Super-Resolution for Satellite Image Segmentation

This project aims to evaluate and benchmark state-of-the-art super-resolution techniques for enhancing satellite imagery at various resolution levels. The primary objective is to assess the performance of these methods in improving image quality, with a focus on their ability to recover fine details from low-resolution satellite data. Additionally, the project investigates the impact of super-resolution-enhanced images on downstream tasks, specifically instance segmentation, by comparing the performance of segmentation models trained on original versus super-resolved images. Through a series of controlled experiments, this research seeks to identify the most effective super-resolution techniques and assess their utility in preprocessing satellite imagery for improved accuracy in geospatial analysis. Ultimately, the findings aim to inform the adoption of super-resolution methods in real-world satellite image processing workflows, with potential applications in environmental monitoring, urban planning, and disaster response. In-depth: