using the datasets' built-in trajectories yielded high-quality meshes. However, replacing them with COLMAP output resulted in poorly reconstructed and incomplete meshes. I attempted to normalize the primary axis in the COLMAP poses, scaling the other two axes by the same factor, but the outcome remained chaotic. Additionally, I observed that adjusting the parameter “3” in the formula scale = 2. / (np.max(max_vertices - min_vertices) + 3.) influenced the results. Could you explain the purpose of this parameter? Do you have any suggestions for effectively incorporating COLMAP trajectories into the training process?

[Golbstein]

you need to find the scale from your 3d points and not from camera positions