The paper introduces AstroGAN-CA, a unified generative adversarial framework for artifact correction and data augmentation in astronomical images. It addresses the limitations of current methods and proposes a multi-task artifact correction module, physics-informed constraints, and a semi-supervised augmentation module.
Key findings
AstroGAN-CA addresses both artifact correction and data augmentation in a single architecture.
The framework introduces a multi-task artifact correction module capable of detecting and removing multiple artifact types simultaneously.
Physics-informed constraints preserve astrophysical properties during image restoration.
A semi-supervised augmentation module generates scientifically plausible training samples for rare astronomical phenomena.
Limitations & open questions
The paper is a research proposal and does not yet include experimental results.
The effectiveness of AstroGAN-CA needs to be evaluated on major astronomical datasets.