The COVID-19 pandemic raises the problem of adapting face recognition systems to the new reality, where people may wear surgical masks to cover their noses and mouths. Traditional data sets (e.g., CelebA, CASIA-WebFace) used for training these systems were released before the pandemic, so they now seem unsuited due to the lack of examples of people wearing masks. We propose a method for enhancing data sets containing faces without masks by creating synthetic masks and overlaying them on faces in the original images. Our method relies on SparkAR Studio, a developer program made by Facebook that is used to create Instagram face filters. In our approach, we use 9 masks of different colors, shapes and fabrics. We employ our method to generate a number of 445,446 (90%) samples of masks for the CASIA-WebFace data set and 196,254 (96.8%) masks for the CelebA data set, releasing the mask images at https://github.com/securifai/masked_faces. We show that our method produces significantly more realistic training examples of masks overlaid on faces by asking volunteers to qualitatively compare it to other methods or data sets designed for the same task. We also demonstrate the usefulness of our method by evaluating state-of-the-art face recognition systems (FaceNet, VGG-face, ArcFace) trained on our enhanced data sets and showing that they outperform equivalent systems trained on original data sets (containing faces without masks) or competing data sets (containing masks generated by related methods), when the test benchmarks contain masked faces.
Ensemble Learning using Transformers and Convolutional Networks for Masked Face Recognition
King Fahd University of Petroleum and Minerals(法赫德国王石油与矿产大学);SDAIA-KFUPM Joint Research Center for Artificial Intelligence(沙特数据与人工智能局-法赫德国王石油与矿产大学人工智能联合研究中心)
FocusFace: Multi-task Contrastive Learning for Masked Face Recognition
INESC TEC;Faculdade de Engenharia da Universidade do Porto(波尔图大学工程学院);Fraunhofer Institute for Computer Graphics Research IGD(弗劳恩霍夫计算机图形学研究所IGD);TU Darmstadt(达姆施塔特工业大学)
Masked Face Recognition with Generative-to-Discriminative Representations
Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);School of Cyber Security at University of Chinese Academy of Sciences(中国科学院大学网络空间安全学院);Cloud Music Inc.(网易云音乐);Department of Communication Engineering, Shanghai University(上海大学通信工程系)
由于COVID-19疫情,佩戴口罩已成为全球公共场所的强制规定。面罩遮挡面部区域的显著部分。此外,人们戴着不同类型的面具,从简单的面具到带有图形和印刷品的面具。这些都对人脸识别算法提出了新的挑战。研究人员最近提出了一些蒙面人脸数据集,用于设计算法来克服蒙面人脸识别的挑战。然而,现有的数据集缺乏文化多样性和收集在无限制的设置。像印度这样的国家,服装多样化,人们不仅限于戴传统的口罩,还穿着薄棉布印花毛巾(当地称为“gamcha”)、“stoles”和“handchiefs”等衣服来遮住脸。在本文中,我们提出了一种新的\textbf{Indian Masked Faces in the Wild (IMFW)}数据集,其中包含的图像的姿态,照明,分辨率的变化,以及由主体所戴的各种面具。我们还对现有人脸识别模型在拟议的IMFW数据集上的性能进行了基准测试。实验结果表明,现有的算法在不同的条件下存在的局限性。
My Eyes Are Up Here: Promoting Focus on Uncovered Regions in Masked Face Recognition
INESC TEC;Faculdade de Engenharia da Universidade do Porto(波尔图大学工程学院);Fraunhofer Institute for Computer Graphics Research IGD(弗劳恩霍夫计算机图形研究所 IGD);TU Darmstadt(达姆施塔特工业大学)
Look Through Masks: Towards Masked Face Recognition with De-Occlusion Distillation
Institute of Information Engineering, Chinese Academy of Sciences(中国科学院信息工程研究所);School of Cyber Security, University of Chinese Academy of Sciences(中国科学院大学网络安全学院);State Key Laboratory of Virtual Reality Technology and Systems, SCSE, Beihang University(北京航空航天大学虚拟现实技术与系统国家重点实验室);Peng Cheng Laboratory(鹏城实验室)
Fraunhofer Institute for Computer Graphics Research IGD(弗劳恩霍夫计算机图形学研究所);TU Darmstadt(达姆施塔特工业大学);Norwegian University of Science and Technology(挪威科技大学);TYAI;VICOMTECH;University of the Basque Country(巴斯克大学);Istanbul Technical University(伊斯坦布尔理工大学);ACES, Inc(ACES公司);Institute of Computing Technology, Chinese Academy of Sciences(中国科学院计算技术研究所)
FaceMAE: Privacy-Preserving Face Recognition via Masked Autoencoders
National University of Singapore(新加坡国立大学);The University of Edinburgh(爱丁堡大学);Tsinghua University(清华大学);InsightFace
人脸识别作为人工智能中最成功的应用之一,已广泛应用于安全、行政、广告、医疗等领域。然而,近年来,公共人脸数据集的隐私问题引起了越来越多的关注。以往的工作只是简单地掩盖大部分地区的人脸或合成样本使用生成模型来构建隐私保护的人脸数据集,忽略了隐私保护和数据效用之间的权衡。在本文中,我们提出了一个新的框架FaceMAE,其中的人脸隐私和识别性能同时考虑。首先,随机掩蔽的人脸图像被用于训练FaceMAE中的重建模块。我们定制的实例关系匹配(REACH)模块,以最大限度地减少真实人脸和FaceMAE重建的分布之间的差距。在部署阶段,我们使用经过训练的FaceMAE从隐藏的身份的面具中重建图像,而无需额外的训练。隐私泄漏的风险是衡量的基础上重建和原始数据集之间的人脸检索。实验证明,重建图像的身份是难以检索的。我们还对几个公共人脸数据集(即CASIA-WebFace和WebFace 260 M)进行了充分的隐私保护人脸识别。与之前的最新技术相比,FaceMAE在LFW、CFP-FP和CFDB上始终保持\textbf{reduces at least 50\% error rate}。
Masked Face Recognition Challenge: The WebFace260M Track Report
Tsinghua University(清华大学);XForwardAI;Imperial College London(伦敦帝国理工学院);InsightFace
根据世卫组织的统计数据,截至2021年8月12日,全球有超过204,617,027例确诊COVID-19病例,其中4,323,247例死亡。在冠状病毒流行期间,几乎每个人都戴着口罩。传统上,人脸识别方法主要处理未被遮挡的人脸,其中包括主要的面部特征,如眼睛,鼻子和嘴巴。在机场或实验室取下口罩进行身份验证,会增加病毒感染的风险,对目前的人脸识别系统构成巨大挑战。由于疫情突然爆发,目前还没有公开的真实世界蒙面人脸识别(MFR)基准。为应对上述问题,我们在ICCV 2021举办了COVID工作坊和面具人脸识别挑战赛。通过超大规模的WebFace 260 M基准测试和推理时间约束下的人脸识别(FRUITS)协议,这项挑战(WebFace 260 M Track)旨在推动实际MFR的前沿。由于公共评估集大多是饱和的或包含噪音,因此收集了一个新的测试集,由2,478位名人和60,926张面孔组成。同时,我们收集了世界上最大的真实世界掩蔽测试集。在WebFace 260 M Track的第一阶段,69支队伍(共833个解决方案)参加了挑战赛,49支队伍的表现超过了我们的基线。挑战赛的第二阶段将持续到2021年10月1日,并将持续排行榜。我们将在未来积极更新此报告。
A novel Facial Recognition technique with Focusing on Masked Faces
College of Science, Knowledge University(知识大学理学院);Faculty of Science, Soran University(索兰大学理学院);Technical College of Informatics, Sulaimani Polytechnic University(苏莱曼尼理工大学信息技术学院);College of Computer and Information Technology, University of Garmian(加尔米安大学计算机与信息技术学院);College of Science, University of Sulaimani(苏莱曼尼大学理学院)
Emotic Masked Autoencoder with Attention Fusion for Facial Expression Recognition
Ho Chi Minh City University of Education(胡志明市师范大学);University of Economics Ho Chi Minh City-UEH(胡志明市经济大学);Vietnamese-German University(越南德国大学);Ho Chi Minh City University of Technology(胡志明市理工大学);Posts and Telecommunications Institute of Technology(邮电技术学院)
PLGSA-Transformer: Periocular Landmark-Guided Attention with Occlusion-Adaptive Cosine Thresholding for Cross-Modal Masked and Unmasked Face Recognition
Faculty of Engineering and Computer Science, Qaiwan International University(工程与计算机科学学院,卡伊万国际大学);Department of Information and Communication Technology Center (ICTC)-System Information, Ministry of Higher Education and Scientific Research(信息与通信技术中心(ICTC)-系统信息,高等教育与科学研究部)