Deep Learning For Land Cover Mapping Using Sentinel-2 Imagery: A Case Study At Greater Cairo, Egypt

Land cover mapping is essential for various applications, and the integration of satellite imagery and deep learning techniques offers an accurate and efficient solution. This study focuses on mapping land cover and change detection in the new extensions of Greater Cairo, Egypt, using Sentinel-2 imagery and convolutional neural networks (CNNs). The CNN model was trained on the BigEarthNet dataset, and transfer learning was applied using a pre-trained U-Net model. The results reveal significant land cover changes in Greater Cairo, particularly in the eastern region due to the construction of the New Administrative Capital. The accuracy assessment metrics, including precision, recall, and F1-score, demonstrate high accuracy levels exceeding 90%. These findings contribute to the advancement of land cover mapping and its applications in urban development.