USVLoc: Infrared Topographic Based Cross-view Localization Dataset in Maritime Environment

USVLoc is a dataset designed for research in autonomous surface vehicle localization. The dataset provides paired visual data (RGB and infrared) with precise geographical coordinates and heading information.

Samples from the training set of the dataset sorted by column, with the corresponding RGB, IR, and rendered images for each sample.

Data Format

Directory Structure

usvloc/
├── train/
│   ├── data.csv          # Metadata: id, lat, lon, hdt
│   ├── rgb/              # RGB images
│   └── ir/               # Infrared images
├── val/
│   ├── data.csv
│   ├── rgb/
│   └── ir/
├── test-0/ through test-4/
│   ├── data.csv
│   ├── rgb/
│   └── ir/
├── grid.csv              # Grid reference metadata
└── grid/                 # Topographic grid images

CSV Format

Each split contains a data.csv file with the following columns:

• id: Unique identifier for the image pair
• lat: Latitude in decimal degrees (WGS84)
• lon: Longitude in decimal degrees (WGS84)
• hdt: Heading/bearing in degrees (0-360, True heading)

Downloads

File Name	Size (GB)	Image Count	Description	Download
train.tar.gz	36.0	57 804	Training split with RGB and IR image pairs and metadata	Download
val.tar.gz	4.0	6 423	Validation split for hyperparameter tuning	Download
test_0.tar.gz	0.6	900	Test split 0 for evaluation	Download
test_1.tar.gz	0.6	900	Test split 1 for evaluation	Download
test_2.tar.gz	0.5	901	Test split 2 for evaluation	Download
test_3.tar.gz	0.6	901	Test split 3 for evaluation	Download
test_4.tar.gz	0.2	753	Test split 4 for evaluation	Download
grid.tar.gz	1.2	91 290	Topographic grid reference images for cross-view localization	Download

✒️ Usage and Citation

If you use this dataset in your research, please cite it as:

@inproceedings{foucher2026usvloc,
  title={USVLoc: Infrared Topographic Based Cross-view Localization in Maritime Environment},
  author={Foucher, Alexandre and Seguin, Cédric and Heller, Dominique and Laurent, Johann},
  booktitle={Workshop on Perception, Interaction and Decision-Making for Human Cyber-Physical Systems, International Conference on Pattern Recognition},
  year={2026}
}

🤝 Acknowledgments

This work is carried out as part of a thesis at the Lab-STICC laboratory within South Brittany University (UBS). This work is financially supported by the Defense Innovation Agency (AID).

📃 License

This dataset is released under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). See LICENSE file for details.