Conventional Face Recognition Systems Research Articles

Masked-face recognition using deep metric learning and FaceMaskNet-21.

The coronavirus disease 2019 (COVID-19) has made it mandatory for people all over the world to wear facial masks to prevent the spread of the virus. The conventional face recognition systems used for security purposes have become ineffective in the current situation since the face mask covers most of the important facial features such as nose, mouth, etc. making it very difficult to recognize the person. We have proposed a system that uses the deep metric learning technique and our own FaceMaskNet-21 deep learning network to produce 128-d encodings that help in the face recognition process from static images, live video streams, as well as, static video files. We achieved a testing accuracy of 88.92% with an execution time of fewer than 10 ms. The ability of the system to perform masked face recognition in real-time makes it suitable to recognize people in CCTV footage in places like malls, banks, ATMs, etc. Due to its fast performance, our system can be used in schools and colleges for attendance, as well as in banks and other high-security zones to grant access to only the authorized ones without asking them to remove the mask.Supplementary InformationThe online version contains supplementary material available at 10.1007/s10489-021-03150-3.

Dual-Proxy Modeling for Masked Face Recognition

With the recent worldwide COVID-19 pandemic, almost everyone wears a mask daily, leading to severe degradation in the accuracy of conventional face recognition systems. Several works improve the performance of masked faces by adopting synthetic masked face images for training. However, such methods often cause performance degradation on unmasked faces, raising the contradiction between the face recognition system's accuracy on unmasked and masked faces. In this paper, we propose a dual-proxy face recognition training method to improve masked faces’ performance while maintaining unmasked faces’ performance. Specifically, we design two fully-connected layers as the unmasked and masked feature space proxies to alleviate the significant difference between the two data distributions. The cross-space constraints are adopted to ensure the intra-class compactness and inter-class discrepancy. Extensive experiments on popular unmasked face benchmarks and masked face benchmarks, including real-world mask faces and the generated mask faces, demonstrate our method's superiority over the state-of-the-art methods on masked faces without incurring a notable accuracy degradation on unmasked faces.

A distinctive landmark-based face recognition system for identical twins by extracting novel weighted features

High similarity in facial appearances of twins has complicated the facial feature-based recognition task. This paper introduces a Distinctive Landmark-based Face Recognition (DLFR) system to mitigate the problem. Novel features are proposed based on the number of key points obtained from a modified scale-invariant feature transform algorithm and also the most distinctive landmark region of the face. A slightly altered genetic algorithm is used to optimize weights for the features. The weighted features are classified using a support vector machine classifier. Extensive experiments on 440 face images of identical twins and non-twin individuals show the higher performance of the DLFR system in comparison to a previous work in distinguishing between identical twins and individuals with similar facial images. Because of substantial differences of the DLFR system with most conventional face recognition systems, the work could motivate the research community to push forward the embodied novel idea further.

Deep Face Rectification for 360° Dual-Fisheye Cameras.

Rectilinear face recognition models suffer from severe performance degradation when applied to fisheye images captured by 360° back-to-back dual fisheye cameras. We propose a novel face rectification method to combat the effect of fisheye image distortion on face recognition. The method consists of a classification network and a restoration network specifically designed to handle the non-linear property of fisheye projection. The classification network classifies an input fisheye image according to its distortion level. The restoration network takes a distorted image as input and restores the rectilinear geometric structure of the face. The performance of the proposed method is tested on an end-to-end face recognition system constructed by integrating the proposed rectification method with a conventional rectilinear face recognition system. The face verification accuracy of the integrated system is 99.18% when tested on images in the synthetic Labeled Faces in the Wild (LFW) dataset and 95.70% for images in a real image dataset, resulting in an average accuracy improvement of 6.57% over the conventional face recognition system. For face identification, the average improvement over the conventional face recognition system is 4.51%.

Face anti-spoofing with generated near-infrared images

Reflection differences between live faces and spoof faces under near-infrared spectrum make near-infrared image based methods obtain superior performance for face anti-spoofing. However, for conventional face recognition systems, near-infrared image based methods need additional near-infrared equipment to capture the input near-infrared images. In this paper, we propose a novel face anti-spoofing method which exploits the clues in both visible light (VIS) images and near-infrared (NIR) images without utilizing any near-infrared equipment during testing. Specifically, we first propose a novel multiple categories image translation generative adversarial network (MCT-GAN) which generates corresponding NIR images for VIS live and spoof face images. Then we utilize convolution neural network to learn fusing features from both VIS images and corresponding generated NIR images for the goal of live and spoof face classification. Qualitative and quantitative experiments demonstrate that our method obtains excellent results compared to the state-of-the-art methods.

Masked Face Recognition Using Deep Metric Learning and FaceMaskNet-21

The coronavirus disease 2019 (COVID-19) has made it mandatory for people all over the world to wear facial masks to prevent the spread of the virus. As the mask hides most of the face, even humans find it difficult to identify a known face. Removing the mask can risk the health of people as it can lead to the transmission of the virus. The conventional face recognition systems used for security purposes have become incapable in the current situation because, the face mask covers most of the important facial features such as nose, mouth, etc. making it very difficult to recognize the person. The efficacy of a face recognition system is affected when trying to identify a masked face. We have proposed a system that uses the deep metric learning technique using our own FaceMaskNet-21 deep learning network to produce 128-d encodings that help in the face recognition process from static images, live video streams, as well as, video files. The HOG technique is used to enable faster recognition of faces with a mask. We achieved an accuracy of 88.92\% with an execution time of less than 10 milliseconds. The ability of the system to perform in real-time makes it suitable to recognize people in CCTV footage in malls, banks, ATMs, etc. Our system, due to its fast performance the system can be used in schools and colleges for attendance as well as in banks and other high-security zones to grant access to only authorized ones.

An extensive review on spectral imaging in biometric systems: Challenges & advancements

Spectral imaging has recently gained traction for face recognition in biometric systems. We investigate the merits of spectral imaging for face recognition and the current challenges that hamper the widespread deployment of spectral sensors for face recognition. The reliability of conventional face recognition systems operating in the visible range is compromised by illumination changes, pose variations and spoof attacks. Recent works have reaped the benefits of spectral imaging to counter these limitations in surveillance activities (defence, airport security checks, etc.). However, the implementation of this technology for biometrics, is still in its infancy due to multiple reasons. We present an overview of the existing work in the domain of spectral imaging for face recognition, different types of modalities and their assessment, availability of public databases for sake of reproducible research as well as evaluation of algorithms, and recent advancements in the field, such as, the use of deep learning-based methods for recognizing faces from spectral images.

Guest editorial: Evolving autonomous systems under realistic environments

In the conventional machine learning approaches, wesometimes have put somewhat unrealistic assumptions fortraining/test data such as sufficiency, ideal statisticalproperty (e.g., Gaussian, i.i.d.), stationary process, com-plete observations (i.d., no missing variable), etc. Theseassumptions ensure the convergence and the universality oflearning algorithms, the stability of system behaviors, andso on. However, considering our daily life, we can easilyfind the cases that violate such assumptions.In face recognition, for example, since human faceshave large variations by expressions, lighting conditions,makeup, hairstyles, and so forth, it is hard to consider allvariations of face in advance. Thus, this makes quitedifficult for a system to ensure the robustness over thespatial and temporal variations of human faces. Actually,conventional face recognition systems can achieve excel-lent performance when tested over a limited set of faceimages. However, it could drop rather drastically whenthey are operated in a practical environment. This isbecause the training set of face images may be eitherinsufficient or inappropriate for future events. Eventhough a large amount of face images are collected whenconstructing a face recognition system, all the variationsthat will happen in future cannot be considered inadvance. Therefore, a realistic solution for this is that asystem is learned incrementally whenever it misclassifiesobjects. Consequently, it is a natural assumption underrealistic environments that face images for training aregiven as a stream.Here is another example of face recognition that weshould consider under a realistic situation. In general, theobjects we encounter in our daily life can be described invarious ways. For example, from a human face, we rec-ognize multiple features such as name, gender, age, healthcondition, etc. Thus, when we see a person, we might haveto answer the name of the person from his/her face. Insome cases, however, we might have to answer the age,gender or health condition. The right answer depends on acontext we are facing. Generally, it is believed that wehumans learn and perform multiple pattern recognitiontasks in parallel or sequentially. In machine learning, therecognition of a feature (e.g., name, gender, age, etc.) isusually defined as an individual task; that is, the recogni-tion task to answer the name of a person is called personidentification, the recognition tasks to answer the genderand the age are called gender recognition and age esti-mation, respectively. From the above facts, it is natural forus to expect that a learning system can also learn multipletasks like humans (Thrun et al. 1998).There are many situations other than the above that weshould consider in constructing a learning systems underrealistic environments. For example, a part of training datamight have no supervised information (e.g., class labelsand target signals); therefore, a learner should predict theinformation to generate training data on its own. In somecase, data distributions could be changed over time withslow or sudden drifts. And there might be no prior infor-mation on useful features of high-dimensional inputs; then,a learner should select or extract optimal features dynam-ically. Furthermore, training data could also include

AAM과 가버 특징 벡터를 이용한 강인한 얼굴 인식 시스템

본 논문에서는 AAM(Active Appearance Model)과 가버 특징 벡터를 이용한 얼굴 인식 시스템을 제안한다. 가버 특징 벡터를 사용하는 대표적인 얼굴 인식 알고리즘인 EBGM(Elastic Bunch Graph Matching)은 가버 특징 벡터를 추출하기 위해 얼굴 특징점들의 검출을 필요로 한다. 그런데, EBGM에서 사용되는 얼굴 특징점 검출 방법은 가버젯 유사도에 기반하는데 이는 초기점에 민감하다. 잘못된 특징점 검출은 얼굴 인식에 영향을 미친다. AAM은 얼굴 특징점 검출에 효과적인 것으로 알려져 있다. 본 논문에서는 AAM으로 얼굴 특징점들을 대략적으로 추정하고 추정된 특징점들을 초기점으로 하여 가버젯 유사도 기반 특징점 검출방법으로 특징점 검출을 정교화하는 얼굴 특징점 검출 방법과 이에 기반한 얼굴 인식 시스템을 제안한다. 실험을 통해 제안된 특징점 검출 방법을 사용한 얼굴 인식 시스템이 EBGM과 같이 기존 가버젯 유사도만의 얼굴 특징점 검출을 이용한 얼굴 인식 시스템보다 더 나은 성능 개선을 보임을 실험을 통해 확인하였다. In this paper, we propose a face recognition system using AAM and Gabor feature vectors. EBGM, which is prominent among face recognition algorithms employing Gabor feature vectors, requires localization of facial feature points where Gabor feature vectors are extracted. However, localization of facial feature points employed in EBGM is based on Gator jet similarity and is sensitive to initial points. Wrong localization of facial feature points affects face recognition rate. AAM is known to be successfully applied to localization of facial feature points. In this paper, we propose a facial feature point localization method which first roughly estimate facial feature points using AAM and refine facial feature points using Gabor jet similarity-based localization method with initial points set by the facial feature points estimated from AAM, and propose a face recognition system based on the proposed localization method. It is verified through experiments that the proposed face recognition system using the combined localization performs better than the conventional face recognition system using the Gabor similarity-based localization only like EBGM.