![[2512.15774] Two-Step Data Augmentation for Masked Face Detection and Recognition: Turning Fake Masks to Real](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2512.15774] Two-Step Data Augmentation for Masked Face Detection and Recognition: Turning Fake Masks to Real
Summary
This article presents a two-step data augmentation framework for improving masked face detection and recognition, addressing challenges of data scarcity and distribution shifts.
Why It Matters
As masked face detection becomes increasingly relevant in security and surveillance, this research offers a novel approach to enhance the robustness of recognition systems. By combining generative techniques with rule-based methods, it provides a pathway for generating realistic training data, which is crucial for developing effective AI models in real-world applications.
Key Takeaways
- Introduces a two-step generative data augmentation framework.
- Combines rule-based mask warping with GANs for enhanced realism.
- Implements a non-mask preservation loss to improve training stability.
- Demonstrates consistent qualitative improvements over existing methods.
- Highlights the importance of data-centric approaches in AI model training.
Computer Science > Computer Vision and Pattern Recognition arXiv:2512.15774 (cs) [Submitted on 13 Dec 2025 (v1), last revised 15 Feb 2026 (this version, v2)] Title:Two-Step Data Augmentation for Masked Face Detection and Recognition: Turning Fake Masks to Real Authors:Yan Yang, George Bebis, Mircea Nicolescu View a PDF of the paper titled Two-Step Data Augmentation for Masked Face Detection and Recognition: Turning Fake Masks to Real, by Yan Yang and 2 other authors View PDF HTML (experimental) Abstract:Data scarcity and distribution shift pose major challenges for masked face detection and recognition. We propose a two-step generative data augmentation framework that combines rule-based mask warping with unpaired image-to-image translation using GANs, enabling the generation of realistic masked-face samples beyond purely synthetic transformations. Compared to rule-based warping alone, the proposed approach yields consistent qualitative improvements and complements existing GAN-based masked face generation methods such as IAMGAN. We introduce a non-mask preservation loss and stochastic noise injection to stabilize training and enhance sample diversity. Experimental observations highlight the effectiveness of the proposed components and suggest directions for future improvements in data-centric augmentation for face recognition tasks. Comments: Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) Cite as: arXiv:2512.15774 [cs.CV] (or arXiv:2...