Masked Face Recognition via CNN Embeddings Optimized with aDiscriminative Quadruplet Loss Function
Abstract
Masked face recognition remains a challenging problem because masks occlude key facial regions that arecrucial for identity verification. In this paper, we propose a deep convolutional architecture composed ofstacked Conv–BatchNorm–MaxPooling blocks followed by dropout, a flatten layer, and a dense embeddinglayer trained with an improved quadruplet loss. Each quadruplet consists of two images from thesame identity and two from different identities, enforcing compact intra-class clusters and well-separatedinter-class distributions in the embedding space. We investigate three similarity measures on the learnedembeddings: Euclidean distance, Manhattan distance, and a learned similarity network. The best performanceis obtained with Euclidean distance: on our ENSA-MFRD dataset of 32,186 masked face imagescollected from university students, the proposed model reaches an accuracy of 99.27%, outperforming thestandard triplet loss and the original quadruplet loss by 0.62% and 0.72%, respectively. Using the learnedsimilarity network, our approach also surpasses the triplet and original quadruplet losses by 37.5% and11.87%, respectively. The model is further evaluated on four public benchmarks—COMASK20, MFRD-80K, CASIA-WebMaskedFace, and LFW-SMFD—where it consistently improves accuracy, F1-score overboth baselines. These results demonstrate that the proposed architecture and enhanced quadruplet lossyield robust and discriminative representations for masked face recognition across diverse datasets andacquisition conditions.References
Organization WH et al. (2020). Advice on the use of masks in the context of covid-19: interim guidance.{it 5 June 2020. Tech. rep.,World Health Organization.}
Ngan M, Grother P, Hanaoka K (2020). Ongoing Face Recognition Vendor Test (FRVT) Part 6B: Face recognition accuracy with face masks using post-COVID-19 algorithms, {it(NISTIR), Gaithersburg, MD.} https://doi.org/10.6028/NIST.IR.8331
Pornpanomchai C, Inkuna C, (2010). Human face recognition by Euclidean distance and neural network, {it Proc. SPIE 7546, in Proc. 2nd ICDIP, 754603.}
Schroff F, Kalenichenko D, Philbin J (2015). Facenet: A unified embedding for face recognition and clustering. {it Proc. CVPR. (pp. 815–823).} https://doi.org/10.1109/CVPR.2015.7298682
Goel R, Mehmood I, Ugail H (2021). A Study of Deep Learning-Based Face Recognition Models for Sibling Identification. {it Sensors 21, 5068.} https://doi.org/10.3390/s21155068
Bromley J, Guyon I, LeCun Y, Sackinger E, Shah R (1993). Signature Verification Using a Siamese Time Delay Neural Network {it Morgan Kaufmann Publishers Inc. San Francisco, CA, USA.} https://doi.org/10.1142/S0218001493000339
Chen W, Chen X, Zhang J, Huang K (2017) Beyond triplet loss: a deep quadruplet network for person re-identification. {it In: Proc. IEEE CVPR (pp. 403-412). } https://doi.org/10.1109/CVPR.2017.145
Boutros F, Damer N, Kirchbuchner F, Kuijper A (2021) Self-restrained triplet loss for accurate masked face recognition. {it Comput. Vis. Pattern Recognit.} https://doi.org/10.1016/j.patcog.2021.108473
Huang B, Wang Z, Wang G, Jiang K, Han Z, Lu T, Liang C (2023) PLFace: Progressive learning for face recognition with mask bias. {it Pattern Recognition, 135, 109142.} https://doi.org/10.1016/j.patcog.2022.109142
Salim RJ, Surantha N (2023) Masked face recognition by zeroing the masked region without model retraining. {it Int. J. Innov. Comput. Inf. Control, 19(4), 1087-1101.}
https://doi.org/10.24507/ijicic.19.04.1087
Golwalkar R, Mehendale N (2022) Masked-face recognition using deep metric learning and FaceMaskNet-21. {it Appl Intell 52, 13268-13279.} https://doi.org/10.1007/s10489-021-03150-3
Sikha OK, Bharath B (2022) VGG16-random Fourier hybrid model for masked face recognition. {it Soft Comput 26, 12795–12810.} https://doi.org/10.1007/s00500-022-07289-0
Du H, Shi H, Liu Y, Zeng D, Mei T (2021) Towards NIR-VIS masked face recognition. {it IEEE Signal Process. Lett., 28, 768-772.} https://doi.org/10.1109/LSP.2021.3076335
Omar M, Rashedul M, Touhid M (2024) Advanced Masked Face Recognition using Robust and Light Weight Deep Learning Model. {it In IJCA (Vol. 186, No. 2, pp. 42-51).} https://doi.org/10.5120/ijca2024923351
Zhang J, An D, Zhang Y, Wang X, Wang X, Wang Q, Pan Z, Yue Y (2025) A Review on Face Mask Recognition. {it Sensors, 25(2), 387. } https://doi.org/10.3390/s25020387
Mahmoud M, Kasem MS, Kang HS (2024) A Comprehensive Survey of Masked Faces: Recognition, Detection, and Unmasking. {it Applied Sciences, 14(19), 8781.} https://doi.org/10.3390/app14198781
Huang YC, Rahardjo DAB, Shiue RH, Chen HH (2024) Masked face recognition using domain adaptation. {it Pattern Recognition, 153, 110574.} https://doi.org/10.1016/j.patcog.2024.110574
Ge Y, Liu H, Du J, Li Z, Wei Y (2023) Masked face recognition with a convolutional visual self-attention network. {it Neurocomputing, 518, 496-506. } https://doi.org/10.1016/j.neucom.2022.10.025
Alqaralleh E, Afaneh A, Toygar Ö (2023) Masked face recognition using frontal and profile faces with multiple fusion levels. {it Signal Image Video Process, 17(4), 1375-1382.}
https://doi.org/10.1007/s11760-022-02345-6
Cabani A, Hammoudi K, Benhabiles H, Melkemi M (2021) MaskedFace-Net–A dataset of correctly/incorrectly masked face images in the context of COVID-19. {it Smart Health, 19, 100144. } https://doi.org/10.1016/j.smhl.2020.100144
Zhang H, Tang J, Wu P, Li H, Zeng N (2023) A novel attention-based enhancement framework for face mask detection in complicated scenarios. {it Signal Process. Image Commun., 116, 116985. } https://doi.org/10.1016/j.image.2023.116985
Oulad-Kaddour M, Haddadou H, Palacios-Alonso D, Conde C, Cabello E (2024) Facial mask-wearing prediction and adaptive gender classification using convolutional neural networks. {it EAI Endorsed Transactions on Industrial Networks and Intelligent Systems 11(2). } http://dx.doi.org/10.4108/eetinis.v11i2.4318
Aly M (2025) Revolutionizing online education: Advanced facial expression recognition for real-time student progress tracking via deep learning model. {it Multimedia Tools and Applications, 84, 12575–12614. } http://doi.org/10.1007/s11042-024-19392-5
Vu HN, Nguyen MH, Pham C (2022) Masked face recognition with convolutional neural networks and local binary patterns. {it Applied Intelligence 52, 5497–5512.}
http://doi.org/10.1007/s10489-021-02728-1
Geekfx. (2023). CASIA WebMaskedFace [Kaggle dataset]. Retrieved from https://www.kaggle.com/datasets/geekfx/casia-webmaskedface
Dalkiran, M (2020) LFW Simulated Masked Face Dataset. {it Kaggle: San Francisco, CA, USA, 2020; pp. 1–8.
Lee CP, Lim KM (2021) Mfrd-80k: A dataset and benchmark for masked face recognition. {it Engineering Letters, 29 (4).}
DOI:
https://doi.org/10.31449/inf.v50i13.12173Downloads
Published
Issue
Section
License
Authors retain copyright in their work. By submitting to and publishing with Informatica, authors grant the publisher (Slovene Society Informatika) the non-exclusive right to publish, reproduce, and distribute the article and to identify itself as the original publisher.
All articles are published under the Creative Commons Attribution license CC BY 3.0. Under this license, others may share and adapt the work for any purpose, provided appropriate credit is given and changes (if any) are indicated.
Authors may deposit and share the submitted version, accepted manuscript, and published version, provided the original publication in Informatica is properly cited.
