You Are Here: Geolocation by Embedding Images and Maps

In this repository, we provide metadata and code for the route-based localisation algorithm described in our papers:

• You Are Here: Geolocation by Embedding Maps and Images
• Image-based Geolocalization by Ground-to-2.5D Map Matching

Currently, we provide:

• Metadata for each of the three testing areas. (latitude, longitude, yaw, neighbors)
• Original single-modal location embeddings in the testing areas. (./features/ES)
• Improved single-modal and newly proposed multi-modal location embeddings in the testing areas. (./features/MES)
• Binary Semantic Descriptors (BSD) for each location in the testing areas. (./features/BSD)
• Testing routes used for the results reported in the paper. (./test_routes)
• Localisation algorithm implemented in Matlab.

Prerequisites

• Matlab
• Tested on Mac and Linux

Instructions to run an experiment

1. Define parameters in the corresponding configuration file depending on the type of features. Use ESParams.m for old-version embedded descriptors, BSDParams.m for the case of BSDs, and MESParams.m for the multi-modal location descriptors.
2. Run Localisation.m. Results are stored in an array called "ranking.mat" which contains the summary of the position of the groud truth route among all possible candidates. Only the top-k positions are considered (defined in the configuration file). A zero rank means the ground truth route is not in the top-k. Also, a struct with the top-5 best-estimated routes is saved.
3. Results will be automatically saved in ./results directory with a path depending on the parameters selected.
4. Plot results using calculate_accuracy.m script.
5. To generate your own testing routes, please run "generate_random_routes.m". To use extra turn information, please run "generate_turns.m".
6. To plot figures presented in the paper, please refer to "bar.m", "culling_plot.m" and "mse_plot.m". All figures in '.eps' format are saved in ./figures directory.
7. To make video of route-based localization, please use "pick_a_route.m" and "video_make.m".

Disclaimer

We make no claims about the stability or usability of the code provided in this repository. We provide no warranty of any kind, and accept no liability for damages of any kind that result from the use of this code.

Citation

@InProceedings{noe2020eccv,
author = {N. Samano, M. Zhou, A. Calway},
title = {{ You Are Here: Geolocation by Embedding Maps and Images}},
booktitle =  In Proc. of the European Conference on Computer Vision (ECCV),
year = {2020},
}

@article{zhou2023image,
  title={Image-based Geolocalization by Ground-to-2.5D Map Matching},
  author={M. Zhou, L. Liu, Y. Zhong, A. Calway},
  journal={arXiv preprint arXiv:2308.05993},
  year={2023}
}