Glasgow 3D city models derived from airborne LiDAR point clouds licensed data

UBDC generates 3D city models via the airborne LiDAR point clouds acquired between 2020-2021 on behalf of Glasgow City Council. We prepared a set of training and validation data to classify the whole LiDAR dataset for subsequent 3D model construction.

The annotated point clouds were generated to train the weakly supervised semantic segmentation algorithm Semantic Query Network (SQN) to classify point clouds [1]-[2]. Four tiles of the 1×1 km2 sparse point clouds were annotated for training and four tiles of 0.5×0.5 km2 full point clouds were annotated for validation. Annotated data contains historical and modern architectures as well as residential and industrial buildings. The point clouds were manually labeled into four categories: ground, trees (including arbors and shrubs but excluding lawn), buildings, and others. The annotated point cloud data can be used to train a deep learning model for point cloud classification or help advance the manipulation within airborne LiDAR.

References: * [1] Hu, Q., Yang, B., Fang, G., Guo, Y., Leonardis, A., Trigoni, N., & Markham, A. (2022, October). Sqn: Weakly-supervised semantic segmentation of large-scale 3d point clouds. In European Conference on Computer Vision (pp. 600-619). Cham: Springer Nature Switzerland. * [2] Li, Q., & Zhao, Q. (2023, May). Weakly-Supervised Semantic Segmentation of Airborne LiDAR Point Clouds in Hong Kong Urban Areas. In 2023 Joint Urban Remote Sensing Event (JURSE) (pp. 1-4). IEEE.

LiDAR data were acquired from 2020-2021