LIBRE:
LiDAR Benchmark Reference dataset
We proudly present the Nagoya University and TierIV multiple 3D LiDARs dataset "LIBRE": LiDAR Benchmarking and Reference, a first-of-its-kind dataset featuring several different 3D LiDAR sensors, covering a range of manufacturers, models, and laser configurations.
Our dataset include LiDAR data from different environments and configurations. Static targets, where objects were placed at known distances and measured from a fixed position within a controlled environment. Adverse weather, where static obstacles were measured from a moving vehicle, captured in a weather chamber where LiDARs were exposed to different conditions (fog, rain, strong light). Finally, dynamic traffic, where dynamic objects were captured from a vehicle driven on public urban roads, multiple times at different times of the day; supporting sensors such as cameras, infrared imaging, and odometry devices were also included.
LIBRE will contribute the research community to (1) provide a means for a fair comparison of currently available LiDARs, and (2) facilitate the improvement of existing self-driving vehicles and robotics-related software, in terms of development and tuning of LiDAR-based perception algorithms.
12 LiDARs
All sensors tested in this study were off-the-shelf production models with the exception of the Velodyne VLS-128. This sensor was a pre-production model, and was tested to provide a preview for the production 128-line Alpha Prime sensor which was unavailable at the time the experiments were carried out. The dataset will be extended with the Alpha Prime results when testing has been completed.
All these sensors correspond to the multi-beam (multi-channel) mechanical scanning type: several pairs of laser diodes and photo-detectors (avalanche photo detector (APD) and single-photon avalanche diode (SPAD)) and corresponding emit-remit optics and mirrors, are rotated by a motor for 360 degrees which defines azimuth, while the vertical angle of a laser and photo-detector pair defines elevation. All sensors in this selection have short-wave infrared (SWIR) wavelengths between 850nm, 903nm and 905nm. While some support multiple returns (echoes), the data collected in our dataset always records only the strongest echo.
Static targets (reflectors, car, mannequin)
Static targets
Point Cloud
Range RSME on frontal (x) axis distance per LiDAR
Expected density obtained from simulation, using each LiDAR's HRes, VRes and VFOV, to find the number of points falling inside the reflector targets at each range.
Actual (measured) density detected on the reflective targets, averaged over 40 frames.
"Bullying" LiDARs under adverse weather conditions
Dense fog
Heavy rain
Strong "sun" light
Dynamic traffic scenes around Nagoya University
Hesai Pandar64
Velodyne VLS128
Dataset general structure
Publications
Don't miss our recent publications derived from LIBRE:
@article{acarballo2020libre,
title={{LIBRE}: The Multiple 3D LiDAR Dataset},
author={Alexander Carballo and Jacob Lambert and Abraham Monrroy and David Wong and Patiphon Narksri and Yuki Kitsukawa and Eijiro Takeuchi and Shinpei Kato and Kazuya Takeda},
journal={arXiv preprint arXiv:2003.06129},
year={2020},
note="(accepted for presentation at IV2020)"
}
@article{acarballo2020characterization,
title={Characterization of Multiple 3D LiDARs for Localization and Mapping using Normal Distributions Transform},
author={Alexander Carballo, Abraham Monrroy, David Wong, Patiphon Narksri, Jacob Lambert, Yuki Kitsukawa, Eijiro Takeuchi, Shinpei Kato, Kazuya Takeda},
journal={arXiv preprint arXiv:2004.01374},
year={2020}
}
For inquires, please contact:
libre-dataset-inquires [at] g [dot] sp [dot] m [dot] is [dot] nagoya-u [dot] ac [dot] jp