Face Detection Research Articles

Multifunctional FFP2 Face Mask for White Light Disinfection and Pathogens Detection using Hybrid Nanostructures and Optical Metasurfaces.

A new generation of an FFP2 (Filtering Face Piece of type 2) smart face mask is achieved by integrating broadband hybrid nanomaterials and a self-assembled optical metasurface. The multifunctional FFP2 face mask shows simultaneously white light-assisted on-demand disinfection properties and versatile biosensing capabilities. These properties are achieved by a powerful combination of white light thermoplasmonic responsive hybrid nanomaterials, which provide excellent photo-thermal disinfection properties, and optical metasurface-based colorimetric biosensors, with a very low limit of pathogens detection. The realized system is studied in optical, morphological, spectroscopic, and cell viability assay experiments and environmental monitoring of harmful pathogens, thus highlighting the extraordinary properties in reusability and pathogens detection of the innovative face mask.

Design and implementation of a driving safety assistant system based on driver behavior

<p>These days, road accidents are one of Morocco's biggest problems. Fatigue, drowsiness, and driver behavior are among the primary causes.This research aims to develop an embedded system by image processing and computer vision to ensure driving safety by monitoring driver behavior and assist drivers to awaken from micro-sleep or fatigue due to long driving hours and various other reasons. Indeed, the driver inattention, drowsiness or driver fatigue can be detected. The suggested method is designed to support drivers if needed, based on the vehicle velocity. Once the driver crosses a certain speed limit, the program starts face detection and analyzing this data to determine whether the driver is tired, sleepy, or inattentive. This activates different alarm depending on the criticality level. It can sound a voice alert to help him wake up and drive more cautiously. The system is based on AI algorithms in image processing based on OpenCV libraries and the Python language to capture the movements of the driver's eyes and head when starting the automobile. Every algorithm is run on a Raspberry-Pi 4 card, and numerous experimentation series have demonstrated overall credible performance with success accuracy of over 93% in EAR and MAR calculations.</p>

Quarantine Order Violators System using Face Recognition (FACID)

Face recognition technology is commonly utilized for user authentication and verification by analysing a digital image of a person's face and matching it against a database of faces for identification purposes. The COVID-19 pandemic has led to mandatory quarantines and the use of quarantine bracelets for some individuals, which can be time-consuming and require a lot of effort. Face recognition technology can help raise awareness about the current situation and alleviate some of the burden associated with quarantine measures. The aim of this research is to concentrate on the face recognition of individuals with a history of quarantine, as a measure to prevent the spread of COVID-19 in educational institutions like universities, colleges, and schools. This research study concentrates on individuals with a history of quarantine orders. The system will employ the Histograms of Oriented Gradients (HOG) algorithm for face detection. Additionally, the system will utilize the Face Landmark Algorithm to compare the 128-d vector with images stored locally. The system will make use of the Helen face collection dataset for its data requirements. The aim of this study is to explore and identify techniques for detecting individuals who violate quarantine measures and issuing notifications through the proposed system. By implementing this system, it could contribute to creating a more secure environment within the educational institution and potentially reduce the spread of the virus.

Facial detection of formaldehyde by using acidichromic carbon dots and the reaction between formaldehyde and ammonium chloride

Herein we report the facial detection of formaldehyde (FA) by using an interesting red acidichromic carbon dots (ACDs) which turns blue when pH gradually decreases. The color change was attributed to the conversion between the double bonds (C=N) and single bonds (C-N) on the surface of the ACDs. Inspired by the reaction between FA and ammonium chloride that produces H+ and methenamine and decrease the pH value of the solution, a fast and simple visual detection method for FA was found with a minimum discriminated concentration of 0.04 mol/L. A fluorescence detection method for FA was also found with LOD of 0.029 mol/L and FA in real sample, e.g., shredded squid was successfully analyzed. This work provides a new idea of developing fast visual detection method for daily monitor or in-site semiquantitative assessment on FA.

Real-Time Human Profiling: Unveiling Age, Gender and Emotions Using Deep Learning

This paper introduces the development of a real-time system, which deploys an integrated use of Flask, deep learning, Convolutional Neural Networks (CNNs) and cascade classifier approach to detect age, gender, and emotion from facial images. While its numerous applications—ranging from marketing and medical services to security surveillance—immediately catches the eye, the facial recognition technology is fast becoming the hot topic. Our suggested system aims at absolutely determining the age, gender or emotional state of a person instantaneously in real life. Flask, Python micro web framework, is at the basis of the software, providing the desired functionalities for the exchange of information between the processing engine and the front end. Convolutional Neural Networks (CNNs) which are a part of deep learning for tasks involving feature extraction and classification are the main tool used by most learning algorithms. CNNs are widely used in face recognition systems mainly due to the fact that they perform very well at such tasks as image processing. On the one hand, the model uses cascade classifiers for superior face detection, thereby finding and separating face regions in input images or video streams. Unlike some of the approaches that require heavy computation, these classifiers are computationally light solutions that can run in real-time even on resource- limited devices. The system capacity to identify age, gender, and emotion accurately in real-time is exemplified through: performance evaluation. We have used various means to stress the system and ensure that it is precise and offers timely results over the given period. By means of varied testing stages, we have discovered that the system usually brings high levels of precision and validity for the numerous data sets and trial situations. In this regard, be it differentiating between individuals’ age or readily identifying the exact gender or recognizing nuanced emotional clues, the system performs at the optimum level.

Study of Multimodal Identification Algorithms using Modern Methods and Tools of Multivariate Analysis

This article aims to comprehensively investigate the theoretical and practical foundations, as well as the distinctive characteristics, underpinning the study of multimodal identification algorithms. This investigation will be conducted using state-of-the-art methods and tools of multidimensional analysis. The development of a multimodal algorithm using the method of modality fusion at the feature level encompasses the integration of various algorithms rooted in multivariate analysis. These include a combined voice activity detector, a face detector utilizing the MTCNN (multi-task cascade convolutional networks) architecture, fine-frequency cepstral coefficients, facial image features, and a decision-making module. To construct a multimodal identification algorithm, a framework for combining these algorithms based on multivariate analysis is proposed. Analysis of the acquired data indicates that “Test 1”, utilizing facial image data, exhibits the highest performance indicators, approaching nearly 100%. Tests 2 and 3 involving voice signals exhibit a minor error in the pre-processing stage, attributed to the inherent delay experienced by participants during the video conference. The proposed multimodal algorithm, integrated within a biometric identification system, enables successful user verification research through the utilization of a combined multidimensional analysis algorithm. Furthermore, the algorithm showcases superior research outcomes in comparison to other analogous multimodal identification algorithms, as it yields precise results.

Driver Facial Detection Across Diverse Road Conditions

This study emphasizes the importance of facial detection for improving road safety through driver behavior analysis. Its employs quantitative methodology to underscore the importance of facial detection in enhancing road safety through driver behavior analysis. The research utilizes the Python programming language and applies the Haar cascade method to investigate how environmental factors such as low light, shadows, and lighting changes influence the reliability of facial detection. Employing the AdaBoost algorithm, the study achieves face detection rates exceeding 95%. Practical testing with an ASUS A416JA laptop and Raspberry Pi under varied lighting conditions and distances demonstrates optimal performance in detecting faces between 30 cm and 70 cm, with reduced efficacy outside this range, particularly in low light conditions and at night. Challenges identified include decreased performance in low light conditions, emphasizing the need for improved algorithmic calibration and enhancement. Future research directions involve refining detection algorithms to effectively handle diverse environmental conditions and integrating advanced machine learning techniques, thereby enhancing the accuracy of driver behavior analysis in real-world scenarios and contributing to advancements in road safety

3D-CNN Method for Drowsy Driving Detection Based on Driving Pattern Recognition

Drowsiness impairs drivers’ concentration and reaction time, doubling the risk of car accidents. Various methods for detecting drowsy driving have been proposed that rely on facial changes. However, they have poor detection for drivers wearing a mask or sunglasses, and they do not reflect the driver’s drowsiness habits. Therefore, this paper proposes a novel method to detect drowsy driving even with facial detection obstructions, such as masks or sunglasses, and regardless of the driver’s different drowsiness habits, by recognizing behavioral patterns. We achieve this by constructing both normal driving and drowsy driving datasets and developing a 3D-CNN (3D Convolutional Neural Network) model reflecting the Inception structure of GoogleNet. This binary classification model classifies normal driving and drowsy driving videos. Using actual videos captured inside real vehicles, this model achieved a classification accuracy of 85% for detecting drowsy driving without facial obstructions and 75% for detecting drowsy driving when masks and sunglasses are worn. Our results demonstrate that the behavioral pattern recognition method is effective in detecting drowsy driving.

Automatic Face Detection of Farm Images Based on an Enhanced Lightweight Deep Learning Model

In the realm of precise management, artificial intelligence has garnered significant attention and adoption, particularly within the domain of smart agriculture. In modern animal husbandry, animal face detection is conducive to individual identification, expression detection and behavior analysis of animals, and this technological advancement holds immense importance in fostering the advancement of intelligent farming practices. In order to solve the challenge of face detection caused by similar appearance features (color, texture, etc.) and no obvious feature differences between the solid-color goats and sheep in the natural environment, this research introduces a novel approach for face detection by combining the capabilities of YOLOv5 and a convolutional block attention module (CBAM). First, datasets of goats and sheep with different angles, scales and densities were constructed. Second, the basic framework of YOLOv5 was used for object detection. To address the obstacle posed by the limited presence of distinguishing features on the faces of goats and sheep, this study aims to overcome the challenge of extracting informative facial characteristics. The CBAM block was introduced to construct the YOLOv5-CBAM model to improve the feature extraction ability. Finally, 2412 images were selected and divided into training set and verification set according to 8:1. The experimental results of this dataset show that the proposed YOLOv5-CBAM model yielded remarkable results with a precision rate of 0.970, a recall rate of 0.890, a mAP@0.5 score of 0.935, an frames per second (FPS) of 140.845, and a model size of 14.680[Formula: see text]MB. In comparison to other approaches such as Faster R-CNN, SSD, YOLOv3, and YOLOv5, the proposed model demonstrated superior performance in some aspects. In addition, it excelled in both lightweight design and overall effectiveness, and it is well-suited for real-time detection of animal faces in real-world farming settings, ensuring efficient identification and monitoring of animals within practical agricultural environments.

Face privacy protection scheme by security-enhanced encryption structure and nonlinear dynamics

This article proposes a secure communication enhancement scheme based on face detection and chaotic partition permutation. The scheme uses edge detection technology to detect facial information, which is then used as an encryption object. The hash value of the plaintext image is extracted as the secret key to the chaotic sequence generated by the chaotic system. Then a series of encryption operations are performed on the face image to obtain the final ciphertext image. In this article, two chaotic systems are used to generate pseudo-random chaotic sequences for different encryption steps. The initial key is computed by combining the hash function of the image and external parameters. The experimental results and security analysis show that the algorithm has excellent encryption effectiveness and security performance against various typical attacks.

AI Prediction for Post-Lower Blepharoplasty Age Reduction.

Aesthetic standards vary and are subjective for each person, and artificial intelligence is booming in this era. the authors aim to provide a relatively objective assessment of the aesthetic outcomes, enhancing the decision-making process and understanding of the surgical results. 150 patients who had undergone lower blepharoplasty-related surgeries were included in our study. FaceAge software was created by our research team, which included four publicly age estimation convolution neural network (CNN) models, Amazon AWS Rekognition (Seattle, WA), Microsoft Azure Face (Redmond, WA), Face++ Detect (Beijing, China), and Inferdo face detection (New York, NY). We then first used this application for age accuracy between the real age and the age estimated by the four CNNs. Second, we used this application to estimate all the preoperative and postoperative images' age of 150 patients and evaluate the effort of lower blepharoplasty. In terms of accuracy in age prediction, all CNN models exhibited a certain degree of accuracy. For all 150 patients undergoing lower blepharoplasty-related surgeries, the surgeries' effect showed about 2 years of rejuvenation with a statistically significant difference; for the sex difference, men had more age reduction than women also with a statistically significant difference; quadrilateral blepharoplasty showed the most significant effect on anti-aging effect. By using the deep learning models, lower blepharoplasty-related surgeries actually had an effect on age reduction. The potential of deep learning models will provide quantitative evidence for the rejuvenation effects of blepharoplasty or other cosmetic surgeries.

Rapid integration of face detection and task set in visually guided reaching.

The superior colliculus (SC) has been increasingly implicated in the rapid processing of evolutionarily relevant stimuli like faces, but the behavioural relevance of such processing is unclear. The SC has also been implicated in the generation of express visuomotor responses (EVR), which are very short-latency (~80 ms) bursts of muscle activity time-locked to visual target presentation. These observations led us to investigate the influence of faces on EVRs. We recorded upper limb muscle activity from healthy participants as they reached toward targets in the presence of a distractor. In some experiments, faces were used as stimuli. Across blocks of trials, we varied the instruction as to which stimulus served as the target or distractor. Doing so allowed us to assess the impact of instruction on muscle recruitment given identical visual stimuli. We found that responses were uniquely modulated in tasks involving high-contrast faces, promoting reaches toward or away from a face depending on instruction. Follow-up experiments confirmed that the phenomenon required highly salient repeated faces and was not observed to non-facial stimuli nor to faces expressing different affects. This study extends the hypothesis that the SC mediates the EVR by demonstrating that faces impact muscle recruitment at short latencies that precede cortical activity for face perception. Our results constitute direct evidence for the behavioural relevance of face detection in the brainstem, and also implicate a role for top-down cortical pre-setting of the EVR depending on task context.

RS-Xception: A Lightweight Network for Facial Expression Recognition

Facial expression recognition (FER) utilizes artificial intelligence for the detection and analysis of human faces, with significant applications across various scenarios. Our objective is to deploy the facial emotion recognition network on mobile devices and extend its application to diverse areas, including classroom effect monitoring, human–computer interaction, specialized training for athletes (such as in figure skating and rhythmic gymnastics), and actor emotion training. Recent studies have employed advanced deep learning models to address this task, though these models often encounter challenges like subpar performance and an excessive number of parameters that do not align with the requirements of FER for embedded devices. To tackle this issue, we have devised a lightweight network structure named RS-Xception, which is straightforward yet highly effective. Drawing on the strengths of ResNet and SENet, this network integrates elements from the Xception architecture. Our models have been trained on FER2013 datasets and demonstrate superior efficiency compared to conventional network models. Furthermore, we have assessed the model’s performance on the CK+, FER2013, and Bigfer2013 datasets, achieving accuracy rates of 97.13%, 69.02%, and 72.06%, respectively. Evaluation on the complex RAF-DB dataset yielded an accuracy rate of 82.98%. The incorporation of transfer learning notably enhanced the model’s accuracy, with a performance of 75.38% on the Bigfer2013 dataset, underscoring its significance in our research. In conclusion, our proposed model proves to be a viable solution for precise sentiment detection and estimation. In the future, our lightweight model may be deployed on embedded devices for research purposes.

Face processing in animal models: implications for autism spectrum disorder.

Processing facial features is crucial to identify social partners (prey, predators, or conspecifics) and recognize and accurately interpret emotional expressions. Numerous studies in both human and non-human primates provided evidence promoting the notion of inherent mechanisms for detecting facial features. These mechanisms support a representation of faces independent of prior experiences and are vital for subsequent development in social and language domains. Moreover, deficits in processing faces are a reliable biomarker of autism spectrum disorder, appearing early and correlating with symptom severity. Face processing, however, is not only a prerogative of humans: other species also show remarkable face detection abilities. In this review, we present an overview of the current literature on face detection in vertebrate models that could be relevant to the study of autism.

Deepfake detection using convolutional vision transformers and convolutional neural networks

Deepfake technology has rapidly advanced in recent years, creating highly realistic fake videos that can be difficult to distinguish from real ones. The rise of social media platforms and online forums has exacerbated the challenges of detecting misinformation and malicious content. This study leverages many papers on artificial intelligence techniques to address deepfake detection. This research proposes a deep learning (DL)-based method for detecting deepfakes. The system comprises three components: preprocessing, detection, and prediction. Preprocessing includes frame extraction, face detection, alignment, and feature cropping. Convolutional neural networks (CNNs) are employed in the eye and nose feature detection phase. A CNN combined with a vision transformer is also used for face detection. The prediction component employs a majority voting approach, merging results from the three models applied to different features, leading to three individual predictions. The model is trained on various face images using FaceForensics++ and DFDC datasets. Multiple performance metrics, including accuracy, precision, F1, and recall, are used to assess the proposed model’s performance. The experimental results indicate the potential and strengths of the proposed CNN that achieved enhanced performance with an accuracy of 97%, while the CViT-based model achieved 85% using the FaceForences++ dataset and demonstrated significant improvements in deepfake detection compared to recent studies, affirming the potential of the suggested framework for detecting deepfakes on social media. This study contributes to a broader understanding of CNN-based DL methods for deepfake detection.

Motion’s privilege in recognizing facial expressions following treatment for blindness

In his 1872 monograph, Charles Darwin posited that "… the habit of expressing our feelings by certain movements, though now rendered innate, had been in some manner gradually acquired."1 Nearly 150 years later, researchers are still teasing apart innate versus experience-dependent contributions to expression recognition. Indeed, studies have shown that face detection is surprisingly resilient to early visual deprivation,2,3,4,5 pointing to plasticity that extends beyond dogmatic critical periods.6,7,8 However, it remains unclear whether such resilience extends to downstream processing, such as the ability to recognize facial expressions. The extent to which innate versus experience-dependent mechanisms contribute to this ability has yet to be fully explored.9,10,11,12,13 To investigate the impact of early visual experience on facial-expression recognition, we studied children with congenital cataracts who have undergone sight-correcting treatment14,15 and tracked their longitudinal skill acquisition as they gain sight late in life. We introduce and explore two potential facilitators of late-life plasticity: the availability of newborn-like coarse visual acuity prior to treatment16 and the privileged role of motion following treatment.4,17,18 We find that early visual deprivation does not preclude partial acquisition of facial-expression recognition. While rudimentary pretreatment vision is sufficient to allow a low level of expression recognition, it does not facilitate post-treatment improvements. Additionally, only children commencing vision with high visual acuity privilege the use of dynamic cues. We conclude that skipping typical visual experience early in development and introducing high-resolution imagery late in development restricts, but does not preclude, facial-expression skill acquisition and that the representational mechanisms driving this learning differ from those that emerge during typical visual development.

CRSFL: Cluster-based Resource-aware Split Federated Learning for Continuous Authentication

In the ever-changing world of technology, continuous authentication and comprehensive access management are essential during user interactions with a device. Split Learning (SL) and Federated Learning (FL) have recently emerged as promising technologies for training a decentralized Machine Learning (ML) model. With the increasing use of smartphones and Internet of Things (IoT) devices, these distributed technologies enable users with limited resources to complete neural network model training with server assistance and collaboratively combine knowledge between different nodes. In this study, we propose combining these technologies to address the continuous authentication challenge while protecting user privacy and limiting device resource usage. However, the model’s training is slowed due to SL sequential training and resource differences between IoT devices with different specifications. Therefore, we use a cluster-based approach to group devices with similar capabilities to mitigate the impact of slow devices while filtering out the devices incapable of training the model. In addition, we address the efficiency and robustness of training ML models by using SL and FL techniques to train the clients simultaneously while analyzing the overhead burden of the process. Following clustering, we select the best set of clients to participate in training through a Genetic Algorithm (GA) optimized on a carefully designed list of objectives. The performance of our proposed framework is compared to baseline methods, and the advantages are demonstrated using a real-life UMDAA-02-FD face detection dataset. The results show that CRSFL, our proposed approach, maintains high accuracy and reduces the overhead burden in continuous authentication scenarios while preserving user privacy.

Fast, Zero-Reference Low-Light Image Enhancement with Camera Response Model.

Low-light images are prevalent in intelligent monitoring and many other applications, with low brightness hindering further processing. Although low-light image enhancement can reduce the influence of such problems, current methods often involve a complex network structure or many iterations, which are not conducive to their efficiency. This paper proposes a Zero-Reference Camera Response Network using a camera response model to achieve efficient enhancement for arbitrary low-light images. A double-layer parameter-generating network with a streamlined structure is established to extract the exposure ratio K from the radiation map, which is obtained by inverting the input through a camera response function. Then, K is used as the parameter of a brightness transformation function for one transformation on the low-light image to realize enhancement. In addition, a contrast-preserving brightness loss and an edge-preserving smoothness loss are designed without the requirement for references from the dataset. Both can further retain some key information in the inputs to improve precision. The enhancement is simplified and can reach more than twice the speed of similar methods. Extensive experiments on several LLIE datasets and the DARK FACE face detection dataset fully demonstrate our method's advantages, both subjectively and objectively.

Reduced spatial frequency differentiation and sex-related specificities in fearful face detection in autism: Insights from EEG and the predictive brain model.

Face processing relies on predictive processes driven by low spatial frequencies (LSF) that convey coarse information prior to fine information conveyed by high spatial frequencies. However, autistic individuals might have atypical predictive processes, contributing to facial processing difficulties. This may be more normalized in autistic females, who often exhibit better socio-communicational abilities than males. We hypothesized that autistic females would display a more typical coarse-to-fine processing for socio-emotional stimuli compared to autistic males. To test this hypothesis, we asked adult participants (44 autistic, 51 non-autistic) to detect fearful faces among neutral faces, filtered in two orders: from coarse-to-fine (CtF) and from fine-to-coarse (FtC). Results show lower d' values and longer reaction times for fearful detection in autism compared to non-autistic (NA) individuals, regardless of the filtering order. Both groups presented shorter P100 latency after CtF compared to FtC, and larger amplitude for N170 after FtC compared to CtF. However, autistic participants presented a reduced difference in source activity between CtF and FtC in the fusiform. There was also a more spatially spread activation pattern in autistic females compared to NA females. Finally, females had faster P100 and N170 latencies, as well as larger occipital activation for FtC sequences than males, irrespective of the group. Overall, the results do not suggest impaired predictive processes from LSF in autism despite behavioral differences in fear detection. However, they do indicate reduced brain modulation by spatial frequency in autism. In addition, the findings highlight sex differences that warrant consideration in understanding autistic females.

Privacy-Protected Contactless Sleep Parameters Measurement Using a Defocused Camera.

Sleep monitoring plays a vital role in various scenarios such as hospitals and living-assisted homes, contributing to the prevention of sleep accidents as well as the assessment of sleep health. Contactless camera-based sleep monitoring is promising due to its user-friendly nature and rich visual semantics. However, the privacy concern of video cameras limits their applications in sleep monitoring. In this paper, we explored the opportunity of using a defocused camera that does not allow identification of the monitored subject when measuring sleep-related parameters, as face detection and recognition are impossible on optically blurred images. We proposed a novel privacy-protected sleep parameters measurement framework, including a physiological measurement branch and a semantic analysis branch based on ResNet-18. Four important sleep parameters are measured: heart rate (HR), respiration rate (RR), sleep posture, and movement. The results of HR, RR, and movement have strong correlations with the reference (HR: R = 0.9076; RR: R = 0.9734; Movement: R = 0.9946). The overall mean absolute errors (MAE) for HR and RR are 5.2 bpm and 1.5 bpm respectively. The measurement of HR and RR achieve reliable estimation coverage of 72.1% and 93.6%, respectively. The sleep posture detection achieves an overall accuracy of 94.5%. Experimental results show that the defocused camera is promising for sleep monitoring as it fundamentally eliminates the privacy issue while still allowing the measurement of multiple parameters that are essential for sleep health informatics.