NCLT Dataset
The University of Michigan North Campus Long-Term Vision and LIDAR Dataset.
Nicholas Carlevaris-Bianco
carlevar@umich.edu
Arash Ushani
aushani@umich.edu
Ryan M. Eustice
eustice@umich.edu
Overview
- Sensors
- Velodyne HDL-32E LIDAR
- Ladybug3 Omnidirectional Vision
- Microstrain GX3 IMU
- KVH Fiber Optic Gyro
- Wheel Odometry
- Standard and RTK GPS
- Data Collection
- 27 Sessions: Varying route and time of day
- 15 Months: Jan. 2012 to April 2013
- Session Duration: ≈ 1.29 hours
- Session Distance: ≈ 5.5 km
- Total Duration: 34.9 hours
- Total Distance: 147.4 km
Documentation
Download the PDF (nclt.pdf), which provides a detailed description of the dataset. If you use this dataset in your research please cite:
Nicholas Carlevaris-Bianco, Arash K. Ushani, and Ryan M. Eustice, University of Michigan North Campus Long-Term Vision and Lidar Dataset, International Journal of Robotics Research, 2016. [BibTeX]
Changelog
Feburary 4, 2017: The covariances for keyframes in the ground truth pose estimate has been added. These marginal covariances are extracted from the SLAM graph and recorded in the same format as other covariances in the dataset.
January 16, 2017: In Table 6, the x and y columns for the camera center coordinates is swapped. The correct values are given in cam_params.zip. Additionally, a python script demonstrating how to use the camera parameter files and project velodyne points to images has been added.
August 14, 2018: Sensor data from the KVH fiber-optic gyro and the left and right wheel velocities have been added to the sen.tar.gz files.
March 24, 2019: The ground truth pose was generated using a SLAM graph, and then interpolation using odometry was used to provide this at roughly 100 Hz. A script showing how to inspect only the ground truth pose from the nodes in the SLAM graph has been added. Thanks to Alexander Schaefer for helping us discover the ambiguity regarding this subtlety.
License
The NCLT Dataset is made available under the Open Database License (available here). Any rights in individual contents of the database are licensed under the Database Contents License (available here).
In short, this means that you are free to use this dataset, share, create derivative works, or adapt it, as long as you credit our work, offer any publically used adapted version of this dataset under the same license, and keep any redistribution of this dataset open.
When using the NCLT Dataset, please credit our work by citing the paper above.
Download
Data
| Session | Images | Sensors | Velodyne | Hokuyo | Ground Truth Pose | Ground TruthCovariance |
|---|---|---|---|---|---|---|
| 2012-01-08 | lb3.tar.gz (106 GB) | sen.tar.gz (114 MB) | vel.tar.gz (16 GB) | hokuyo.tar.gz (137 MB) | groundtruth.csv (111 MB) | cov.csv (448 KB) |
| 2012-01-15 | lb3.tar.gz (130 GB) | sen.tar.gz (136 MB) | vel.tar.gz (20 GB) | hokuyo.tar.gz (157 MB) | groundtruth.csv (132 MB) | cov.csv (522 KB) |
| 2012-01-22 | lb3.tar.gz (92 GB) | sen.tar.gz (107 MB) | vel.tar.gz (15 GB) | hokuyo.tar.gz (126 MB) | groundtruth.csv (103 MB) | cov.csv (421 KB) |
| 2012-02-02 | lb3.tar.gz (105 GB) | sen.tar.gz (119 MB) | vel.tar.gz (16 GB) | hokuyo.tar.gz (148 MB) | groundtruth.csv (115 MB) | cov.csv (427 KB) |
| 2012-02-04 | lb3.tar.gz (85 GB) | sen.tar.gz (96 MB) | vel.tar.gz (13 GB) | hokuyo.tar.gz (118 GB) | groundtruth.csv (92 MB) | cov.csv (383 KB) |
| 2012-02-05 | lb3.tar.gz (109 GB) | sen.tar.gz (114 MB) | vel.tar.gz (16 GB) | hokuyo.tar.gz (136 MB) | groundtruth.csv (110 MB) | cov.csv (451 KB) |
| 2012-02-12 | lb3.tar.gz (101 GB) | sen.tar.gz (105 MB) | vel.tar.gz (15 GB) | hokuyo.tar.gz (113 MB) | groundtruth.csv (101 MB) | cov.csv (396 KB) |
| 2012-02-18 | lb3.tar.gz (102 GB) | sen.tar.gz (108 MB) | vel.tar.gz (15 GB) | hokuyo.tar.gz (135 MB) | groundtruth.csv (105 MB) | cov.csv (425 KB) |
| 2012-02-19 | lb3.tar.gz (101 GB) | sen.tar.gz (109 MB) | vel.tar.gz (15 GB) | hokuyo.tar.gz (126 MB) | groundtruth.csv (105 MB) | cov.csv (422 KB) |
| 2012-03-17 | lb3.tar.gz (100 GB) | sen.tar.gz (101 MB) | vel.tar.gz (13 GB) | hokuyo.tar.gz (118 MB) | groundtruth.csv (98 MB) | cov.csv (401 KB) |
| 2012-03-25 | lb3.tar.gz (107 GB) | sen.tar.gz (98 MB) | vel.tar.gz (13 GB) | hokuyo.tar.gz (90 MB) | groundtruth.csv (95 MB) | cov.csv (401 KB) |
| 2012-03-31 | lb3.tar.gz (95 GB) | sen.tar.gz (107 MB) | vel.tar.gz (95 GB) | hokuyo.tar.gz (132 MB) | groundtruth.csv (103 MB) | cov.csv (413 KB) |
| 2012-04-29 | lb3.tar.gz (87 GB) | sen.tar.gz (53 MB) | vel.tar.gz (8 GB) | hokuyo.tar.gz (57 MB) | groundtruth.csv (51 MB) | cov.csv (343 KB) |
| 2012-05-11 | lb3.tar.gz (105 GB) | sen.tar.gz (104 MB) | vel.tar.gz (14 GB) | hokuyo.tar.gz (95 MB) | groundtruth.csv (100 MB) | cov.csv (418 KB) |
| 2012-05-26 | lb3.tar.gz (87 GB) | sen.tar.gz (108 MB) | vel.tar.gz (15 GB) | hokuyo.tar.gz (130 MB) | groundtruth.csv (104 MB) | cov.csv (433 KB) |
| 2012-06-15 | lb3.tar.gz (94 GB) | sen.tar.gz (67 MB) | vel.tar.gz (10 GB) | hokuyo.tar.gz (81 MB) | groundtruth.csv (64 MB) | cov.csv (396 KB) |
| 2012-08-04 | lb3.tar.gz (90 GB) | sen.tar.gz (97 MB) | vel.tar.gz (14 GB) | hokuyo.tar.gz (133 MB) | groundtruth.csv (94 MB) | cov.csv (380 KB) |
| 2012-08-20 | lb3.tar.gz (89 GB) | sen.tar.gz (101 MB) | vel.tar.gz (15 GB) | hokuyo.tar.gz (129 MB) | groundtruth.csv (94 MB) | cov.csv (416 KB) |
| 2012-09-28 | lb3.tar.gz (84 GB) | sen.tar.gz (94 MB) | vel.tar.gz (13 GB) | hokuyo.tar.gz (126 MB) | groundtruth.csv (90 MB) | cov.csv (387 KB) |
| 2012-10-28 | lb3.tar.gz (98 GB) | sen.tar.gz (104 MB) | vel.tar.gz (14 GB) | hokuyo.tar.gz (138 MB) | groundtruth.csv (100 MB) | cov.csv (394 KB) |
| 2012-11-04 | lb3.tar.gz (88 GB) | sen.tar.gz (97 MB) | vel.tar.gz (13 GB) | hokuyo.tar.gz (128 MB) | groundtruth.csv (93 MB) | cov.csv (335 KB) |
| 2012-11-16 | lb3.tar.gz (89 GB) | sen.tar.gz (98 MB) | vel.tar.gz (13 GB) | hokuyo.tar.gz (137 MB) | groundtruth.csv (95 MB) | cov.csv (337 KB) |
| 2012-11-17 | lb3.tar.gz (107 GB) | sen.tar.gz (107 MB) | vel.tar.gz (14 GB) | hokuyo.tar.gz (133 MB) | groundtruth.csv (103 MB) | cov.csv (400 KB) |
| 2012-12-01 | lb3.tar.gz (94 GB) | sen.tar.gz (93 MB) | vel.tar.gz (12 GB) | hokuyo.tar.gz (126 MB) | groundtruth.csv (89 MB) | cov.csv (348 KB) |
| 2013-01-10 | lb3.tar.gz (37 GB) | sen.tar.gz (21 MB) | vel.tar.gz (3 GB) | hokuyo.tar.gz (26 MB) | groundtruth.csv (20 MB) | cov.csv (80 KB) |
| 2013-02-23 | lb3.tar.gz (88 GB) | sen.tar.gz (97 MB) | vel.tar.gz (15 GB) | hokuyo.tar.gz (120 MB) | groundtruth.csv (93 MB) | cov.csv (368 KB) |
| 2013-04-05 | lb3.tar.gz (88 GB) | sen.tar.gz (84 MB) | vel.tar.gz (13 GB) | hokuyo.tar.gz (92 MB) | groundtruth.csv (81 MB) | cov.csv (317 KB) |
Python Downloader
For systems with python and wget, we provide a downloader script (downloader.py) that can be used to download different subsets of the data. In this script wget calls are made with the --continue flag, which allows you to restart a download if the connection is interrupted. Example usage:
Download the entire dataset
./downloader.py --all
Download all data from 2012-01-08
./downloader.py --date="2012-01-08" --all
Download all sensor data
./downloader.py --sen
Download imagery from 2012-01-08
./downloader.py --date="2012-01-08" --lb3
Download imagery and Velodyne data from 2012-01-08
./downloader.py --date="2012-01-08" --lb3 --vel
LadyBug3 Parameters and Undistort Maps
| Camera Parameters | cam_params.zip (2 KB) | |
| All Maps | U2D_ALL_1616X1232.tar.gz (151 MB) | D2U_ALL_1616X1232.tar.gz (148 MB) |
| Cam 0 | U2D_Cam0_1616X1232.txt (61 MB) | D2U_Cam0_1616X1232.txt (61 MB) |
| Cam 1 | U2D_Cam1_1616X1232.txt (61 MB) | D2U_Cam1_1616X1232.txt (61 MB) |
| Cam 2 | U2D_Cam2_1616X1232.txt (61 MB) | D2U_Cam2_1616X1232.txt (61 MB) |
| Cam 3 | U2D_Cam3_1616X1232.txt (61 MB) | D2U_Cam3_1616X1232.txt (61 MB) |
| Cam 4 | U2D_Cam4_1616X1232.txt (61 MB) | D2U_Cam4_1616X1232.txt (61 MB) |
| Cam 5 | U2D_Cam5_1616X1232.txt (61 MB) | D2U_Cam5_1616X1232.txt (61 MB) |
Sample Python Scripts
| File | Description |
|---|---|
| python.tar.gz | All sample python scripts |
| read_gps.py | Read GPS from data file |
| read_ground_truth.py | Read ground truth from data file |
| read_hokuyo_30m.py | Read Hokuyo UTM-30LX from data file |
| read_hokuyo_4m.py | Read Hokuyo URG-04LX from data file |
| read_ms25.py | Read Microstrain data from data file |
| read_ms25_euler.py | Read Microstrain raw Euler angles from data file |
| read_odom.py | Read odometry from data file |
| read_vel_hits.py | Read Velodyne data from data file |
| read_vel_sync.py | Read one revolution of Velodyne from data file |
| project_vel_to_cam.py | Project velodyne points to camera image |
| Undistort.py | Undistort LadyBug3 images |
| gps_to_rosbag.py | Convert GPS file to ROSbag |
| hokuyo_30m_to_rosbag.py | Convert Hokuyo UTM-30LX file to ROSbag |
| hokuyo_4m_to_rosbag.py | Convert Hokuyo URG-04LX file to ROSbag |
| ms25_euler_to_rosbag.py | Convert Microstrain Euler file to ROSbag |
| ms25_to_rosbag.py | Convert Microstrain file to ROSbag |
| vel_to_rosbag.py | Convert Velodyne file to ROSbag |
| sensordata_to_rosbag.py | Convert all sensor data to ROSbag |
Other Files
| File | Description |
|---|---|
| segway_cad.tar.gz | Solidworks CAD Model of the Segway |
| laser_angles.csv | Laser angles for the Velodyne |
| DSP3000_Manual.pdf | KVH DSP-3000 Fiber Optic Gyro Technical Manual |
| URG-04LX_spec_en.pdf | Scanning Laser Range Finder URG-04LX Specification |
| UTM-30LX_spec_en.pdf | Scanning Laser Range Finder UTM-30LX/LN Specification |
| GPS_18x_Tech_Specs.pdf | GPS 18x Technical Specifications |
| 3DM-GX3-45-Theory_of_Operation.pdf | 3DM-GX3-45 Theory of Operation |
| Ladybug3-Technical-Reference.pdf | Ladybug3 Technical Reference Manual |
| HDL-32E_manual.pdf | User's Manual and Programming Guide: HDL-32E High Definition LiDAR Sensor |
| dl4_manual.pdf | DL-4plus User Manual |