POSS Odometry: A Visual and LiDAR Odometry Dataset in Campus Environment


    We present the POSS odometry dataset, which contains 36924 frames captured in the campus of Peking University. Each frame contains LiDAR data, grayscale monocular camera data, colorful monocular data and ground truth pose information. LiDAR data are captured with Velodyne 32e, and each frame of the LiDAR data is saved in a seperate bin file. Camera data are saved in loss less png format. With this dataset, you may test monocular visual odometry algorithms, laser-based SLAM algorithms or algorithms that combine visual and LIDAR information. This dataset is used for Localization and Mapping task. And for convenience, the provieded dataset uses the same data format and API as KITTI Odometry.

    The following figure demonstrates the ground truth trajectory of the dataset on the Google map.


    2023.03.27: The calibration file has been updated with an associated readme.

Data format

    The folder structure and data format are same as the KITTI odometry dataset. File XXXXXX.bin in velodyne folder is the point cloud data. File XXXXXX.png in image_0 folder is the grayscale image data. File XXXXXX.png in the image_2 folder is the colorful image data. poses.txt contains the ground truth trajectory information. times.txt contains the time of the each frame from the first frame. calib.txt contains the calibraiton information which is in the same format as KITTI.



    title={Learning Scene Adaptive Covariance Error Model of LiDAR Scan Matching for Fusion Based Localization},
    author={X. Ju and D. Xu and H. Zhao},
    journal={IEEE Transactions on Intelligent Transportation Systems},
    title={Scene-Aware Error Modeling of LiDAR/Visual Odometry for Fusion-Based Vehicle Localization},


    This dataset follow Creative Commons Attribution-NonCommercial-ShareAlike 3.0 License.