The Institut Cartogràfic i Geològic de Catalunya (ICGC), as a premier mapping agency, undertakes an annual aerial imagery survey of Cataluña to monitor urban changes and update geospatial basemaps. This project focused on automating the detection of new building footprints between 2018 and 2022 using deep learning techniques, aiming to overcome the traditionally labour-intensive process of change detection and processing large datasets.

ICGC leveraged the pretrained Building Footprint Extraction model from ArcGIS Living Atlas for high-quality, out-of-the-box results. Using aerial imagery of 25 cm resolution, the workflow included data preparation via mosaic datasets, creation of map tile packages for efficient data sharing, and application of inference using deep learning. Multi-resolution analysis was conducted to optimize results, adjusting cell sizes to capture buildings of varying sizes effectively. This end-to-end workflow not only delivered accurate building footprints but also offered a detailed view of urban growth patterns, supporting urban planning and decision-making processes.