Euclid Quick Data Release (Q1). Searching for giant gravitational arcs in galaxy clusters with mask region-based convolutional neural networks

TL;DR

This work develops ARTEMIDE, an automated arc-detection framework for Euclid-like cluster imaging using Mask R-CNN. It leverages a sophisticated data-generation pipeline (HST2EUCLID) to create realistic, multi-band training data by injecting lensed arcs into Euclid-like images derived from real HST clusters, with ground-truth masks for supervised learning. On a held-out test set, the model achieves a precision of 76% and a recall of 58% for arc detection, and demonstrates scalable inference for 2′×2′ fields, while revealing limitations in detecting smaller or fainter arcs and managing false positives. Inference on real Q1 clusters confirms the approach’s potential to automate large-scale lens searches, though improvements via more diverse training data and fine-tuning on real observations are identified as key steps toward robust deployment in upcoming surveys; the work provides an open-source implementation, ARTEMIDE, to accelerate community adoption and further development.

Abstract

Strong gravitational lensing (SL) by galaxy clusters is a powerful probe of their inner mass distribution and a key test bed for cosmological models. However, the detection of SL events in wide-field surveys such as Euclid requires robust, automated methods capable of handling the immense data volume generated. In this work, we present an advanced deep learning (DL) framework based on mask region-based convolutional neural networks (Mask R-CNNs), designed to autonomously detect and segment bright, strongly-lensed arcs in Euclid's multi-band imaging of galaxy clusters. The model is trained on a realistic simulated data set of cluster-scale SL events, constructed by injecting mock background sources into Euclidised Hubble Space Telescope images of 10 massive lensing clusters, exploiting their high-precision mass models constructed with extensive spectroscopic data. The network is trained and validated on over 4500 simulated images, and tested on an independent set of 500 simulations, as well as real Euclid Quick Data Release (Q1) observations. The trained network achieves high performance in identifying gravitational arcs in the test set, with a precision and recall of 76% and 58%, respectively, processing 2'x2' images in a fraction of a second. When applied to a sample of visually confirmed Euclid Q1 cluster-scale lenses, our model recovers 66% of gravitational arcs above the area threshold used during training. While the model shows promising results, limitations include the production of some false positives and challenges in detecting smaller, fainter arcs. Our results demonstrate the potential of advanced DL computer vision techniques for efficient and scalable arc detection, enabling the automated analysis of SL systems in current and future wide-field surveys. The code, ARTEMIDE, is open source and will be available at github.com/LBasz/ARTEMIDE.

Figures (10)

• Figure 1: Steps of a GCSL simulation. Left: HST2EUCLID$2'\times2'$ RGB image of the galaxy cluster Abell S1063 ($z_\mathrm{cl}=0.348$), with the main critical line (in red) for a source at $z_{\mathrm{s}}=2.92$, based on the lens model by bergamini2019. The critical line has a circularised Einstein radius of $\theta_\mathrm{E} \simeq 33"$. The green cross marks the position of the injected source to be lensed. Middle: Source plane at $z_{\mathrm{s}}=2.92$, displaying the caustic (in red) related to the main critical line. The injected source (green cross) features a Sérsic profile (index $n=1.22$, $r_{\mathrm{eff}}=\ang{;;0.11}$), with $\YE=24.8$ and the SED of a star-forming galaxy; these parameters are sampled according to the procedure described in Sect. \ref{['subsec:simulations']}. Right: Colour-composite image of the simulated GCSL system, including the critical line (red dotted line). Green boxes enclose the gravitational arcs resulting from the lensing simulation, which are also shown in the bottom inset.
• Figure 2: Mask R-CNN architecture. Adapted from jung2019.
• Figure 3: Training history of the total loss and all its individual components as a function of the training epoch, for the training (solid lines) and validation (dashed lines) data sets. The total loss (black lines) is the sum of the classification, box, and mask losses from the Mask R-CNN, along with the classification and box losses from the RPN.
• Figure 4: Left: Euclidised $2' \times 2'$ RGB image ($R=JH_\sfont{E}$, $G=\YE$, $B=\IE$) of the galaxy cluster Abell 1063, belonging to the test set; the right-most arc is the one injected via SL simulation, while the others are real. Right: single channel $2' \times 2'$ Euclidised $\IE$ image of the same cluster. The green boxes enclose the gravitational arcs (both real and simulated) present in the field, i.e. the ground truth, while the red dashed boxes are the 'gravitational arcs' found by the NN, having an object confidence score greater than $0.996$.
• Figure 5: In panel (a) we show two sets of three $P$--$R$ curves for $\mathrm{IoU}_{\mathrm{thr}} = 0.5, 0.75$, @$50$:$5$:$95$, each obtained by varying the object score threshold $p_{\mathrm{thr}}$. In panel (b) we show the $P$--$R$ curve for $\mathrm{IoU}_{\mathrm{thr}} = 0.5$, where the dots are colour-coded by the score threshold $p_{\mathrm{thr}}$ used to evaluate them. In particular, we emphasise the $P$--$R$ pair associated to our choice of score threshold $p_{\mathrm{thr}}=0.996$. Note that the recall remains relatively high even at moderate precision, reflecting the network's ability to recover most bright arcs while keeping the number of false positives manageable. Increasing the threshold shifts the operating point toward higher precision at the expense of recall. In panel (c) we show two precision and the recall curves as a function of the score threshold for $\mathrm{IoU}_{\mathrm{thr}} = 0.5$.