Ivan DeAndres-Tame,Ruben Tolosana,Pietro Melzi,Ruben Vera-Rodriguez,Minchul Kim,Christian Rathgeb,Xiaoming Liu,Aythami Morales,Julian Fierrez,Javier Ortega-Garcia,Zhizhou Zhong,Yuge Huang,Yuxi Mi,Shouhong Ding,Shuigeng Zhou,Shuai He,Lingzhi Fu,Heng Cong,Rongyu Zhang,Zhihong Xiao,Evgeny Smirnov,Anton Pimenov,Aleksei Grigorev,Denis Timoshenko,Kaleb Mesfin Asfaw,Cheng Yaw Low,Hao Liu,Chuyi Wang,Qing Zuo,Zhixiang He,Hatef Otroshi Shahreza,Anjith George,Alexander Unnervik,Parsa Rahimi,Sébastien Marcel,Pedro C. Neto,Marco Huber,Jan Niklas Kolf,Naser Damer,Fadi Boutros,Jaime S. Cardoso,Ana F. Sequeira,Andrea Atzori,Gianni Fenu,Mirko Marras,Vitomir Štruc,Jiang Yu,Zhangjie Li,Jichun Li,Weisong Zhao,Zhen Lei,Xiangyu Zhu,Xiao-Yu Zhang,Bernardo Biesseck,Pedro Vidal,Luiz Coelho,Roger Granada,David Menotti

Abstract:Synthetic data is gaining increasing relevance for training machine learning models. This is mainly motivated due to several factors such as the lack of real data and intra-class variability, time and errors produced in manual labeling, and in some cases privacy concerns, among others. This paper presents an overview of the 2nd edition of the Face Recognition Challenge in the Era of Synthetic Data (FRCSyn) organized at CVPR 2024. FRCSyn aims to investigate the use of synthetic data in face recognition to address current technological limitations, including data privacy concerns, demographic biases, generalization to novel scenarios, and performance constraints in challenging situations such as aging, pose variations, and occlusions. Unlike the 1st edition, in which synthetic data from DCFace and GANDiffFace methods was only allowed to train face recognition systems, in this 2nd edition we propose new sub-tasks that allow participants to explore novel face generative methods. The outcomes of the 2nd FRCSyn Challenge, along with the proposed experimental protocol and benchmarking contribute significantly to the application of synthetic data to face recognition.

What problem does this paper attempt to address?

This paper introduces the second edition of the FRCSyn challenge, which is a competition on face recognition organized at the 2024 CVPR conference. The competition focuses on the problem of face recognition in the era of synthetic data. Despite the significant advancements in face recognition technology with the development of deep learning, there still exist challenges such as privacy issues, demographic biases, generalization to new scenarios, and performance limitations under complex conditions like aging, pose variations, and occlusions. The paper highlights that synthetic data can alleviate these issues by generating a large number of face images with identities that do not exist in reality, along with variations in demographic attributes and scene conditions. The first edition of the FRCSyn challenge allowed the use of synthetic data generated by the DCFace and GANDiffFace methods to train face recognition systems. The second edition introduces new subtasks, allowing participants to explore novel methods for face generation. The competition comprises two main tasks aimed at studying the limitations of face recognition technology trained solely on synthetic data and how synthetic data can help mitigate the limitations of current techniques. Each task consists of three subtasks, focusing on different types and quantities of training data (real/synthetic) and different application scenarios. Evaluation criteria include accuracy, fairness, and adaptability to different challenging conditions. Participants are allowed to use various existing or novel face generation frameworks and can choose any architecture for the face recognition system as long as its computational complexity does not exceed 50 GFLOPs. This encourages the design of innovative models and avoids the dominance of overly complex models. The paper also outlines the methods and strategies employed by top participating teams from research institutions and companies worldwide. They adopt different techniques such as diffusion-based face image synthesis, data augmentation, and pre-training strategies to improve the performance and fairness of face recognition systems. In summary, this paper describes a competition aimed at advancing face recognition technology, particularly by addressing real-world challenges through synthetic data.

Second Edition FRCSyn Challenge at CVPR 2024: Face Recognition Challenge in the Era of Synthetic Data

Second FRCSyn-onGoing: Winning Solutions and Post-Challenge Analysis to Improve Face Recognition with Synthetic Data

FRCSyn Challenge at WACV 2024:Face Recognition Challenge in the Era of Synthetic Data

FRCSyn-onGoing: Benchmarking and comprehensive evaluation of real and synthetic data to improve face recognition systems

SDFR: Synthetic Data for Face Recognition Competition

Face Recognition Using Synthetic Face Data

SynFace: Face Recognition with Synthetic Data

Synthetic Data for Face Recognition: Current State and Future Prospects

If It's Not Enough, Make It So: Reducing Authentic Data Demand in Face Recognition through Synthetic Faces

GANDiffFace: Controllable Generation of Synthetic Datasets for Face Recognition with Realistic Variations

On the use of automatically generated synthetic image datasets for benchmarking face recognition

Synthetic Data for the Mitigation of Demographic Biases in Face Recognition

Domain-Specific Face Synthesis for Video Face Recognition from a Single Sample Per Person

DCFace: Synthetic Face Generation with Dual Condition Diffusion Model

SynFER: Towards Boosting Facial Expression Recognition with Synthetic Data

CemiFace: Center-based Semi-hard Synthetic Face Generation for Face Recognition

OCFR 2022: Competition on Occluded Face Recognition From Synthetically Generated Structure-Aware Occlusions

Digi2Real: Bridging the Realism Gap in Synthetic Data Face Recognition via Foundation Models

Synthetic Face Ageing: Evaluation, Analysis and Facilitation of Age-Robust Facial Recognition Algorithms

SYN-MAD 2022: Competition on Face Morphing Attack Detection Based on Privacy-aware Synthetic Training Data