We use a contrastive self-supervised learning framework to estimate distancesto galaxies from their photometric images . We incorporate data augmentationsfrom computer vision as well as an application-specific augmentation accountingfor galactic dust . We find that the resulting visual representations of galaxyimages are semantically useful and allow for fast similarity searches, and can be successfully fine-tuned for the task of redshift estimation . We show that pretraining on a large corpus of unlabeled data followed by fine tuning onsome labels can attain the accuracy of a fully-supervisory model which requires2-4x more labeled data .

Author(s) : Md Abul Hayat, Peter Harrington, George Stein, Zarija Lukić, Mustafa Mustafa

Links : PDF - Abstract

Code :

Keywords : data - supervised - learning - images - fine -

Leave a Reply

Your email address will not be published. Required fields are marked *