Research on face recognition based on residual neural network

An empirical study of the impact of masks on face recognition

Face recognition is a kind of biometric technology that recognizes identities through human faces. At first, the speed of machine recognition of human faces was slow and the accuracy was lower than manual recognition. With the rapid development of deep learning and the application of Convolutional Neural Network (CNN) in the field of face recognition, the accuracy of face recognition has greatly improved. FaceNet is a deep learning framework commonly used in face recognition in recent years. FaceNet uses the deep learning model GoogLeNet, which has a high accuracy in face recognition. However, its network structure is too large, which causes the FaceNet to run at a low speed. Therefore, to improve the running speed without affecting the recognition accuracy of FaceNet, this paper proposes a lightweight FaceNet model based on MobileNet. This article mainly does the following works: Based on the analysis of the low running speed of FaceNet and the principle of MobileNet, a lightweight FaceNet model based on MobileNet is proposed. The model would reduce the overall calculation of the network by using deep separable convolutions. In this paper, the model is trained on the CASIA-WebFace and VGGFace2 datasets, and tested on the LFW dataset. Experimental results show that the model reduces the network parameters to a large extent while ensuring the accuracy and hence an increase in system computing speed. The model can also perform face recognition on a specific person in the video.

The paper made trend analysis on global patent in the field of face recognition by taking advantage of retrieval and analysis tool of Innography, conducted specific comparison on high-intensive patents of three countries, China, America and Korea which were top three in number of global patent applications, and analyzed existent advantages and gap in patent application and layout in China. Although our country started later in the field of face recognition, it had been prosperous development trend on number of patent applications due to technology innovation and policy support. In terms of high-intensive patent, there was a certain gap with the United States. Accordingly, This paper proposed suggestions for face recognition technology in China: To improve quality of patent application; to strengthen cooperation among enterprises, universities and other scientific research institutions in order to realize a win-win situation; To fully cultivate the talents of face recognition and artificial intelligence.

With the rapid development of electronic computers and information technology, face recognition is widely used in fields such as enterprises, entertainment, information security, and daily life. However, the current face recognition technology is still relatively poor in distinguishing facial features, resulting in a low accuracy of face recognition, which cannot meet increasing application requirements. For this reason, it is necessary to develop more accurate face recognition technology. Residual neural network (ResNet) neural network is a deep residual learning convolutional neural network, which can be used to solve the degeneration problem (i.e., after adding more layers to the neural network, the performance drops rapidly) of deep-learning neural networks. We aim to study the specific principles, calculation methods, and characteristics of the ResNet neural network, analyze the complexity of the classroom environment, propose the use of ResNet network for face feature extraction, and use additive angular margin loss for deep face recognition (ArcFace) as the loss function of the ResNet network to improve the ResNet network. Because the ArcFace function has the advantages of high performance, easy programming, low complexity and high training efficiency. We use this scheme to conduct field tests of face recognition on many students in the classroom. The test results show that this scheme enhances the discrimination of facial features and improves the accuracy of face recognition. In the case of different numbers of people, facial defects, and strong light exposure, the system can still detect and recognize faces stably. In the case of facial defects, the recognition accuracy rate is still 63.3%, and the recognition accuracy rate is still more than 60% under the illumination of strong light, and the recognition accuracy rate under correct conditions is more than 70%.

The study presents the results of the development and testing of deep learning neural network architectures, which demonstrate high accuracy rates in classifying neurophysiological data, in particular hemodynamic brain activation patterns obtained by functional near-infrared spectroscopy, during solving mathematical problems on spatial-numerical associations. The analyzed signal represents a multidimensional time series of oxyhemoglobin and deoxyhemoglobin dynamics. Taking the specificity of the fNIRS signal into account, a comparative analysis of 2 types of neural network architectures was carried out: (1) architectures based on recurrent neural networks: recurrent neural network with long short-term memory, recurrent neural network with long short-term memory with fully connected layers, bidirectional recurrent neural network with long short-term memory, convolutional recurrent neural network with long short-term memory; (2) architectures based on convolutional neural networks with 1D convolutions: convolutional neural network, fully convolutional neural network, residual neural network. Trained long short-term memory recurrent neural network architectures showed worse results in accuracy in comparison with 1D convolutional neural network architectures. Residual neural network (model_Resnet) demonstrated the highest accuracy rates in three experimental conditions more than 88% in detecting age-related differences in brain activation during spatial-numerical association tasks considering the individual characteristics of the respondents’ signal.

The advancement in the history of computer vision utilizing deep learning approaches especially convolutional neural networks have accomplished to solve difficult problems in face recognition field. Face recognition-based approach is one amongst the important identification methods which can be used as a possible substitution for conventional system of marking attendance manually, especially if a huge classroom of students is addressed for an hour session. In this proposed paper, an automated attendance system is implemented utilizing algorithms such as Histogram of Oriented Gradients (HOG) and Face landmark Estimation for face detection and finally, the deep convolutional neural network is trained for face recognition. In this system, a database containing the student’s images is created. A camera installed in the classroom captures the image of all the student in the classroom. This image is processed using machine learning algorithms to detect faces and to mark the attendance automatically in an excel sheet.KeywordsFace detectionFace recognitionDeep learningMachine learningAttendance systemPythonHOGFace landmark estimationSVMTriplet learning

Adversarial attacks have become an essential direction for security research in this area, as they generate adversarial samples to deceive face recognition systems. To defend against adversarial attacks, people need various methods to cope with the attacks, and the methods based on image denoising have widely attracted the attention of scholars for their High applicability and robustness. This paper aims to explore the application of three types of anti-attack defence in the field of face recognition by traditional filtering, JPEG compression denoising and deep learning denoising. The paper concluded the high efficiency of traditional filtering and JPEG compression denoising, with the lack of effect when defending against non-noise type attacking, however, deep learning noise reduction is far superior to traditional filtering or compression noise reduction in effect. On this basis, this paper further explores the insufficient attention to high-frequency noise denoising in the current study, and looks forward to the future research trend of combining image denoising and face recognition security, providing a reference for subsequent research.

Deep learning, in particular Convolutional Neural Network (CNN), has achieved promising results in face recognition recently. However, it remains an open question: why CNNs work well and how to design a 'good' architecture. The existing works tend to focus on reporting CNN architectures that work well for face recognition rather than investigate the reason. In this work, we conduct an extensive evaluation of CNN-based face recognition systems (CNN-FRS) on a common ground to make our work easily reproducible. Specifically, we use public database LFW (Labeled Faces in the Wild) to train CNNs, unlike most existing CNNs trained on private databases. We propose three CNN architectures which are the first reported architectures trained using LFW data. This paper quantitatively compares the architectures of CNNs and evaluate the effect of different implementation choices. We identify several useful properties of CNN-FRS. For instance, the dimensionality of the learned features can be significantly reduced without adverse effect on face recognition accuracy. In addition, traditional metric learning method exploiting CNN-learned features is evaluated. Experiments show two crucial factors to good CNN-FRS performance are the fusion of multiple CNNs and metric learning. To make our work reproducible, source code and models will be made publicly available.

In recent years, face recognition technology, and face matching particularly have broadened the application fields in various aspects of society. It is considered a combination of deep learning architecture and face recognition technology, which has been used for personal information security and safety efficiently for many years. For this, this paper aims to investigate the practical method of utilizing Siamese models to enhance the accuracy and efficiency of face matching systems. The existing challenges of low accuracy and slow recognition rates in face matching applications have been approved to be solvable by utilizing the capabilities of the Siamese model. Experimental analysis and comments from relevant practitioners demonstrate the effectiveness and potential of the Siamese model in enhancing the performance of face matching systems. To conclude, the Siamese model is introduced as a robust and efficient tool in the field of face recognition. It provides higher accuracy and efficiency compared to the traditional feature-based models. Its adaptability and advancements bring the potential to revolutionize face-matching applications and overcome current limitations. The findings from the experiments demonstrate that the utilization of the joint model can significantly enhance the performance of the matching system. The proposed model offers a potential solution to address the issue of low accuracy during the face matching phase.

Big data refers to huge information whose quantity is beyond the capacity of the system to manage, process and capture. On the basis of the biological and behavioral characteristics of the human, information gathered from which person can be recognized and referred as biometrics. Example includes finger print, face, voice and behavioral analysis etc. Among these, face recognition will not make use of any physical contact with the biometrics system; it is more secure and effective. Hence face recognition becomes one of the best technologies in the biometrics field. Although plenty of efforts have been carried out in the field of face recognition, this work has mark with many challenges in general settings. The criteria behind the successful face recognition systems are developed only under constrained situations with small sized data bases. Hence there is necessary to work under general setting that fit to big data for face recognition. This paper, aims to increase performance and accuracy of face recognition by considering face spoofing attack (veracity). Result analysis of our proposed work ensures increase in accuracy compared to other methods.

In the field of face recognition and analysis, eye state detection is an essential step, which is the prerequisite and breakthrough of drowsiness estimation and auxiliary driving. This paper presents an eye state detection method based on Weight Binarization Convolution Neural Network (WBCNN). The weight of the network is constrained by binarization, which can limit the weight to 1 or -1, reducing the power dissipation and internal storage considerably. The human eye state features which can be extracted by convolution neural network effectively, and binary network not only contributes to reducing the storage size of the model, but also accelerates the computation. Experiments on eye state detection were conducted on the Closed Eyes in the wild (CEW) and FER2013 Databases, from which the results show that our method achieved average test accuracy of 97.41% on CEW. We used the FER2013 facial expression database for pre-training, which can make up for the lack of CEW training samples. The computational speed of non-binary is slower than binary network. Moreover, less storage capacity is required by our method.

A comparative study on local binary pattern (LBP) based face recognition: LBP histogram versus LBP image

ABSTRACTPartially obscuring the face with a hat, sunglasses, scarf, or beard is referred to as partial occlusions. They are very common in forensic face recognition applications because facial images are captured in unconstrained environments. To address the issue of partial occlusion, two significant spatial domain approaches using deep learning techniques are proposed in this research work. The first approach exploits deep metric‐based learning (DML) and spatial features for face recognition. Convolution‐based face finding (CFF) technique using CNN for digital forensic face recognition tasks is employed in the second approach. Proposed approaches are experimented on the Disguised Faces in Wild (DFW) dataset, and a comparative analysis of the proposed approaches with existing approaches is presented. It is observed that the CFF technique using CNN provides a recognition accuracy of 93.75%, which performs better than DML and the existing approaches. The research work presented has significance, and results are promising in the field of unconstrained face recognition.

Photorealistic frontal view synthesis from a single face image has a wide range of applications in the field of face recognition. Although data-driven deep learning methods have been proposed to address this problem by seeking solutions from ample face data, this problem is still challenging because it is intrinsically ill-posed. This paper proposes a Two-Pathway Generative Adversarial Network (TP-GAN) for photorealistic frontal view synthesis by simultaneously perceiving global structures and local details. Four landmark located patch networks are proposed to attend to local textures in addition to the commonly used global encoder-decoder network. Except for the novel architecture, we make this ill-posed problem well constrained by introducing a combination of adversarial loss, symmetry loss and identity preserving loss. The combined loss function leverages both frontal face distribution and pre-trained discriminative deep face models to guide an identity preserving inference of frontal views from profiles. Different from previous deep learning methods that mainly rely on intermediate features for recognition, our method directly leverages the synthesized identity preserving image for downstream tasks like face recognition and attribution estimation. Experimental results demonstrate that our method not only presents compelling perceptual results but also outperforms state-of-the-art results on large pose face recognition.

In recent years, deep learning that can learn features from a dataset has been remarkably developing in the field of face recognition and voice recognition and so on. However, it is difficult to pursue cause of misjudgment result because input-output relation of deep learning is a black box. Furthermore, the content has yet to be elucidated what the judgment is based on. Therefore, when introducing deep Learning into multiple fields, it is important to understand the reason. This study aims to pursue cause of misjudgment result by intervening in the preprocessing part of deep learning using 2-dimensional discrete wavelet packet transform.