Satellite Image Segmentation Using Federated Learning: A Privacy Preserving Approach
S. Shivkumar, Suja Palaniswamy · 2023
Satellite images are images captured by satellites orbiting the Earth. They provide valuable information about the Earth's surface, including land cover, vegetation, urban areas, and more. Image segmentation is particularly useful when analysing satellite images because it allows researchers and analysts to extract meaningful information from the images. By segmenting the image into different regions, it becomes easier to identify and classify various features such as buildings, roads, water bodies, forests, and agricultural land. Satellite images can raise privacy concerns due to their ability to capture detailed information about the Earth's surface, including private properties, sensitive locations, and personal activities. In this paper, a federated learning approach is incorporated to perform semantic image segmentation tasks on satellite images. This would attempt in overcoming the privacy issue and reducing the communication load without compromising the overall performance of the model. The results obtained show that the federated learning approach is performing better than the conventional deep learning approach with a loss value of 0.0503 and 0.255 on two different datasets.