|
The estimation of building height using optical and synthetic aperture radar (SAR) remote sensing imagery is of critical importance for advancing our understanding of urban morphology and facilitating the spatial optimization of urban resources. However, current building height datasets subject to several restrictions: the small sample sizes hinder the effective extraction of remote sensing information required for robust, data-driven deep learning models; the limited spatial coverage of these datasets restricts the geographic diversity and representativeness necessary for capturing spatial heterogeneity; and the lack of open access to these datasets constrains their applicability and validation in broader research contexts. To address these challenges, this study constructs a large-scale building height regression dataset tailored for deep learning applications, encompassing the central urban areas of 62 cities in China and comprising 5606 samples, including Sentinel-1 and Sentinel-2 imagery alongside ground truth height values. Using this dataset, we evaluate the performance of multiple deep learning models for building height estimation, providing insights and references for future research in this domain. The dataset can be downloaded at (https://github.com/CSPON2035/SSBH).
|
