Self-Attending Task Generative Adversarial Network for Realistic Satellite Image Creation

TL;DR

This paper presents SATGAN, a novel generative adversarial network that enhances synthetic satellite images with realistic noise and sensor effects, preserving semantic content and reducing hallucinations.

Contribution

Introduction of SATGAN, a self-attending task GAN that improves synthetic satellite image realism while maintaining semantic integrity and reducing artifacts.

Findings

Successfully augments synthetic images with realistic noise patterns.

Preserves semantic content during image augmentation.

Reduces hallucinations and obfuscation in generated images.

Abstract

We introduce a self-attending task generative adversarial network (SATGAN) and apply it to the problem of augmenting synthetic high contrast scientific imagery of resident space objects with realistic noise patterns and sensor characteristics learned from collected data. Augmenting these synthetic data is challenging due to the highly localized nature of semantic content in the data that must be preserved. Real collected images are used to train a network what a given class of sensor's images should look like. The trained network then acts as a filter on noiseless context images and outputs realistic-looking fakes with semantic content unaltered. The architecture is inspired by conditional GANs but is modified to include a task network that preserves semantic information through augmentation. Additionally, the architecture is shown to reduce instances of hallucinatory objects or obfuscation of semantic content in context images representing space observation scenes.