Transformer-Based Source Detection and Morphological Classification in LOFAR Deep-Field Continuum Images

Abstract

Radio source detection and morphological classification are fundamental for exploiting the scientific potential of modern radio continuum surveys. However, the rapidly increasing data volumes and the wide diversity of radio morphologies make traditional visual inspection infeasible and pose significant challenges for automated source finding. We apply a transformer-based set-prediction detector (RF-DETR) to 150\,MHz continuum images from the LOFAR Deep Fields for instance-level source detection and morphological classification. The method is adapted to multi-frequency-synthesis images of interferometric data and trained with a morphology-driven scheme using five mutually exclusive classes. The model is trained on the ELAIS-N1 Deep Field, where it achieves high detection and classification performance ($F1\simeq 91$ per cent), and is then applied without retraining to the other three LOFAR Deep Fields. Across all four fields, the model yields consistent catalogues with modest field-to-field differences arising from survey depth and calibration. Compared with widely used PyBDSF catalogues, RF-DETR recovers the majority of PyBDSF sources while representing classical multi-component radio galaxies as single source-level detections rather than fragmented Gaussian components. Artefact-affected and spurious detections are identified as explicit classes, allowing these detections to be distinguished from general astrophysical sources in the resulting catalogues. As external validation, RF-DETR recovers the majority of visually identified extended and giant radio galaxies in the LOFAR Deep Fields and assigns them predominantly to extended morphological classes. These results indicate that transformer-based detectors provide a practical, scalable, morphology-aware approach to source finding in deep radio surveys, with clear relevance for forthcoming facilities such as SKA-Low.

17 pages, 10 figures, 6 tables, accepted for publication in MNRAS

References in corpus (46)

• LOFAR: The LOw-Frequency ARray
• The UKIRT Infrared Deep Sky Survey (UKIDSS)
• The Murchison Widefield Array: the Square Kilometre Array Precursor at low radio frequencies
• On the fundamental dichotomy in the local radio-AGN population: accretion, evolution, and host galaxy properties
• The LOFAR Two-metre Sky Survey - I. Survey Description and Preliminary Data Release
• The LOFAR Two-metre Sky Survey - II. First data release
• The LOFAR Two-metre Sky Survey -- V. Second data release
• Australian Square Kilometre Array Pathfinder: I. System Description
• Radio galaxies and feedback from AGN jets
• The LOFAR Two Meter Sky Survey: Deep Fields, I -- Direction-dependent calibration and imaging
• Compact continuum source-finding for next generation radio surveys
• Revisiting the Fanaroff-Riley dichotomy and radio-galaxy morphology with the LOFAR Two-Metre Sky Survey (LoTSS)
• Source finding in the era of the SKA (precursors): Aegean 2.0
• The LOFAR Two-metre Sky Survey (LoTSS) III. First Data Release: optical/IR identifications and value-added catalogue
• Radio Galaxy Zoo: host galaxies and radio morphologies derived from visual inspection
• The LOFAR Two Metre Sky Survey: Deep Fields Data Release 1 -- III. Host-galaxy identifications and value added catalogues
• Anticipated Performance of the Square Kilometre Array -- Phase 1 (SKA1)
• Radio Galaxy Zoo: ClaRAN - A Deep Learning Classifier for Radio Morphologies
• Accretion mode versus radio morphology in the LOFAR Deep Fields
• ALMA observations of 99 GHz free-free and H40$\alpha$ line emission from star formation in the centre of NGC 253
• Morphological classification of radio galaxies: Capsule Networks versus Convolutional Neural Networks
• Square Kilometre Array Science Data Challenge 1: analysis and results
• DeepSource: Point Source Detection using Deep Learning
• The FIRST Classifier: Compact and Extended Radio Galaxy Classification using Deep Convolutional Neural Networks
• Free-free and H42alpha emission from the dusty starburst within NGC 4945 as observed by ALMA
• The LOFAR Two-metre Sky Survey: the radio view of the cosmic star formation history
• Cross-correlating radio continuum surveys and CMB lensing: constraining redshift distributions, galaxy bias and cosmology
• Radio astronomical images object detection and segmentation: A benchmark on deep learning methods
• YOLO-CIANNA: Galaxy detection with deep learning in radio data. I. A new YOLO-inspired source detection method applied to the SKAO SDC1
• Feature Guided Training and Rotational Standardisation for the Morphological Classification of Radio Galaxies
• Data-Efficient Classification of Radio Galaxies
• Evolutionary Map of the Universe (EMU):Compact radio sources in the SCORPIO field towards the Galactic plane
• Radio Sources Segmentation and Classification with Deep Learning
• LW-DETR: A Transformer Replacement to YOLO for Real-Time Detection
• The LOFAR Two-metre Sky Survey: Deep Fields Data Release 2. I. The ELAIS-N1 field
• Discovery of a Bimodal Environmental Distribution of Compact Ellipticals in the Local Universe
• The Most Predictive Physical Properties for the Stellar Population Radial Profiles of Nearby Galaxies
• Square Kilometre Array Science Data Challenge 3a: foreground removal for an EoR experiment
• Self-supervised contrastive learning of radio data for source detection, classification and peculiar object discovery
• Complex morphology and precession indicators of AGN jets in LoTSS DR2
• RF-DETR Object Detection vs YOLOv12 : A Study of Transformer-based and CNN-based Architectures for Single-Class and Multi-Class Greenfruit Detection in Complex Orchard Environments Under Label Ambiguity
• Star formation in the centre of NGC 1808 as observed by ALMA
• Pulsar B1237+25 Aberration/Retardation Analysis from Decimeter to Decameter Wavelength: Challenge to "Radius-to-Frequency Mapping"
• Classification of Radio Sources Through Self-Supervised Learning
• Radio-to-Submillimetre Spectral Energy Distributions of NGC 1365
• Cross-correlating the EMU Pilot Survey 1 with CMB lensing: Constraints on cosmology and galaxy bias with harmonic-space power spectra