Super-Resolution for Overhead Imagery Using DenseNets and Adversarial Learning

TL;DR

This paper introduces a GAN-based super-resolution method using DenseNets for overhead imagery, achieving up to 8x resolution enhancement and evaluating its performance on multiple satellite datasets.

Contribution

It presents a novel DenseNet-GAN architecture for high-factor super-resolution of overhead images, surpassing existing methods in resolution and quality.

Findings

Achieved up to 8x super-resolution on overhead imagery.

Demonstrated superior performance over other state-of-the-art architectures.

Evaluated on SpaceNet and IARPA datasets with promising results.

Abstract

Recent advances in Generative Adversarial Learning allow for new modalities of image super-resolution by learning low to high resolution mappings. In this paper we present our work using Generative Adversarial Networks (GANs) with applications to overhead and satellite imagery. We have experimented with several state-of-the-art architectures. We propose a GAN-based architecture using densely connected convolutional neural networks (DenseNets) to be able to super-resolve overhead imagery with a factor of up to 8x. We have also investigated resolution limits of these networks. We report results on several publicly available datasets, including SpaceNet data and IARPA Multi-View Stereo Challenge, and compare performance with other state-of-the-art architectures.