Mask region-based convolutional neural network (R-CNN) models for extracting building footprints from remote sensing data

These trained deep learning are trained to extract building footprints from high-resolution aerial imagery and LiDAR-derived nDSM with a spatial resolution of 20cm or less. These models were used to extract footprints in formal residential zones, industrial zones, and informal settlement zones within the City of Cape Town. It must be noted that the trained Mask R-CNN models are scalable to extract building footprints across different South African’ Metropolitans, as formal and informal zones co-exist in these areas and they have similar environmental settings.

本数据集采用的经训练深度学习模型，可从空间分辨率不超过20cm的高分辨率航空影像与激光雷达（LiDAR）派生的归一化数字表面模型（nDSM）中提取建筑基底轮廓（building footprint）。上述模型已被应用于提取开普敦市内正规住宅区、工业区以及非正规定居区的建筑基底轮廓。需要说明的是，经训练的掩码区域卷积神经网络（Mask R-CNN）模型具备可扩展性，可用于提取南非不同大都市区内的建筑基底轮廓；此类区域同时并存正规与非正规区域，且环境条件相似。