amuya_sfm_lines_beta

This repository contains an implementation of the line-based SfM pipeline proposed in the paper "Structure from Motion with Line Segments Under Relaxed Endpoint Constraints" by Branislav Micusik & Horst Wildenauer.

Link to the paper: Click here

Please note that this repository does not include all of the steps mentioned in the paper and some of the steps have also been modified.

How to use it?

1. Clone repository

All necessary data and codes are inside the src directory.

2. Create a conda environment and install python packages

conda create -n vis3d python=3.7.10

pip install -r requirements.txt

3. Run examples

There are the three following examples:

1. main.py: It implements the general pipeline to perform line-based SfM on natural images.
2. main_toy_example1.py and main_toy_example2.py: To test our pipeline, we created these two toy examples.

To test on your data, follow the template of main.py

4. How to cite us

If you find this implementation useful, please cite us as:

@software{bryan_german_pantoja_rosero_2021_4946844,
  author       = {Bryan German Pantoja-Rosero and Amir Rezaie},
  title        = {Amuya Line-based SfM},
  month        = jun,
  year         = 2021,
  publisher    = {Zenodo},
  version      = {v0.0},
  doi          = {10.5281/zenodo.4946844},
  url          = {https://doi.org/10.5281/zenodo.4946844}
}