Face Recognition Task Research Articles

Self-supervised Learning for Expression Recognition on Small-scale Data Set

In this paper, we propose a facial micro-expression recognition method utilizing self-supervised learning and Vision Transformer. We employ a contrast learning approach for self-supervision and extract image features using an attention mechanism. Various data augmentation techniques were utilized, and we specifically designed an enhancement method for facial recognition. By combining the strengths of the Vision Transformer and CNN models for feature extraction, our approach achieves improved recognition accuracy, even with limited labeled data. Experimental evaluation shows that our proposed method has good results in facial micro-expression recognition tasks.

Can machine recognize a long-missed old friend? A test to the FaceNet face recognition algorithm

Facial recognition receives heated discussion in recent years. Given the increasing demand for portable and easy-to-use face recognition programs, several coding libraries have been developed that includes the algorithm for identifying and classifying faces. One example of such libraries, the “face_recognition” library developed by Adam Geitgey allows fast execution for face recognition tasks and requires only 1 photo as training input. It adopts the Convolutional Neural Network (CNN) from OpenFace, which is built based on the FaceNet algorithm. Although the study claims that the algorithm has a high base on tests from common face dataset, it is not clear whether the algorithm can identify people consistently when they are in their different ages. In this paper, a face dataset is constructed by collecting publicly available photos from celebrities that can be traced to a certain date of taken. These photos are grouped by 5 pieces per person with each photo taken roughly 4 years apart from the previous one. Then, experiments are conducted to test the effectiveness of face_recognition library on identifying the person from these groups of photos. The results report an over 75% success rate for the algorithm on image identification tasks, indicating a good ability for the face encoding to resist the gap of time.

Neurophysiological responses to emotional faces predict dynamic fluctuations in affect in adolescents.

The ability to accurately identify and interpret others' emotions is critical for social and emotional functioning during adolescence. Indeed, previous research has identified that laboratory-based indices of facial emotion recognition and engagement with emotional faces predict adolescent mood states. Whether socioemotional information processing relates to real-world affective dynamics using an ecologically sensitive approach, however, has rarely been assessed. In the present study, adolescents (N = 62; ages 13-18) completed a Facial Recognition Task, including happy, angry, and sad stimuli, while EEG data were acquired. Participants also provided ecological momentary assessment (EMA) data probing their current level of happiness, anger, and sadness for 1-week, resulting in indices of emotion (mean-level, inertia, instability). Analyses focused on relations between (1) accuracy for and (2) prolonged engagement with (LPP) emotional faces and EMA-reported emotions. Greater prolonged engagement with happy faces was related to less resistance to changes in happiness (i.e., less happiness inertia), whereas greater prolonged engagement with angry faces associated with more resistance to changes in anger (i.e., greater anger inertia). Results suggest that socioemotional processes captured by laboratory measures have real-world implications for adolescent affective states and highlight potentially actionable targets for novel treatment approaches (e.g., just-in-time interventions). Future studies should continue to assess relations among socioemotional informational processes and dynamic fluctuations in adolescent affective states.

An attentional residual feature fusion mechanism for sheep face recognition

In the era of globalization and digitization of livestock markets, sheep are considered an essential source of food production worldwide. However, sheep behavior monitoring, disease prevention, and precise management pose urgent challenges in the development of smart ranches. To address these problems, individual identification of sheep has become an increasingly viable solution. Despite the benefits of traditional sheep individual identification methods, such as accurate tracking and record-keeping, they are labor-intensive and inefficient. Popular convolutional neural networks (CNNs) are unable to extract features for specific problems, further complicating the issue. To overcome these limitations, an Attention Residual Module (ARM) is proposed to aggregate the feature mapping between different layers of the CNN. This approach enables the general model of the CNN to be more adaptable to task-specific feature extraction. Additionally, a targeted sheep face recognition dataset containing 4490 images of 38 individual sheep has been constructed. Furthermore, the experimental data was expanded using image enhancement techniques such as rotation and panning. The results of the experiments indicate that the accuracy of the VGG16, GoogLeNet, and ResNet50 networks with the ARM improved by 10.2%, 6.65%, and 4.38%, respectively, compared to these recognition networks without the ARM. Therefore, the proposed method for specific sheep face recognition tasks has been proven effective.

Neutral Face Learning and Progressive Fusion Synthesis Network for NIR-VIS Face Recognition

To meet the strong demand for deploying face recognition systems in low-light scenarios, the Near-InfraRed and VISible (NIR-VIS) face recognition task is receiving increasing attention. However, heterogeneous faces have the characteristics of heterogeneity and non-neutrality. Heterogeneity refers to the fact that the matching images are in different modalities, and non-neutrality means that the matching images are significantly different in pose, expression, lighting, etc. Both situations pose challenges for NIR-VIS face matching. To address this problem, we propose a novel Neutral face Learning and Progressive Fusion synthesis (NLPF) network to disentangle the latent attributes of heterogeneous faces and learn neutral face representations. Our approach naturally integrates Identity-related Neutral face Learning (INL) and Attribute Progressive Fusion (APF) into a joint framework. Firstly, INL eliminates modal variations and residual variations by guiding the network to learn homogeneous neutral face feature representations, which tackles the challenge of heterogeneity and non-neutrality by mapping cross-modal images to a common neutral representation subspace. Besides, APF is presented to perform the disentanglement and reintegration of identity-related features, modality-related features and residual features in a progressive fusion manner, which helps to further purify identity-related features. Comprehensive evaluations are carried out on three mainstream NIR-VIS datasets to verify the robustness and effectiveness of the NLPF model. In particular, NLPF has competitive recognition performance on LAMP-HQ, the most challenging NIR-VIS dataset so far.

The positive effects of symbolized characteristics of Luojing cross pattern decoration on public anxiety disorders based on social psychology

BackgroundIn recent years, the incidence of anxiety disorders has been increasing. The cause of anxiety disorders is not clear, but it may be related to genetic factors, personality characteristics, cognitive processes, etc. In order to study the positive impact of anxiety disorder among people in intangible cultural heritage, the social behavior of Luojing cross pattern symbolization and the emotional face of people with anxiety disorder were identified.Subjects and MethodsThe study used task-state fMRI technology, which required participants to complete the task of emotional face recognition during the MRI scan. The study compared the difference in activation of brain regions under positive and negative emotional faces between the GAD patient group and the healthy control group. SPSS24.0 was used for data analysis.ResultsIn positive mood, compared with healthy controls, the brain regions with significantly enhanced activation in the anxiety disorder group included the right posterior central gyrus (t=3.02, P<0.05), the right superior temporal gyrus (t=3.86, P<0.05), and the right caudate nucleus (t=4.02, P<0.05). The brain regions with significantly weakened activation in the anxiety disorder group included the right superior frontal gyrus (t=3.11, P<0.05), the right medial frontal lobe (t=4.01, P<0.05), the left middle frontal gyrus (t=2.69, P<0.05), the left insula (t=3.08, P<0.05), the left anterior cingulate gyrus (t=2.38, P<0.05), the left hippocampus (t=3.48, P<0.05), etc.ConclusionsWhen the Luojing cross pattern ornament was symbolized, the mood of the people with anxiety disorders was more positive. This helps in the diagnosis and treatment of anxiety disorders from a social perspective.AcknowledgementA School-level Educational Teaching Research Project at Shanghai Business School in 2023 (No.SBS-2023-XJJG-07); Shanghai Cultural and Educational Integration project in 2023.

Sex Differences in Recognition of Face Expressions

Abstract: Emotional facial expressions are a crucial non-verbal communication skill for humans’ interactions. The current study assessed the impact of sex on emotional face recognition. A total of 125 individuals per-formed an online emotional face recognition task. The stimuli were created recutting male and female Saudi volunteers. Results showed that, except of the male participants having faster responses compared to females in general, no main significant differences between the sex group in accuracy nor any significant interaction between participant sex and the sex of the faces. Our findings suggests that the effect of sex on emotional face recognition needs further investigation with well calibrated stimuli. Limitations and the direction of future research in this area were discussed.

Implementation of Face Recognition for Attendance Recording in Online Learning

The utilization of facial recognition technology has become increasingly imperative within the realm of online learning. The current study introduces a novel system that utilizes face recognition technology to record attendance in online learning environments. The attendance system necessitates students to activate an attendance button, whereby their attendance is subsequently documented through facial recognition technology. The system recognizes students as present solely based on facial recognition. The system stores the duration of online learning activities in a database. Implementing machine learning methodologies, specifically face detection algorithms, improves precision and efficacy in administering student attendance in online education. The system utilizes Haar cascades in OpenCV to detect faces, extract features such as eyes, nose, and mouth, and classify them using LBPH. Through extensive experiments, an accuracy rate of 93.55% was achieved. The study demonstrates the effectiveness of the combined approach, showcasing the potential of Haar cascades and LBPH in face recognition tasks. The present study makes a valuable contribution to the domains of computer vision and educational technology by offering a pragmatic remedy for attendance tracking in virtual learning settings.

Elevated EEG heartbeat-evoked potentials in adolescents with more ADHD symptoms

IntroductionSymptoms of attention deficit hyperactivity disorder (ADHD) are associated with a variety of mental abnormalities, but little is known about the perception and processing of internal signals, i.e., interoception, in individuals with ADHD symptoms. This study aimed to investigate the association between ADHD symptoms and the heartbeat-evoked potential (HEP), known as a neural correlate of automatic interoceptive processing of cardiac signals, in adolescents. MethodsHEPs of 47 healthy adolescent participants (53.2 % female) with a mean age of 14.29 years were measured during an emotional face recognition task. In addition, participants completed a self-report screening for ADHD symptoms. ResultsADHD symptoms were positively related to the HEP activity during the task in three of eight EEG sectors in the left hemisphere, as well as in all sectors in the right hemisphere. DiscussionThis study is the first to demonstrate preliminary a relationship between the strength of HEP activity and ADHD symptoms in awake subjects. This finding of higher HEP amplitudes in subjects with more ADHD symptoms can be interpreted in terms of (i) increased arousal, (ii) altered neural processing of internal processes in an emotion-relevant task, and (iii) a misaligned precision-weighting process of task-irrelevant stimuli according to the predictive coding framework. These different interpretations could be reflected by previous studies showing heterogeneity of psychological deficits in individuals with ADHD symptoms. However, the generalizability to patients with diagnosed ADHD is limited due to the measurement tool for ADHD symptoms and the sample characteristics.

DAFL: Domain adaptation-based federated learning for privacy-preserving biometric recognition

With the introduction of data protection regulation in various countries, traditional centralized learning for the exploitation of sensitive biological information will gradually become unsustainable. We take face and speaker recognition systems as examples, to consider the possibility of implementing them using federated learning. It is known that federated learning methods still have some issues to be solved, e.g., parameter updates delivered can still reveal user privacy. Besides, existing privacy-preserving methods such as encryption and perturbation could pose other problems, including low accuracy and high training overhead, which can be more prominent in biometric recognition scenarios. We adopt the idea of domain adaptation and propose a new federated learning framework, DAFL, focusing on the improvement of the training efficiency and model performance, to achieve a practical privacy-preserving biometric recognition scheme for the first time. We also design a special dropout method for it to address the possible imbalance of biometric samples across users. Experiments on real-world datasets show that our approach can achieve up to 95.6% and 97.9% accuracy in face recognition and speaker recognition tasks, respectively, under medium privacy settings. Compared with existing work, it can save 56.0% of training time with a 3.1% accuracy loss in the best case.

Hypersphere guided embedding for masked face recognition

Hypersphere Guided Embedding for Masked Face Recognition has been proposed to address the problem encountered in the Masked Face Recognition task, which arises due to non-biological information from occlusions. While some existing algorithms prefer to digesting the existence of masks by probing and covering, others aim to integrate face recognition and masked face recognition tasks into a unified solution domain. In this paper, we propose a framework to enable existing methods to accommodate multiple data distributions by orthogonal subspaces. Specifically, We introduce constraints on multiple hypersphere manifolds via Multi-Center Loss and employ a Spatial Split Strategy to ensure the orthogonality of base vectors associated with different hypersphere manifolds, corresponding to distinct distribution. Our method is extensively evaluated on publicly available datasets on face recognition, mask face recognition and occlusion, demonstrating promising performance. Our code is available on an anonymous website: https://github.com/CaptainKai/HE_MFR.

Development of Deep Learning Algorithms for Improved Facial Recognition in Security Applications

This research aims to develop artificial intelligence (AI) algorithms in the context of facial recognition with a focus on increasing accuracy in difficult environmental conditions. Although facial recognition technology has made great progress, challenges such as poor lighting, variations in facial expressions, and head rotation are still problems that must be overcome. The research methodology involved collecting a wide dataset covering a wide variety of faces under various environmental conditions. This data is then processed and its features are extracted using computer image processing techniques. Furthermore, several deep neural network architectures, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), were developed, trained, and evaluated for face recognition tasks. The expected result is the development of an AI algorithm that is able to overcome challenges in facial recognition with higher accuracy than existing methods. In particular, significant improvements in facial recognition accuracy are expected especially under low lighting conditions and variations in facial expressions. This research has a major impact in a variety of security applications, such as border surveillance, building access control, and corporate security. With higher facial recognition accuracy, security risks can be significantly reduced, resulting in safer and more efficient security solutions. In conclusion, this research aims to bring innovation in facial recognition technology through advanced AI approaches, with the potential to improve security in various contexts.

Multi-label recognition of facial body constitution via MLP-like deep networks with non-equilibrium weight loss

In multi-label image classification tasks, uncertainty in the number of sample labels is a common problem, which can lead to the existence of a unbalance in the labels of samples. Facial body constitution recognition is one of the important methods of traditional Chinese medicine (TCM) body constitution recognition, which has important significance for body constitution conditioning and disease prevention. Based on the sandwich constitution of the TCM body constitution theory, the multi-label constitution corresponding to the facial image is regarded as a multi-label recognition task. A new multi-label body constitution (MLBC) dataset is proposed using facial images in this paper. We propose a novel deep network via the MLP-like architectures, including Wave-MLP, Vip, and Cycle-MLP, for the multi-label facial body constitution recognition task with a non-equilibrium weight (NEBW) loss function. The experiments compare five commonly used multi-label classification loss functions for analyzing their recognition performance. The results show the effectiveness of the proposed NEBW loss function on this task. Notably, compared with the benchmark binary cross-entropy (BCE) loss function, the three kinds of MLP-like models improve by 1.50, 1.78, and 1.44 percentage points from Acc, respectively; the improvements are 1.09, 2.00, and 2.19 percentage points from mAP, respectively. And the improvements are 1.06, 2.43, and 0.77 percentage points from AUC, respectively.

Chronotype and Time of Day Effects on a Famous Face Recognition Task with Dynamic Stimuli.

Chronotype and Time of Day (ToD) can modulate several aspects of cognitive performance. However, there is limited evidence about the effect of these variables on face recognition performance, so the aim of the present study is to investigate this influence. For this, 274 participants (82.5% females; age 18-49 years old, mean = 27.2, SD = 1.82) were shown 20 short videoclips, each gradually morphing from a general identity unfamiliar face to a famous face. Participants should press the spacebar to stop each video as soon as they could identify the famous face, and then provide the name or an unequivocal description of the person. Analysis of response times (RT) showed that evening-types recognised the faces faster than morning-types. Considering different ToD windows, the effect of chronotype was only significant in the 13h-17h and in the 21h-6h time-windows. Altogether, results suggest an advantage of evening-types on famous face recognition using dynamic stimuli with morning-types, being particularly slower during their non-optimal period.

Investigating the role of the fusiform face area and occipital face area using multifocal transcranial direct current stimulation

The functional role of the occipital face area (OFA) and the fusiform face area (FFA) in face recognition is inconclusive to date. While some research has shown that the OFA and FFA are involved in early (i.e., featural processing) and late (i.e., holistic processing) stages of face recognition respectively, other research suggests that both regions are involved in both early and late stages of face recognition. Thus, the current study aims to further examine the role of the OFA and the FFA using multifocal transcranial direct current stimulation (tDCS). In Experiment 1, we used computer-generated faces. Thirty-five participants completed whole face and facial features (i.e., eyes, nose, mouth) recognition tasks after OFA and FFA stimulation in a within-subject design. No difference was found in recognition performance after either OFA or FFA stimulation. In Experiment 2 with 60 participants, we used real faces, provided stimulation following a between-subjects design and included a sham control group. Results showed that FFA stimulation led to enhanced efficiency of facial features recognition. Additionally, no effect of OFA stimulation was found for either facial feature or whole face recognition. These results suggest the involvement of FFA in the recognition of facial features.

Enhancing Smart Home Security: Anomaly Detection and Face Recognition in Smart Home IoT Devices Using Logit-Boosted CNN Models.

Internet of Things (IoT) devices for the home have made a lot of people's lives better, but their popularity has also raised privacy and safety concerns. This study explores the application of deep learning models for anomaly detection and face recognition in IoT devices within the context of smart homes. Six models, namely, LR-XGB-CNN, LR-GBC-CNN, LR-CBC-CNN, LR-HGBC-CNN, LR-ABC-CNN, and LR-LGBM-CNN, were proposed and evaluated for their performance. The models were trained and tested on labeled datasets of sensor readings and face images, using a range of performance metrics to assess their effectiveness. Performance evaluations were conducted for each of the proposed models, revealing their strengths and areas for improvement. Comparative analysis of the models showed that the LR-HGBC-CNN model consistently outperformed the others in both anomaly detection and face recognition tasks, achieving high accuracy, precision, recall, F1 score, and AUC-ROC values. For anomaly detection, the LR-HGBC-CNN model achieved an accuracy of 94%, a precision of 91%, a recall of 96%, an F1 score of 93%, and an AUC-ROC of 0.96. In face recognition, the LR-HGBC-CNN model demonstrated an accuracy of 88%, precision of 86%, recall of 90%, F1 score of 88%, and an AUC-ROC of 0.92. The models exhibited promising capabilities in detecting anomalies, recognizing faces, and integrating these functionalities within smart home IoT devices. The study's findings underscore the potential of deep learning approaches for enhancing security and privacy in smart homes. However, further research is warranted to evaluate the models' generalizability, explore advanced techniques such as transfer learning and hybrid methods, investigate privacy-preserving mechanisms, and address deployment challenges.

A high-precision facial recognition method for small-tailed Han sheep based on an optimised Vision Transformer.

Accurate identification of individual animals plays a pivotal role in enhancing animal welfare and optimising farm production. Although Radio Frequency Identification technology has been widely applied in animal identification, this method still exhibits several limitations that make it difficult to meet current practical application requirements. In this study, we proposed ViT-Sheep, a sheep face recognition model based on the Vision Transformer (ViT) architecture, to facilitate precise animal management and enhance livestock welfare. Compared to Convolutional Neural Network (CNN), ViT is renowned for its competitive performance. The experimental procedure of this study consisted of three main steps. Firstly, we collected face images of 160 experimental sheep to construct the sheep face image dataset. Secondly, we developed two sets of sheep face recognition models based on CNN and ViT, respectively. To enhance the ability to learn sheep face biological features, we proposed targeted improvement strategies for the sheep face recognition model. Specifically, we introduced the LayerScale module into the encoder of the ViT-Base-16 model and employed transfer learning to improve recognition accuracy. Finally, we compared the training results of different recognition models and the ViT-Sheep model. The results demonstrated that our proposed method achieved the highest performance on the sheep face image dataset, with a recognition accuracy of 97.9%. This study demonstrates that ViT can successfully achieve sheep face recognition tasks with good robustness. Furthermore, the findings of this research will promote the practical application of artificial intelligence animal recognition technology in sheep production.

Behavioral signatures of face perception emerge in deep neural networks optimized for face recognition

Human face recognition is highly accurate and exhibits a number of distinctive and well-documented behavioral "signatures" such as the use of a characteristic representational space, the disproportionate performance cost when stimuli are presented upside down, and the drop in accuracy for faces from races the participant is less familiar with. These and other phenomena have long been taken as evidence that face recognition is "special". But why does human face perception exhibit these properties in the first place? Here, we use deep convolutional neural networks (CNNs) to test the hypothesis that all of these signatures of human face perception result from optimization for the task of face recognition. Indeed, as predicted by this hypothesis, these phenomena are all found in CNNs trained on face recognition, but not in CNNs trained on object recognition, even when additionally trained to detect faces while matching the amount of face experience. To test whether these signatures are in principle specific to faces, we optimized a CNN on car discrimination and tested it on upright and inverted car images. As we found for face perception, the car-trained network showed a drop in performance for inverted vs. upright cars. Similarly, CNNs trained on inverted faces produced an inverted face inversion effect. These findings show that the behavioral signatures of human face perception reflect and are well explained as the result of optimization for the task of face recognition, and that the nature of the computations underlying this task may not be so special after all.

Modeling face recognition in the predictive coding framework: A combined computational modeling and functional imaging study

The learning of new facial identities and the recognition of familiar faces are crucial processes for social interactions. Recently, a combined computational modeling and functional magnetic resonance imaging (fMRI) study used predictive coding as a biologically plausible framework to model face identity learning and to relate specific model parameters with brain activity (Apps and Tsakiris, Nat Commun 4, 2698, 2013). On the one hand, it was shown that behavioral responses on a two-option face recognition task could be predicted by the level of contextual and facial familiarity in a computational model derived from predictive-coding principles. On the other hand, brain activity in specific brain regions was associated with these parameters. More specifically, brain activity in the superior temporal sulcus (STS) varied with contextual familiarity, whereas activity in the fusiform face area (FFA) covaried with the prediction error parameter that updated facial familiarity.Literature combining fMRI assessments and computational modeling in humans still needs to be expanded. Furthermore, prior results are largely not replicated. The present study was, therefore, specifically set up to replicate these previous findings. Our results support the original findings in two critical aspects. First, on a group level, the behavioral responses were modeled best by the same computational model reported by the original authors. Second, we showed that estimates of these model parameters covary with brain activity in specific, face-sensitive brain regions. Our results thus provide further evidence that the functional properties of the face perception network conform to central principles of predictive coding. However, our study yielded diverging findings on specific computational model parameters reflected in brain activity. On the one hand, we did not find any evidence of a computational involvement of the STS. On the other hand, our results showed that activity in the right FFA was associated with multiple computational model parameters. Our data do not provide evidence for functional segregation between particular face-sensitive brain regions, as previously proposed.

Why can people with developmental prosopagnosia recognise some familiar faces? Insights from subjective experience.

Developmental prosopagnosia is a relatively common visuo-cognitive condition, characterised by impaired facial identity recognition. Impairment severity appears to reside on a continuum, however, it is unknown whether instances of milder deficits reflect the successful use of spontaneous (typical) face recognition strategies, or the application of extraneous compensatory cues to recognition. Here, we explore this issue in two studies. First, 23 adults with developmental prosopagnosia were asked about their use of spontaneous versus compensatory face recognition techniques in everyday life, using a series of closed- and open-ended questions. Second, the same participants performed a computerised famous face recognition task where they were asked to provide reasons why they could make any successful identifications. Findings from both studies suggest that people with developmental prosopagnosia can successfully, and quite frequently, use compensatory strategies to recognition, and that these cues support the majority of instances of preserved familiar face recognition. In contrast, 16 of the 23 participants were able to spontaneously recognise familiar faces on at least some occasions, but there were vast individual differences in frequencies of success. These findings have important implications for our conceptualisation of the condition, as well as for diagnostic practice.