Cycle-Consistent Adversarial Networks for Realistic Pervasive Change Generation in Remote Sensing Imagery

TL;DR

This paper presents a cycle-consistent adversarial network that generates realistic pervasive changes in remote sensing images, aiding the evaluation of change detection algorithms with minimal training data.

Contribution

It introduces a novel application of CycleGAN for creating realistic change scenarios in satellite imagery, specifically snow-covered scenes, for testing change detection methods.

Findings

Successfully generated realistic snow-covered Sentinel-2 images

Demonstrated use of generated images as a test bed for anomaly detection

Low data requirement for realistic change generation

Abstract

This paper introduces a new method of generating realistic pervasive changes in the context of evaluating the effectiveness of change detection algorithms in controlled settings. The method, a cycle-consistent adversarial network (CycleGAN), requires low quantities of training data to generate realistic changes. Here we show an application of CycleGAN in creating realistic snow-covered scenes of multispectral Sentinel-2 imagery, and demonstrate how these images can be used as a test bed for anomalous change detection algorithms.