2026· IEEE Transactions on Information Forensics and Security· Vol 21, pp. 7666-7680· 0 citations· 60 references
Abstract
The purpose of face enhancement tasks is to improve the recognition of faces, thus adapting to diverse visualization and recognition demands. However, the performance of the majority methods is drastically degraded under extreme conditions, including large pose variations, low resolution, blur, occlusion, and illumination changes, which can distort facial geometry and identity related details. In this work, we construct a simple and effective face robust enhancement method. In particular, in order to maintain the identity consistency of the reconstructed face, an evolutionary learning framework for face disentanglement representation is proposed, in which we disentangle the identity and pose information of the face and unite it with identity recognition as a multi-objective optimization problem, where reconstruction, adversarial, and identity-preserving objectives are adaptively balanced. Further, in order to maintain the pose consistency of reconstructed faces, we construct a unified face pose dictionary, which forms a robust and standard pose representation by statistics and induction of the geometric structure of a large number of face images. In the conditional generation architecture, the pose dictionary could accurately guide the model to realize face reconstruction with desired poses. Extensive benchmark experiments on MS1M, LFW, CPLFW, CFP-FF, CFP-FP, and AgeDB show that the proposed method not only significantly outperforms state-of-the-art methods, but also can further stimulate the discrimination potential of existing face recognition models. Specifically, DiEL achieves an average improvement of 4.66% over the SOTA methods across six benchmark datasets, with particularly significant gains on challenging cross-pose benchmarks such as CPLFW and CFP-FP.
The concealed facial features make it difficult to identify masked faces. The features are further distorted and deteriorated when masked faces are combined with low resolution and pose variation in unrestricted contexts. Pose variation and low-resolution circumstances, along with face masks, are not assessed for current masked face recognition systems. We developed a transformer-based model a mask-aware geometry-guided vision transformer (MGViT), to address these issues. First, a learnable geometry-guided patch weighting (LGW) is used in the proposed model to suppress occluded regions and concentrate on the key face regions. Second, a mask-aware feature adaptor is created to improve the domain embeddings between masked and unmasked faces. Following that, by combining Identity and Consistency Loss functions to align identity, a strong consistency learning is integrated. Experiments with various challenges are carried out on the various masked face datasets. The proposed model, MGViT, performs well in identifying and verifying low-quality and cross-pose masked faces. Additionally, the model achieves 94.70%, 95.25%, 87.67%, and 78.56% accuracy on RMFRD, Masked LFW, Masked Multi-PIE, and Masked LR, respectively, outperforming the various state-of-the-art methods and previously proposed methods. Facial identity identification in unrestricted real-world environments may benefit from this model.
P. Kaur, Taqdir Kaur, Sahezpreet Singh· Engineering Research Express· 0 citations
Emergence of masked face recognition (MFR) as a pivotal area in biometric identification has been significantly accelerated by the global COVID-19 pandemic. In response, the research community has developed a variety of innovative techniques to address recognition and detection under occlusion, with a growing emphasis on Generative Adversarial Networks (GANs) for masked face restoration and inpainting. We examined three interconnected sub-domains: Masked Face Recognition (MFR), Face Mask Detection, and Face Unmasking (FU), each addressing unique aspects of the problem from identifying individuals with partially or fully covered faces to reconstructing occluded facial regions for improved accuracy. The core focus of this paper is on the role of GANs in overcoming occlusion by synthesizing realistic facial textures in the masked regions, thereby restoring the identity cues. Beyond technical developments, the paper analyzes the limitations and open research problems, such as maintaining identity consistency in restored images, handling diverse mask types and occlusion levels, and ensuring generalizability across different demographic groups and environments. By integrating insights from recent advances and identifying existing research gaps, this survey aims to serve as a comprehensive reference for academics and practitioners engaged in the development of robust, privacy-aware, and ethically responsible masked face recognition systems enhanced by GANs.
Payal Parekh, Hina Choksi, Mahesh Goyani et al.· ITEGAM- Journal of Engineeri...· 0 citations
Face pose variation is easy to cause feature loss and face recognition rate decline, which is a key problem in the field of face image generation and recognition. Existing face generation methods based on encoder-decoder often focus on pose conversion and lose facial detail features. This paper proposes a Multitask Detail Compensated Generative Adversarial Networks (MDC-GAN), which improves the effect of generating face details and preserving identity through multi-task learning and multi-scale feature fusion. It has achieved good face reconstruction results on the FERET database, and the results are better than other current methods.
Shasha Wu, Hualong Zhang, Tianci Liu et al.· International Conference on...· 0 citations
Face recognition becomes an important biometric implementation in any surveillance, access control, forensic and intelligent security system applications. Despite of technical advances in deep learning models, recognition accuracy normally affected by facial pose, illumination, occlusion, expression, aging factors and limited availability of labelled dataset. These challenges reduce the robustness of face recognition models in real world environments [1][2][3]. Recent researches demonstrated that the super-resolution techniques based on Generative Adversarial Networks (GANs) reconstruct high-quality facial images very effectively from low-resolution input images. It supports to improve feature representation and face recognition [4][5]. Therefore, this research motivated from this advancement and proposes an Adaptive Super-Resolution Generative Adversarial Network (Adaptive SRGAN) for face recognition. It integrates adaptive learning with image super resolution to reconstruct identity preserving high resolution facial images by employing adaptive learning rate optimization, dynamic loss weighting, attention guided feature enhancement and identity preserving loss functions. However, it enhances reconstruction quality by preserving discriminative facial characteristics [6]. The proposed model is expected to achieve higher Peak Signal to Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), recognition accuracy, precision, recall, and F1-score while reducing false acceptance and false rejection rates. Simultaneously, Adaptive SRGAN delivers a robust and scalable solution for improving face recognition systems.
M. Kirubakaran, A. S. Aneeshkumar· International journal of com...· 0 citations
Face recognition in unconstrained environments remains a challenging problem in computer vision due to variations in pose, illumination, expression, and occlusion. This paper proposes a novel attention-enhanced ArcFace-based deep learning framework that integrates a Residual CNN backbone with Convolutional Block Attention Module (CBAM) and ArcFace loss for robust face recognition. Unlike existing approaches that rely on large-scale external pretraining datasets, the proposed framework is trained exclusively on the Labelled Faces in the Wild (LFW) dataset, demonstrating data-efficient learning. The system is evaluated on both 1:1 verification and 1:N identification protocols. Experimental results demonstrate superior performance with verification accuracy of 95.70%, identification accuracy of 89.75%, ROC-AUC of 99.16%, and True Positive Rate (TPR) of approximately 92% at a 1% False Positive Rate (FPR). The novelty lies in the synergistic integration of attention mechanisms with angular margin-based metric learning, achieving competitive performance without external pretraining. Comparative analysis with state-of-the-art methods including DeepFace, FaceNet, VGGFace, SphereFace, and baseline ArcFace validates the effectiveness of the proposed attention-guided approach for unconstrained face recognition tasks.
Samadhan S. Ghodke, Prapti D. Deshmukh· International journal of com...· 0 citations
Diffusion models have achieved strong results in high-fidelity image synthesis, but their iterative sampling process makes large-scale generation computationally expensive. This limitation is especially relevant when generating synthetic face datasets for face recognition, where a large number of subjects with many samples in different poses, expressions, ages, etc., are required. In this work, we show that identity-conditioned face synthesis can be performed at a substantially lower computational cost by a latent Consistency Model with few iterations, without compromising image quality. For training, we distill knowledge from the foundation Diffusion Model Arc2Face (teacher) by adapting its original text-to-image pipeline to an embedding-to-face setting, replacing textual prompts with ArcFace identity embeddings. Our distilled model (student) generates identity-conditioned face images with an average inference time of 0.4819 seconds per image, compared with 2.102 seconds for Arc2Face, resulting in a 4.36$\times$ speed-up. Quantitative results, based on FID scores, show that the distilled model remains competitive with Arc2Face across all evaluation protocols. On 100k generated images, it achieves near-parity on CelebA (13.921 vs. 12.928) and outperforms the teacher on WebFace42M (9.317 vs. 9.802). Further evaluations on Synth-500 and AgeDB show a moderate performance gap for the former but comparable results for the latter. These results indicate that Arc2Face can be accelerated through task-specific latent consistency distillation while preserving high image quality for large-scale synthetic face generation. Our proposal is publicly available at https://github.com/UFPR-IPASP-PR/FaceRec-IdentityConsistency.
Tiago Kienen Chaves, Bernardo Biesseck, David Menotti· 0 citations
