Dataset and Model Code for Built-up Area Recognition in Chongqing (2015)

This dataset and code support the study of high-precision built-up area recognition in Chongqing, China, using multi-source spatial data and attention-based deep learning models. The dataset includes raster layers for key physical and socio-economic indicators in 2015, such as population density (PD), gross domestic product (GDP), normalized difference vegetation index (NDVI), land surface temperature (LST), digital elevation model (DEM), and nighttime lights (NTL). All rasters are resampled to a 300-meter resolution and aligned in the same spatial coordinate system. Labeled built-up area positions are provided as raster masks and NumPy arrays. The repository also contains Python scripts for data preprocessing, feature normalization, and label generation. Additionally, we provide full training and validation code for the deep learning model, including the model architecture, loss functions, training loop, and evaluation logic. The model employs spatial attention mechanisms to effectively capture long-range dependencies in the urban environment. All data and scripts are shared under the Creative Commons Attribution 4.0 (CC-BY 4.0) license. This repository enables full reproducibility of the study’s results and figures, in line with the Open Data policy of the journal.