GAN network galaxy probe

My first machine learning project, inspitrated by GalaxyGAN .

In the present era of astronomy, we are expecting a large amount of data that can not be fully processed by humans. Because of this, it is convenient to develop a method that analyzes the data efficiently and reliable. Our work reacts to the current situation and it is inspired by the GalaxyGAN project we have been exploring the use of neural networks in the processing of astronomical images. Observation of astronomical objects is limited by many factors and one of them is noise. Our research has focused on the removal of noise through generative adversarial networks (GANs). The results have shown that neural networks can be used for a given problem, and in comparison with available methods, the results comparable and sometimes even better.

From above: Original images, noisy images, TV-chambolle denoise, image generated by GAN.