Detecting pattern irregularities in astronomical images using deep learning

Classification of astronomical images is a field of long standing interest. With modern technology, we have been able to obtain many techniques and consequently generated huge amounts and type of data. This amount of new and variety in astronomical data along with the complexities presents itself as a affair that a human analysis is not possible. In our project, we use deep learning techniques to explore the various challenges faced by astronomers when processing huge amounts of data, looking for information relevant to each target. We have implemented existing methods like InceptionV3 and CNN to detect and classify galaxies. Then we proposed to use VGG16, VGG19, and MobileNetV2 for classification and image segmentation to detect and classify galaxies. And we compared the performance of existing and proposed methods. Our proposal was approximately 20% more accurate than existing methods CNN and InceptionV3.