Planck-cmb-allsky

Euclid Quick Data Release (Q1): XLI. AgileLens: A scalable CNN-based pipeline for strong gravitational lens identification

August 2026 • 2026A&A...712A.207E

Authors • Euclid Collaboration • Xu, X. • Chen, R. • Li, T. • Cooray, A. R. • Schuldt, S. • Acevedo Barroso, J. A. • Stern, D. • Scott, D. • Meneghetti, M. • Despali, G. • Chopra, J. • Cao, Y. • Cheng, M. • Buda, J. • Zhang, J. • Furumizo, J. • Valencia, R. • Jiang, Z. • Tortora, C. • Lines, N. E. P. • Collett, T. E. • Fotopoulou, S. • Galan, A. • Manjón-García, A. • Gavazzi, R. • Iwamoto, L. • Kruk, S. • Millon, M. • Nugent, P. • Saulder, C. • Sluse, D. • Wilde, J. • Walmsley, M. • Courbin, F. • Metcalf, R. B. • Altieri, B. • Amara, A. • Andreon, S. • Auricchio, N. • Baccigalupi, C. • Baldi, M. • Balestra, A. • Bardelli, S. • Battaglia, P. • Bender, R. • Biviano, A. • Branchini, E. • Brescia, M. • Camera, S. • Capobianco, V. • Carbone, C. • Cardone, V. F. • Carretero, J. • Casas, S. • Castellano, M. • Castignani, G. • Cavuoti, S. • Cimatti, A. • Colodro-Conde, C. • Congedo, G. • Conselice, C. J. • Conversi, L. • Copin, Y. • Courtois, H. M. • Cropper, M. • Da Silva, A. • Degaudenzi, H. • De Lucia, G. • Dolding, C. • Dole, H. • Dubath, F. • Dupac, X. • Dusini, S. • Escoffier, S. • Farina, M. • Farinelli, R. • Farrens, S. • Ferriol, S. • Finelli, F. • Fosalba, P. • Frailis, M. • Franceschi, E. • Fumana, M. • Galeotta, S. • George, K. • Gillard, W. • Gillis, B. • Giocoli, C. • Gómez-Alvarez, P. • Gracia-Carpio, J. • Grazian, A. • Grupp, F. • Haugan, S. V. H. • Holmes, W. • Hormuth, F. • Hornstrup, A. • Jahnke, K. • Jhabvala, M. • Joachimi, B. • Kermiche, S. • Kiessling, A. • Kubik, B. • Kümmel, M. • Kunz, M. • Kurki-Suonio, H. • Le Brun, A. M. C. • Ligori, S. • Lilje, P. B. • Lindholm, V. • Lloro, I. • Mainetti, G. • Maiorano, E. • Mansutti, O. • Marcin, S. • Marggraf, O. • Martinelli, M. • Martinet, N. • Marulli, F. • Massey, R. J. • Medinaceli, E. • Mei, S. • Melchior, M. • Merlin, E. • Meylan, G. • Mora, A. • Moresco, M. • Moscardini, L. • Nakajima, R. • Neissner, C. • Nichol, R. C. • Niemi, S.-M. • Nightingale, J. W. • Padilla, C. • Paltani, S. • Pasian, F. • Pedersen, K. • Percival, W. J. • Pettorino, V. • Polenta, G. • Poncet, M. • Popa, L. A. • Raison, F. • Renzi, A. • Rhodes, J. • Riccio, G. • Romelli, E. • Roncarelli, M. • Saglia, R. • Sakr, Z. • Sapone, D. • Schirmer, M. • Schneider, P. • Schrabback, T. • Secroun, A. • Seidel, G. • Sihvola, E. • Simon, P. • Sirignano, C. • Sirri, G. • Stanco, L. • Tallada-Crespí, P. • Taylor, A. N. • Tereno, I. • Tessore, N. • Toft, S. • Toledo-Moreo, R. • Torradeflot, F. • Tutusaus, I. • Valenziano, L. • Valiviita, J. • Vassallo, T. • Verdoes Kleijn, G. • Veropalumbo, A. • Wang, Y. • Weller, J. • Zacchei, A. • Zamorani, G. • Zerbi, F. M. • Zucca, E. • Ballardini, M. • Bolzonella, M. • Burigana, C. • Cabanac, R. • Calabrese, M. • Cappi, A. • Castro, T. • Escartin Vigo, J. A. • Gabarra, L. • Hemmati, S. • Macias-Perez, J. • Maoli, R. • Martín-Fleitas, J. • Mauri, N. • Monaco, P. • Nucita, A. A. • Pezzotta, A. • Pöntinen, M. • Risso, I. • Scottez, V. • Sereno, M. • Tenti, M. • Tucci, M. • Viel, M. • Wiesmann, M. • Akrami, Y. • Andika, I. T. • Angora, G. • Anselmi, S. • Archidiacono, M. • Atrio-Barandela, F. • Bazzanini, L. • Bergamini, P. • Bertacca, D. • Bethermin, M. • Beutler, F. • Blot, L. • Borgani, S. • Brown, M. L. • Bruton, S. • Calabro, A. • Camacho Quevedo, B. • Caro, F. • Carvalho, C. S. • Cogato, F. • Conseil, S. • Cucciati, O. • Davini, S. • Desprez, G. • Díaz-Sánchez, A. • Di Domizio, S. • Diego, J. M. • Duc, P.-A. • Duret, V. • Elkhashab, M. Y. • Enia, A. • Fang, Y. • Finoguenov, A. • Franco, A. • Ganga, K. • Gasparetto, T. • Gaztanaga, E. • Giacomini, F. • Gianotti, F. • Gozaliasl, G. • Guidi, M. • Gutierrez, C. M. • Hall, A. • Hernández-Monteagudo, C. • Hildebrandt, H. • Hjorth, J. • Kajava, J. J. E. • Kang, Y. • Kansal, V. • Karagiannis, D. • Kiiveri, K. • Kim, J. • Kirkpatrick, C. C. • Lepori, F. • Leroy, G. • Lesci, G. F. • Lesgourgues, J. • Liaudat, T. I. • Liu, S. J. • Magliocchetti, M. • Magnier, E. A. • Mannucci, F. • Martins, C. J. A. P. • Maurin, L. • Miluzio, M. • Moretti, C. • Morgante, G. • Naidoo, K. • Navarro-Alsina, A. • Nesseris, S. • Paoletti, D. • Passalacqua, F. • Paterson, K. • Patrizii, L. • Pisani, A. • Potter, D. • Pratt, G. W. • Quai, S. • Radovich, M. • Rojas, K. • Roster, W. • Sacquegna, S. • Sahlén, M. • Sanders, D. B. • Sarpa, E. • Scarlata, C. • Schneider, A. • Schultheis, M. • Sciotti, D. • Sellentin, E. • Smith, L. C. • Tanidis, K. • Tao, C. • Tarsitano, F. • Testera, G. • Teyssier, R. • Tosi, S. • Troja, A. • Venhola, A. • Vergani, D. • Vernardos, G. • Verza, G. • Vinciguerra, S. • Walton, N. A. • Wright, A. H. • Yeung, H. W.

Abstract • We present an end-to-end, iterative pipeline for efficient identification of strong galaxy-galaxy lensing systems, applied to the Euclid Q1 imaging data. Starting from the VIS catalogues, we rejected point sources, applied a magnitude cut (IE <= 24) on the deflectors, and ran a pixel-level artefact-and-noise filter to build 96 × 96 pixel cutouts. The VIS+NISP colour composites were constructed with a VIS-anchored luminance scheme that preserves VIS morphology and NISP colour contrast. A VIS-only seed classifier supplies clear positives and typical impostors, which we used to curate a morphology-balanced negative set and augment scarce positives. Among the six compact convolutional neural networks (CNNs) that had been studied initially, the modified VGG16 (GlobalAveragePooling + 256/128 dense layers with the last nine layers trainable) exhibited the best performance. In our run, the training set grew from 27 seed lenses (augmented 67× to 1809) plus 2000 negatives to a colour dataset of 30 686 images. After three rounds of iterative fine-tuning, a human grading of the top 4000 candidates ranked by the final model yielded 441 Grade A/B candidate lensing systems, including 311 overlapping with the existing Q1 strong-lens catalogue, and 130 additional A/B candidates (9 As and 121 Bs) not previously reported. Independently, the model recovered 740 out of 905 (81.8%) candidate Q1 lenses within its top 20 000 predictions, considering off-centred samples. Candidates span IE ≃ 17--24 AB mag (median 21.3 AB mag) and are redder in YE - HE than the parent population, consistent with massive early-type deflectors. Each training iteration required about a week for a small team and the approach can easily be scaled to future wide-area Euclid releases. Subsequent works will focus on calibrating the selection function via lens injection, extending recall through uncertainty-aware active learning, and exploring multi-scale or attention-based neural networks with fast post hoc vetters that incorporate lens models into the classification.

Links


IPAC Authors
(alphabetical)

Shooby

Shoubaneh Hemmati

Staff Scientist


Yun_may2018

Yun Wang

Staff Scientist