[scalemvo] Author-collected SNU underground parking lots datasets

The datasets are used in our submission to scalemvo paper.


We collect our own driving datasets, called SNU underground parking lots datasets. Different from the KITTI outdoor datasets, due to the absence of the external ground truth measurement such as the GPS/INS solution, we additionally record the 3-D LiDAR pointcloud, and we execute the LiDAR odometry and mapping (LOAM) on our datasets to obtain the accurate metric trajectory.

We install three global shutter grayscale cameras, a 32-channel 3-D LiDAR, and a 6-axis IMU on the roof of the vehicle. All cameras are triggered to capture time-synchronized 10 Hz images by the digital signal from the Arduino MKR Zero microcontroller. All the sensors and the microcontroller communicate to the Linux laptop computer by the ethernet interface.

Our camera setting has L=1.45 m and the height of the cameras is H=1.55 m from the ground. Two main cameras face front, and an auxiliary camera is rotated left to 20 degrees.Intrinsic and extrinsic parameters of cameras, 3-D LiDAR and IMU are calibrated by using the LiDAR and camera extrinsic calibration module.

Vehicle and Sensor Rigs

We have various sensors on the vehicle: three monocular cameras, a 3-D LiDAR, and an inertial measurement unit (IMU). All images are accurately time-synchronized by a hardware trigger generated by the microcontroller unit (Arduino MKR Zero). Cameras, LiDAR, and microcontroller communicate to the Linux laptop by the ethernet interface.


  • 1 x Hyundai Elantra CN7 2021
  • length: 4.68 m, width: 1.82 m, height: 1.41 m, wheel base: 2.72 m


  • 3 x Matrixvision mvBlueCOUGAR-X104iG
  • Resolution: 1032 X 772 (8-bits monochrome)
  • Frequency: 10 Hz (hardware synchronized by Arduino MRK Zero)


  • 1 x Velodyne VLP-32C
  • 32-channel 360 degrees laser scans at 10 Hz, 20 degrees vertical field of view


  • 1 x Lord Microstrain 3DM-GX3-25 AHRS
  • 3-axis acceleration, 3-axis gyroscope at 250 Hz


  • 1 x Arduino MKR Zero with the Ethernet Shield

How to use

After reading details below, you can download each dataset at the Download List section below. The datasets are in the ROSBAG format.

Download List

| Sequence Name | Duration [s] | Length of the trajectory [m]| | bldg_39
ROSBAG(5.48 GB)| 137.5 | 332 | | bldg_220
ROSBAG(11.51 GB)| 272.1 | 687 |

  • For questions, comments or suggestions, please feel free to send me an e-mail.