Deep Guidance Network for Satellite Image Segmentation using U-NET Models

Abstract

This project describes satellite image segmentation using a U-net-based satellite image-based segmentation algorithm, The purpose of this research is to develop a convolutional machine learning used to create a land cover sorting map based on satellite images prototype by a reformed U-Net arrangement. The goal of this project is to develop and evaluate convolutional prototypes for automated land cover mapping. The utility for enhancing the accuracy of land cover mapping and identifying changes. For land cover sorting and semantic sorting, a dataset was created, and machine learning models were trained by the authors. The findings were analyzed across three different geographical classification levels using picture segmentation based on satellite images. One of the two key datasets for the investigation was the BigEarthNet satellite picture database. This unique and most recent collection, which comprises Sentinel-2 satellite photos from ten European nations, was released in 2019.