RGC-2

RGC-2 is the ongoing successor to RGC-1. Where the first model was trained on the 2,060-source FIRST-2060 set, RGC-2 scales to a much larger labelled dataset and moves to a multimodal architecture — combining the radio imagery with complementary information (source photometry, spectra and catalogue metadata) rather than pixels alone — to push accuracy and generalisation well beyond the first release.

The goal is a classifier robust enough to run across full radio surveys and to extend cleanly to new morphological classes, providing the environment-sensitive labels that downstream cluster science needs. Its training sets are curated through CASSA’s GAZE annotation platform.