GLOMAP support

[danylomm]

GLOMAP, the COLMAP's successor from the same authors looks very promising:

In terms of accuracy and robustness, we achieve results on-par or superior to COLMAP, the most widely used incremental SfM, while being orders of magnitude faster.

Project page: https://lpanaf.github.io/eccv24_glomap/ Code: https://github.com/colmap/glomap

[pablovela5620]

I wonder if its possible to already support this by just installing glomap via pixi/conda https://prefix.dev/channels/conda-forge/packages/glomap and modifying colmap-command to glomap. I don't know if this will outright work because it would be calling for feature extraction/matching, but you could probably just modify it https://github.com/nerfstudio-project/nerfstudio/blob/d395aa2e537244b6f54fab2082b01b0edb589e62/nerfstudio/process_data/colmap_utils.py#L156 to call glomap and everything else colmap

[danylomm]

According to e.g. this tutorial - https://www.youtube.com/watch?v=QIxXuilEEVw you should use GLOMAP only for the mapper part, and regular COLMAP for steps before mapping. Simply setting colmap-command argument to glomap doesn't work for the same reason, already tried it out.

I think the best integration option from user perspective would be to add glomap to --sfm-tool argument. From documentation:

--sfm-tool {any,colmap,hloc}                                   
    Structure from motion tool to use. Colmap will use sift    
    features, hloc can use                                     
    many modern methods such as superpoint features and    
    superglue matcher (default: any)