ulsrgs/hloc-glomap

Hierarchical Localization and GLOMAP

A repo to use GLOMAP Structure-From-Motion mapper with Hierarchical Localization toolbox for a faster mapper with deep learned features and matchers. Uses Rerun to visualize, Gradio for an interactive UI, and Pixi for a easy installation

Install and Run

Using Pixi

Make sure you have the Pixi package manager installed

first clone the repo and go to the parent directory

https://github.com/pablovela5620/hloc-glomap.git
cd hloc-glomap

then within the parent directory to use gradio interface

pixi run app

to use cli

pixi shell
python tools/reconstruct.py --help

Running splatting DATA_DIR="$OUTPUT-PATH" pixi run train-splat

All commands can be listed using pixi task list

Time Comparison

**Note**: The time comparison shown above is based on a single example and is very anecdotal. It may not represent the performance in all scenarios.

The above compares glomap with colmap using the same feature+matcher but different pair generations (sequential v.s. vocab tree) on 300 images

Matching Type	GLOMAP Reconstruction Time	COLMAP Reconstruction Time
Vocabulary Tree Matching	18 minutes 40 seconds	31 minutes 52 seconds
Sequential w/ Loop Closure	8 minutes 51 seconds	19 minutes 21 seconds

Acknowledgements

Thanks to the original GLOMAP and hloc repos!

hloc

@inproceedings{sarlin2019coarse,
  title     = {From Coarse to Fine: Robust Hierarchical Localization at Large Scale},
  author    = {Paul-Edouard Sarlin and
               Cesar Cadena and
               Roland Siegwart and
               Marcin Dymczyk},
  booktitle = {CVPR},
  year      = {2019}
}

GLOMAP

@inproceedings{pan2024glomap,
    author={Pan, Linfei and Baráth, Dániel and Pollefeys, Marc and Sch\"{o}nberger, Johannes Lutz},
    title={Global Structure-from-Motion Revisited},
    booktitle={European Conference on Computer Vision (ECCV)},
    year={2024},
}