{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/aerial-imagery-for-roof-segmentation-a-large","title":"Aerial Imagery for Roof Segmentation: A Large-Scale Dataset towards Automatic Mapping of Buildings","arxiv_id":"1807.09532","date":"2018-07-25","proceeding":null,"authors":["Qi Chen","Lei Wang","Yifan Wu","Guangming Wu","Zhiling Guo","Steven L. Waslander"],"abstract":"As an important branch of deep learning, convolutional neural network has\nlargely improved the performance of building detection. For further\naccelerating the development of building detection toward automatic mapping, a\nbenchmark dataset bears significance in fair comparisons. However, several\nproblems still remain in the current public datasets that address this task.\nFirst, although building detection is generally considered equivalent to\nextracting roof outlines, most datasets directly provide building footprints as\nground truths for testing and evaluation; the challenges of these benchmarks\nare more complicated than roof segmentation, as relief displacement leads to\nvarying degrees of misalignment between roof outlines and footprints. On the\nother hand, an image dataset should feature a large quantity and high spatial\nresolution to effectively train a high-performance deep learning model for\naccurate mapping of buildings. Unfortunately, the remote sensing community\nstill lacks proper benchmark datasets that can simultaneously satisfy these\nrequirements. In this paper, we present a new large-scale benchmark dataset\ntermed Aerial Imagery for Roof Segmentation (AIRS). This dataset provides a\nwide coverage of aerial imagery with 7.5 cm resolution and contains over\n220,000 buildings. The task posed for AIRS is defined as roof segmentation. We\nimplement several state-of-the-art deep learning methods of semantic\nsegmentation for performance evaluation and analysis of the proposed dataset.\nThe results can serve as the baseline for future work.","url_abs":"http://arxiv.org/abs/1807.09532v2","url_pdf":"http://arxiv.org/pdf/1807.09532v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[{"slug":"airs","name":"AIRS","full_name":""}],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}