Semantic segmentation of remote sensing images using U-net and its variants : Conference New Technologies of Information and Communication (NTIC 2022)

The process of dividing aerial images into distinct segments based on their semantic content is a crucial aspect of computer vision research that has numerous real-world applications, including disaster monitoring, land mapping, weather forecasting, and agriculture. This work provides a comprehensive overview of the methods used for semantic segmentation of aerial images and how deep neural networks, especially convolutional neural networks and the U-net architecture, can be employed to achieve this. The methods discussed are trained on aerial image datasets, with the results demonstrating the effectiveness of using U-net and its variations for semantic segmentation of aerial imagery.