Improved UNet for Semantic Segmentation in Remote Sensing

In the field of remote sensing, Semantic Segmentation has become an important technique because it allows to precisely classify each image pixel to obtain more general and useful interpretations. This aspect becomes even more critical in the context of UAV technology, which has completely changed several fields by providing outstanding spatial resolution possibilities for image sensing. In our study, we present an innovative approach to improve the accuracy of semantic segmentation networks used in remote sensing. The standard UNet architecture has been modified in our proposed approach by introducing a Spatial Alignment Module (SAM). This multi-scale approach dynamically optimizes the use of contextual information at different scales. This innovative approach leverages UAV technology to provide improved spatial resolution and flexibility in image acquisition, enhancing remote sensing applications.