Description
In the era of SKA, AI has emerged as a powerful approach for addressing source identification challenges in large-scale radio continuum surveys (e.g., EMU, LoTSS, and VLASS). While the community has devoted considerable effort to automatically identifying radio sources in large surveys, we find that even state-of-the-art algorithms struggle to accurately identify source-hosting galaxies and to recognize complex source radio structures. Building a large, diverse, and well-labelled dataset of sources with complex structures is essential for developing and validating automated pipelines. Robust training samples of rare, complex, and morphologically diverse radio sources are also critical for both iterating comprehensive active-learning frameworks and developing rapid-hunting algorithms for rare sources.Radio Galaxy Zoo: EMU (RGZ EMU) was launched in mid-2024 to address these challenges by combining human pattern recognition with modern machine-learning techniques¹. Using radio continuum data from the
EMU Pilot Survey conducted by ASKAP, RGZ EMU engages a global community of citizen scientists to produce high-quality classifications of radio sources. These human-generated labels support scientific discovery while also serving as training and validation data for deep learning models designed for scalable, automated source classification. The project workflow incorporates novel machine-learning components, including anomaly detection and natural language processing, to guide volunteer participation, highlight unusual sources,
and optimise volunteer effort.
Since its launch, RGZ EMU has engaged more than 3,000 volunteers worldwide and collected approximately 120,000 classifications. Phase I has already identified some rare and scientifically compelling sources, including Odd Radio Circles and candidate Giant Radio Galaxies. In this talk, we present early results from RGZ EMU Phase I, including an overview of classification data analysis, the discovery of rare source populations, and contributions to the EMUCAT cataloguing pipeline. We also describe the project’s outreach and education
activities, outline plans for Phase II, and discuss opportunities for cross-survey synergy in the context of next-generation radio continuum surveys.
| Period | 19 Aug 2026 |
|---|---|
| Event title | The 36th URSI General Assembly and Scientific Symposium |
| Event type | Conference |
| Location | Kraków, PolandShow on map |
| Degree of Recognition | International |