NTU VIRAL: A Visual-Inertial-Ranging-Lidar Dataset for Autonomous Aerial Vehicle

📦 Latest complete dataset: Download NTU VIRAL DATASET, including all 18 sequences, calibration data and ground truth, from Hugging Face.

This site presents the datasets collected from our research Unmanned Aerial Vehicle (UAV) platform, featuring an extensive set of sensors:

• Two 3D lidars
• Two time-synchronized cameras
• Multiple Inertial Measurement Units (IMUs)
• Four Ultra-wideband (UWB) nodes on UAV, ranging to three anchor nodes.

The comprehensive sensor suite resembles that of an autonomous driving car, but features distinct and challenging characteristics of aerial operations. The flight tests are conducted in a variety of both indoor and outdoor conditions.

Citation

If you use some resource from this data suite, please cite it as

@article{nguyen2022ntu,
  title     = {NTU VIRAL: A Visual-Inertial-Ranging-Lidar Dataset, From an Aerial Vehicle Viewpoint},
  author    = {Nguyen, Thien-Minh and Yuan, Shenghai and Cao, Muqing and Lyu, Yang and Nguyen, Thien Hoang and Xie, Lihua},
  journal   = {The International Journal of Robotics Research},
  volume    = {41},
  number    = {3},
  pages     = {270--280},
  year      = {2022},
  publisher = {SAGE Publications Sage UK: London, England}
}

[Journal][Preprint]

Updates

05/02/2023: Update works using the datasets

26/09/2022: Extra sequences (rtp_01, rtp_02, rtp_03, tnp_01, tnp_02, tnp_03, spms_01, spms_02, spms_03) in more challenging scenarios are added.

12/07/2022: The ouster pointcloud and IMU messages are found to jitter due to synchronization issues. A script to regularize the ouster pointcloud and imu topics can be downloaded here.

Downloads

The primary download links below point to NTU VIRAL DATASET on Hugging Face. The original source remains NTU Data Repository. An additional original-source mirror is available in this OneDrive folder.

Groundtruth is included in the bag. Or you can download them in csv formats here: ground-truth CSVs on Hugging Face.

We note that many users forget to check for the 0.4m offset from the IMU to the prism, which is where ground truth is measured. To help evaluating things more easily, we have written a jupyter and MATLAB scripts to help with the evalution. Please check out the tutorial for their use.

Quick use

💻 Baseline code

Choose a baseline for your sensor setup. This catalogue brings together 63 methods and implementations and 14 tools or reference resources for NTU VIRAL DATASET, including the original 10 dataset-adapted baselines and additional public implementations.

📖 Baseline guide · 📚 Papers and BibTeX · 🔎 Code and configuration audit

The guide covers installation, preprocessing, sensor settings, evaluation and licenses. NTU setup links open the checked configuration, launch file or usage guide. Original credits the upstream project for dataset-adapted forks; Software notes identifies resources without a separate paper linked in the catalogue.

For evaluation: use the stated sensor topics and calibration, account for the prism offset, and report development sequences and loop-closure settings. See the linked papers for experimental results.

Visual and visual–inertial methods

LiDAR odometry

LiDAR–inertial methods

Multiple-LiDAR methods

LiDAR–visual–inertial methods

Mapping and localization

Evaluation, calibration and data tools

Additional ports and variants

These entries extend methods listed above and are grouped with their parent method.

• FAST-LIVO — Experimental time-offset variant: Code · NTU setup. Software notes · BibTeX.
• FAST-LIVO2 — ROS 2 port: Code · NTU setup. Software notes · BibTeX.
• FAST-LIVO2 — SC_FAST_LIVO2 front end; external PGO required: Code · NTU setup. Software notes · BibTeX.
• Traj-LO — AdaTrajLo fork retaining NTU config: Code · NTU setup. Software notes · BibTeX.
• Coco-LIC — LICO-mid360 fork retaining NTU config: Code · NTU setup. Software notes · BibTeX.
• ORB-SLAM3 — CARES stereo-inertial adapter: Code · NTU setup. Software notes · BibTeX.
• GenZ-ICP — MapsHD ROS 2 guide: Code · NTU setup. Software notes · BibTeX.
• FAST-LIO2 — Surfel map / Eigen implementation (ROS 1 and ROS 2): Code · NTU setup. Software notes · BibTeX.
• FAST-LIO2 — FLiVIS community variant: Code · NTU setup. Software notes · BibTeX.

Last catalogue check: 26 October 2026.

Related works

The datasets were used in the following papers. Please checkout these works if you are interested. (Please contact us if you would like your work mentioned here).

• FAST-LIVO: Fast and Tightly-coupled Sparse-Direct LiDAR-Inertial-Visual Odometry
• VIRAL SLAM: Tightly Coupled Camera-IMU-UWB-Lidar SLAM
• MILIOM: Tightly Coupled Multi-Input Lidar-Inertia Odometry and Mapping (RAL 2021)
• LIRO: Tightly Coupled Lidar-Inertia-Ranging Odometry (ICRA 2021)

Notes:

For more information on the sensors and how to use the dataset, please checkout the other sections.

For resources and other works of our group please checkout our github.

If you have some inquiry, please raise an issue on github.

Licence

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License and is intended for non-commercial academic use. If you are interested in using the dataset for commercial purposes please contact us.