GL-D-FINE: A Transformer-based Object Detector with Lens-Topology Relation for Strong Gravitational Lensing

This repository is for the paper "Detection of Strong Gravitational Lenses with Lens-Topology Relation and Arc-Assisted Annotations" by Junyu Meng et al. It is associated with manuscript number AAS72899. GL-D-FINE is a transformer-based object detector designed for identifying strong gravitational lenses in wide-field imaging surveys. Building on the D-FINE architecture, it incorporates a Lens-Topology Relation (LTR) module and utilizes an arc-assisted annotation strategy. This allows the model to explicitly capture the fine-grained morphology of lensed arcs and the unique geometric topology of strong lens systems, significantly improving localization accuracy and detection recall.