New Radio Dataset Using Multiple Surveys for Machine Learning Applications
URSI Radio Science Letters · 2025
This paper introduces the Radio Analysis of Duplicates Catalogue (RADCAT) dataset, the largest dataset to date of radio galaxies classified into Fanaroff and Riley Class I, edge-darkened (FRI) and Fanaroff and Riley Class II, edge-brightened (FRII), and compact morphologies, based on data from the Faint Images of the Radio Sky at Twenty Centimeters, the National Radio Astronomy Observatory Very Large Array Sky Survey, and the Low-Frequency Array Twometer Sky Survey.The dataset employed a novel graph theory-based method to remove duplicate sources, resulting in a final collection of 2510 unique radio sources.The dataset was further processed into multimodal images and classified using a convolutional neural network model, achieving an average accuracy of 0.80 6 0.02 on the test set, showcasing the validity of RADCAT and the potential of multimodal datasets for radio galaxy classification.