PAIR360: A Paired Dataset of High-Resolution 360° Panoramic Images and LiDAR Scans

GeunU Kim, Daeho Kim, Jaeyun Jang, Hyoseok Hwang
{kyakl7232, kdh2769, yoon2926, hyoseok}@khu.ac.kr
Kyung Hee University AIRLAB
RA-L 2024
Paper Dataset

PAIR360 Dataset

  • For access to the dataset, please request the password by email.
  • This dataset provides 360 equirectangular image set with LiDAR point cloud from Insta360 pro2 - Velodyne VLP-32C system.
  • Moreover, We use pre-trained foundation model, Depth-Anything, to make depth map and semantic segmentation labels.
Overview

Data

  • 360 equirectangular images (7680 x 3840)
    • RGB
    • Depth
  • Fish eye images (3840 x 2880)
    • RGB
  • LiDAR point clouds (32 channels)
  • IMU and GPS
  • Panormic annotations
    • Semantic segmentation labels
    • Depth map
  • 3D point cloud map

Weather condition

  • Sunny
  • Cloudy
  • Sunrising

Area

  • We capture 7 areas for 3 times in the Kyung Hee University International Campus
    • College of Engineering
    • College of Life Science
    • College of Physical Education
    • Central Library
    • 1st Dormitory
    • 2nd Dormitory
    • Parking Lot & Peace Amphitheater

Data Directory

Sensor Setup & Area Layout.

MY ALT TEXT

The Insta360 pro2 has six lenses, and they store original fisheye video. we store extract fisheye images and saved each origin'#' folder. The z-axis is front of camera.

Sensor Setup & Data Area

MY ALT TEXT

Sequences

GPS Trajectories in traversal 1

GPS Trajectories in traversal 2

GPS Trajectories in traversal 3

Dataset

engineer sport
1stdorm 2nddorm

Trajectory Estimation

Tools

File Description
pcd2bag.py Convert pcd files, GPS.csv and GYRO.csv file to ROSbag
Lidar_fisheye_projection.py Project onto fisheye image
Lidar_pano_projection.py Project onto panorama image
GPS_mapping.py Map GPS data and save html file
segmentation_labels.txt Category and RGB value of segmentation labels

License

The PAIR360 dataset is provided under the Open Database License (available here). This means that you are free to use, distribute, modify and create derivative works from this dataset, provided that you give appropriate credit to our work. In addition, any adapted version of this dataset that is used publicly must be released under the same licence, and any redistribution must remain open. When using the PAIR360 dataset, please ensure that you acknowledge our work by citing the referenced paper.

BibTeX


        @article{kim2024pair360,
                  title={PAIR360: A Paired Dataset of High-Resolution 360$$\{$$\}$\^{}$\{$$\backslash$circ$\}$ $ Panoramic Images and LiDAR Scans},
                  author={Kim, Geunu and Kim, Daeho and Jang, Jaeyun and Hwang, Hyoseok},
                  journal={IEEE Robotics and Automation Letters},
                  year={2024},
                  publisher={IEEE}
                }