CITYLID: A large-scale categorized aerial lidar dataset for street-level research

Urban point cloud datasets are crucial for understanding the depth and physical structure of environmental features. These details hold significance in urban planning, providing precise measurements of the space upon which novel development plans and strategies can be formulated. However, such datasets, when uncategorized, lack information, rendering them much less helpful in utilizing them in urban planning and design projects. This documentation provides a methodical framework to create the CITYLID dataset, which uses an openly available citywide aerial Lidar dataset, categorizes it with standard urban classes such as buildings, trees, and ground, and fuses with it detailed street features information such as driveways, medians, bikepaths, walkways, and on-street parking. Since the point cloud provides the required height information, shadow maps are also generated utilizing the entire point cloud dataset and further integrated with the point clouds. The resulting dataset includes 3 standard and 5 street feature classes, along with 5 classes representing shadows. Apart from the categorized point cloud dataset, we additionally provide the detailed methodology to generate Lidar categorization and starter codes to extract subsets of point clouds, which can be used to analyze and study urban environments such as street cross-sections, neighborhood comparisons, tree inventory, etc.