Traces of ancient irrigation in central Iraq detected using deep learning model

The dataset consists of predicted levees between Babylon and Uruk, utilising the Swin UNETR 48-band deep learning model with multitemporal multisource satellite data. The data used in the study were assembled from multitemporal Landsat 5, Sentinel-1 SAR, Sentinel-2 multispectral imagery, and the TanDEM-X 30 Edited DSM, combined with vegetation and moisture indices, PCA reductions of seasonal variability, and Multi-Scale Relief Model (MSRM) outputs. The post-processing included the morphological operations, such as filtering by size and closing. The results of the model after post-processing enabled it to achieve a precision of 0.5118 and a recall of 0.5908 at the pixel level.The dataset is supplemented with the results of training and the final version of the model.