The New Zealand Point Cloud Classification Deep Learning Package is designed to classify aerial LiDAR point clouds into two categories: Building and Background. Optimized specifically for New Zealand’s terrain and architectural styles, this model delivers high-accuracy results when applied to LiDAR data collected across both urban and rural areas.

Accurate classification of buildings within point cloud datasets is essential for applications such as 3D basemap generation, urban development, and climate change planning. Traditional classification methods often struggle with the complex and irregular geometries of buildings. In contrast, this deep learning model excels at learning and identifying such structures, producing more reliable outputs.

The model was trained, tested, and validated entirely on LiDAR data from New Zealand, ensuring strong generalization to the region’s typical building forms and increasing its effectiveness for local use cases.