Presentation #301.06 in the session “Machine Learning in Astronomy: Data Compression, Representation and Visualization (Meeting-in-a-Meeting)”.
At present it is only possible to search astronomical archives of images via metadata. Here we present a concept and prototype systems for providing “search by image”, allowing users to find complex images similar to user-provided or user-selected examples. We will discuss transfer learning approaches to image sorting, as well as interdisciplinary work using self-supervised learning. We will also discuss our citizen science project that provides a test data for these systems, and the future of such endeavors.