Mihai Dusmanu

Hey. Thanks for reaching out - sounds like a neat idea! I was planning to work on a cleaner implementation at some point, so why not integrate it with Kornia....

I have never installed PyCOLMAP on Windows so I am not sure how to fix this. Are you following the official installation instructions using VCPKG - https://github.com/colmap/pycolmap#windows?

I tested this on a local reconstruction and it's working fine (low reprojection errors). Could you provide the sparse folder such that we can reproduce the issue?

Currently, there is no easy way to run the refinement on two-views only. I will try to explain it here; the process is similar to [`compute_match_graph.py`](https://github.com/mihaidusmanu/local-feature-refinement/blob/master/two-view-refinement/compute_match_graph.py). You first have to...

Hello. Inside our pipeline, we use 33x33 patches for refinement and the coordinates inside these patches are normalized such that the top left corner is (-1, -1) and bottom right...

Hello. I remember that some scenes used the default COLMAP parameters (notably the `image_undistorter --max_image_size`), while some others used some custom values. This is why in the script I try...

Another reason why the image sizes might be mismatched is due to EXIF information (notably image rotation) > maybe at some point they were ignored but now they are taken...

Do you by any chance have one such image where the undistortion fails at hand? I can try to look into it and submit a fix upstream if I figure...

Fixed in the latest commit. The link was moved to dusmanu.com

Current ETA for the code is mid-January. I need some extra time to clean up everything and prepare the release! Leaving this issue open in the meanwhile!