language:
- en
license: cc-by-4.0
size_categories:
- 1M<n<10M
task_categories:
- robotics
tags:
- space
- exploration
- lunar
- navigation
- science
- spad
arxiv: 2506.22956
SPICE-HL3: Single-Photon, Inertial, and Stereo Camera dataset for Exploration of High-Latitude Lunar Landscapes
Paper | Code | Data (Zenodo)
This dataset presents data recorded at the LunaLab from the Interdisciplinary Research Center for Security reliability and Trust (SnT) at the University of Luxembourg, an indoor test facility designed to replicate the optical characteristics of multiple lunar latitudes. The data recorded is specifically intended for engineers and researches working on vision-based robotic navigation (and associated tools) across high-latitude lunar regions; although it could also serve those working on computer vision applications at the edges of the spectrum of perceptually degraded scenarios. It includes images, inertial, and wheel odometry data from robots navigating seven distinct trajectories under multiple high-latitude illumination scenarios, simulating conditions from lunar dawn to night time with and without the aid of headlights. Data was captured using Leo rovers incorporating a ZED2 stereo-inertial sensor, a FLIR Blackfly-S monocular monochrome camera, and a novel single-photon avalanche diode (SPAD) camera. We recorded both static and dynamic image sequences, with robots navigating at slow (5 cm/s) and fast (50 cm/s) speeds. All data is calibrated, synchronized, and timestamped, providing a valuable resource for validating vision-based autonomous navigation pipelines for future lunar missions targeting high-latitude regions.
Citation
Rodríguez-Martínez, D., van der Meer, D., Song, J., Bera, A., Pérez del Pulgar, C., & Olivares-Mendez, M. A. (2026). SPICE-HL3: Single-Photon, Inertial, and Stereo Camera dataset for Exploration of High-Latitude Lunar Landscapes. Scientific Data. https://doi.org/10.1038/s41597-026-06668-8
@article{rodriguez2025spicehl3,
title={{SPICE}-{HL3}: Single-Photon, Inertial, and Stereo Camera dataset for Exploration of High-Latitude Lunar Landscapes},
author = {Rodríguez-Martínez, David and van der Meer, Dave and Song, Junlin and Bera, Abishek and Pérez del Pulgar, C.J. and Olivares-Mendez, Miguel Angel},
journal={Scientific Data},
volume={},
number={},
pages={},
doi = {10.1038/s41597-026-06668-8},
url = {https://www.nature.com/articles/s41597-026-06668-8},
year = {2026}
}
Leaderboard
This is a public leaderboard showcasing the performance of various Visual Odometry and SLAM methods evaluated on different trajectories from the SPICE-HL3 dataset.
Fast Sequences SPICE-HL3 Leaderboard
| Rank | Method | Sensor | Trj_A | Trj_B | Trj_C | Trj_D | Trj_E | Trj_F | Trj_G | avATE RMSE [cm] | avATE Max [cm] |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 :fire: | Wheel Odometry | WODO | :heavy_minus_sign: | 34.66 (63.91) | 124.87 (213.86) | 85.40 (123.33) | :x: | :heavy_minus_sign: | :heavy_minus_sign: | 140.70 | 442.36 |
| 2 | RTAB-Map | Stereo | :heavy_minus_sign: | 64.77 (97.44) | :x: | :x: | 29.12 (35.29) | 64.50 (95.65) | 145.56 (202.12) | 253.50 | 965.61 |
| 3 | ORB-SLAM3 | Stereo | :heavy_minus_sign: | 83.72 (135.86) | 83.23 (135.47) | :x: | 17.31 (31.72) | :x: | :x: | 841.15 | 4161.01 |
| 4 | ORB-SLAM3 | Mono | :heavy_minus_sign: | 100.27 (168.82) | 94.04 (159.92) | :x: | :x: | :x: | :x: | 859.51 | 4247.57 |
| 5 | Inertial Odometry (naive) | IMU | :heavy_minus_sign: | 473.97 (1183.15) | 928.11 (2341.06) | 816.84 (2384.15) | :x: | :x: | :x: | 938.21 | 4265.62 |
:heavy_minus_sign: : Denotes trajectories that have not been evaluated.
:x: : Describe trajectories in which localization is lost and unrecovered; the method fails to provide a final estimate.