license: cc-by-nc-sa-4.0
task_categories:
- robotics
- image-to-3d
tags:
- slam
- lidar
- 3d-reconstruction
- nerf
- 3d-gaussian-splatting
- localization
- sfm
- mvs
- multimodal
- oxford
We present the Oxford Spires Dataset, captured in and around well-known landmarks in Oxford using a custom-built multi-sensor perception unit as well as a millimetre-accurate map from a terrestrial LiDAR scanner (TLS). The perception unit includes three global shutter colour cameras, an automotive 3D LiDAR scanner, and an inertial sensor — all precisely calibrated.
Sample Usage
Download the Dataset
You can download the dataset from Hugging Face using the provided script. You can specify which folders to download by changing the example_pattern. Core sequences are also defined in the script.
python scripts/dataset_download.py
Install Python Tools
Install oxspires_tools to access Python utilities for using the dataset:
pip install .
To enable C++/Python bindings (requires PCL and Octomap):
BUILD_CPP=1 pip install .
Alternatively, use the provided Docker container:
docker compose -f .docker/oxspires/docker-compose.yml run --build oxspires_utils
Generate Depth Images
The following script downloads synchronised images and LiDAR data from a sequence on Hugging Face and generates depth images, LiDAR overlaid on camera images, and surface normal images:
python scripts/generate_depth.py
Citation
If you use The Oxford Spires Dataset in your research, please cite the following paper:
@article{tao2025spires,
title={The Oxford Spires Dataset: Benchmarking Large-Scale LiDAR-Visual Localisation, Reconstruction and Radiance Field Methods},
author={Tao, Yifu and Mu{\~n}oz-Ba{\~n}{\'o}n, Miguel {\'A}ngel and Zhang, Lintong and Wang, Jiahao and Fu, Lanke Frank Tarimo and Fallon, Maurice},
journal={International Journal of Robotics Research},
year={2025},
}