Face Detection Task Research Articles

ADYOLOv5-Face: An Enhanced YOLO-Based Face Detector for Small Target Faces

Benefiting from advancements in generic object detectors, significant progress has been achieved in the field of face detection. Among these algorithms, the You Only Look Once (YOLO) series plays an important role due to its low training computation cost. However, we have observed that face detectors based on lightweight YOLO models struggle with accurately detecting small faces. This is because they preserve more semantic information for large faces while compromising the detailed information for small faces. To address this issue, this study makes two contributions to enhance detection performance, particularly for small faces: (1) modifying the neck part of the architecture by integrating a Gather-and-Distribute mechanism instead of the traditional Feature Pyramid Network to tackle the information fusion challenges inherent in YOLO-based models; and (2) incorporating an additional detection head specifically designed for detecting small faces. To evaluate the performance of the proposed face detector, we introduce a new dataset named XD-Face for the face detection task. In the experimental section, the proposed model is trained using the Wider Face dataset and evaluated on both Wider Face and XD-face datasets. Experimental results demonstrate that the proposed face detector outperforms other excellent face detectors across all datasets involving small faces and achieved improvements of 1.1%, 1.09%, and 1.35% in the AP50 metric on the WiderFace validation dataset compared to the baseline YOLOv5s-based face detector.

Dense pedestrian face detection in complex environments

To address the problem of dense crowd face detection in complex environments, this paper proposes a face detection model named Deep and Compact Face Detection (DCFD), which adopts an improved lightweight EfficientNetV2 network to replace the backbone network of RetinaFace. A large kernel attention mechanism is introduced to address the face detection task more accurately. The backbone network, an improved efficient channel attention (ECA) mechanism, is added to further improve the algorithm performance. The feature fusion module is an improved neural architecture search feature pyramid network (NAS-FPN) that significantly improves the face detection accuracy in different scenes. To balance the training process of positive and negative samples, we use the focus loss function to replace the traditional cross-entropy loss function. In different environments, the DCFD algorithm has shown efficient face detection performance. This algorithm provides not only a feasible and effective solution for solving the problem of face detection in dense groups but also an important basis for improving the accuracy of face detection models in practical applications.

Multi-model fusion of DSFD and Zero-DCE for low-light face detection

Due to the wide application of face detection, face recognition task under extreme lighting conditions inevitably becomes one of the challenges. Under low-light conditions, the face detection task becomes more complex and difficult due to the presence of more noise and lower visibility in the image. For this extreme environment, this paper proposes a multi-model-assisted solution that combines image enhancement and face detection models to handle this task. The DarkFace dataset was used in this study's validation of a number of distinct face identification models and picture enhancement techniques. The experimental findings demonstrate that, on this dataset, the DSFD (Dual Shot Face Detector) and Zero-DCE (Zero-Reference Deep Curve Estimation) methods perform noticeably better than the other techniques, achieving a recall of 24.61%. Therefore, these two algorithms are selected as the core processing flow in this paper to improve the facial detection system's low-light performance. This combination offers a workable answer to the problem of face detection in low light by effectively increasing the visibility of the picture and raising the accuracy of face detection.

Face detection in contextual scenes.

Object and scene perception are intertwined. When objects are expected to appear within a particular scene, they are detected and categorised with greater speed and accuracy. This study examined whether such context effects also moderate the perception of social objects such as faces. Female and male faces were embedded in scenes with a stereotypical female or male context. Semantic congruency of these scene contexts influenced the categorisation of faces (Experiment 1). These effects were bi-directional, such that face sex also affected scene categorisation (Experiment 2), suggesting concurrent automatic processing of both levels. In contrast, the more elementary task of face detection was not affected by semantic scene congruency (Experiment 3), even when scenes were previewed prior to face presentation (Experiment 4). This pattern of results indicates that semantic scene context can affect categorisation of faces. However, the earlier perceptual stage of detection appears to be encapsulated from the cognitive processes that give rise to this contextual interference.

Functional performance of a vibrotactile sensory substitution device in people with profound vision loss.

This study has shown a vibrotactile sensory substitution device (SSD) prototype, VibroSight, has the potential to improve functional outcomes (i.e., obstacle avoidance, face detection) for people with profound vision loss, even with brief familiarization (<20 minutes). Mobility aids such as long canes are still the mainstay of support for most people with vision loss, but they do have limitations. Emerging technologies such as SSDs are gaining widespread interest in the low vision community. The aim of this project was to assess the efficacy of a prototype vibrotactile SSD for people with profound vision loss in the face detection and obstacle avoidance tasks. The VibroSight device was tested in a movement laboratory setting. The first task involved obstacle avoidance, in which participants were asked to walk through an obstacle course. The second was a face detection task, in which participants were asked to step toward the first face they detected. Exit interviews were also conducted to gather user experience data. Both people with low vision (n = 7) and orientation and mobility instructors (n = 4) completed the tasks. In obstacle avoidance task, participants were able to use the device to detect (p<0.001) and avoid (p<0.001) the obstacles within a significantly larger range, but were slower (p<0.001), when compared with without the device. In face detection task, participants demonstrated a great level of accuracy, precision, and sensitivity when using the device. Interviews revealed a positive user experience, although participants identified that they would require a lighter and compact design for real-world use. Overall, the results verified the functionality of vibrotactile SSD prototype. Further research is warranted to evaluate the user performance after an extended training program and to add new features, such as object recognition software algorithms, into the device.

Autistic traits specific to communication ability are associated with performance on a Mooney face detection task.

Difficulties in global face processing have been associated with autism. However, autism is heterogenous, and it is not known which dimensions of autistic traits are implicated in face-processing difficulties. To address this gap in knowledge, we conducted two experiments to examine how identification of Mooney face stimuli (stylized, black-and-white images of faces without details) related to the six subscales of the Comprehensive Autistic Trait Inventory in young adults. In Experiment 1, regression analyses indicated that participants with poorer communication skills had lower task sensitivity when discriminating between face-present and face-absent images, whilst other autistic traits had no unique predictive value. Experiment 2 replicated these findings and additionally showed that autistic traits were linked to a reduced face inversion effect. Taken together, these results indicate autistic traits, especially communication difficulties, are associated with reduced configural processing of face stimuli. It follows that both reduced sensitivity for identifying upright faces amongst similar-looking distractors and reduced susceptibility to face inversion effects may be linked to relatively decreased reliance on configural processing of faces in autism. This study also reinforces the need to consider the different facets of autism independently.

Learning of the same task subserved by substantially different mechanisms between patients with body dysmorphic disorder and healthy individuals.

It has remained unclear whether individuals with psychiatric disorders involving altered visual processing employ similar neuronal mechanisms during perceptual learning of a visual task. We investigated this question by training patients with body dysmorphic disorder, a psychiatric disorder characterized by distressing or impairing preoccupation with nonexistent or slight defects in one's physical appearance, and healthy controls on a visual detection task for human faces with low spatial frequency components. Brain activation during task performance was measured with functional magnetic resonance imaging before the beginning and after the end of behavioral training. Both groups of participants improved performance on the trained task to a similar extent. However, neuronal changes in the fusiform face area were substantially different between groups such that activation for low spatial frequency faces in the right fusiform face area increased after training in body dysmorphic disorder patients but decreased in controls. Moreover, functional connectivity between left and right fusiform face area decreased after training in patients but increased in controls. Our results indicate that neuronal mechanisms involved in perceptual learning of a face detection task differ fundamentally between body dysmorphic disorder patients and controls. Such different neuronal mechanisms in body dysmorphic disorder patients might reflect the brain's adaptations to altered functions imposed by the psychiatric disorder.

A Multimodel Edge Computing Offloading Framework for Deep-Learning Application Based on Bayesian Optimization

With the rapid development of Internet of Things (IoT), data generated by IoT devices are also increasing exponentially. The edge computing has alleviated the problems of limited network and transmission delay when processing tasks of IoT devices in traditional cloud computing. And with the popularity of deep learning, more and more terminal devices are embedded with AI (Artificial Intelligence) processors for higher processing capability at the edge. However, the problems of Deep Learning task offloading in heterogeneous edge computing environment have not been fully investigated. In this paper, a multi-model edge computing offloading framework is proposed, using NVIDIA Jetson edge devices (Jetson TX2, Jetson Xavier NX, and Jetson Nano) and GeForce RTX GPU servers (RTX3080 and RTX2080) to simulate the edge computing environment, and make binary computational offloading decisions for face detection tasks. We also introduce a Bayesian Optimization algorithm, namely MTPE (Modified Tree-structured Parzen Estimator), to reduce the total cost of edge computation within a time slot including response time and energy consumption, and ensure the accuracy requirements of face detection. In addition, we employ the Lyapunov model to obtain the harvesting energy between time slots to keep the energy queue stable. Experiments reveal that MTPE algorithm can achieve the globally optimal solution in fewer iterations. The total cost of multi-model edge computing framework is reduced by an average of 17.94% compared to a single-model framework. In contrast to the Double Deep Q-Network (DDQN), our proposed algorithm can decrease the computational consumption by 23.01% for obtaining the offloading decision.

Face Detection Using a Capsule Network for Driver Monitoring Application

Bus driver distraction and cognitive load lead to higher accident risk. Driver distraction sources and complex physical and psychological effects must be recognized and analyzed in real-world driving conditions to reduce risk and enhance overall road safety. The implementation of a camera-based system utilizing computer vision for face recognition emerges as a highly viable and effective driver monitoring approach applicable in public transport. Reliable, accurate, and unnoticeable software solutions need to be developed to reach the appropriate robustness of the system. The reliability of data recording depends mainly on external factors, such as vibration, camera lens contamination, lighting conditions, and other optical performance degradations. The current study introduces Capsule Networks (CapsNets) for image processing and face detection tasks. The authors’ goal is to create a fast and accurate system compared to state-of-the-art Neural Network (NN) algorithms. Based on the seven tests completed, the authors’ solution outperformed the other networks in terms of performance degradation in six out of seven cases. The results show that the applied capsule-based solution performs well, and the degradation in efficiency is noticeably smaller than for the presented convolutional neural networks when adversarial attack methods are used. From an application standpoint, ensuring the security and effectiveness of an image-based driver monitoring system relies heavily on the mitigation of disruptive occurrences, commonly referred to as “image distractions,” which represent attacks on the neural network.

ADD: An automatic desensitization fisheye dataset for autonomous driving

Autonomous driving systems require many images for analyzing the surrounding environment. However, there is fewer data protection for private information among these captured images, such as pedestrian faces or vehicle license plates, which has become a significant issue. In this paper, in response to the call for data security laws and regulations and based on the advantages of large Field of View (FoV) of the fisheye camera, we build the first Autopilot Desensitization Dataset, called ADD, and formulate the first deep-learning-based image desensitization framework, to promote the study of image desensitization in autonomous driving scenarios. The compiled dataset consists of 650K images, including different face and vehicle license plate information captured by the surround-view fisheye camera. It covers various autonomous driving scenarios, including diverse facial characteristics and license plate colors. Then, we propose an efficient multitask desensitization network called DesCenterNet as a benchmark on the ADD dataset, which can perform face and vehicle license plate detection and desensitization tasks. Based on ADD, we further provide an evaluation criterion for desensitization performance, and extensive comparison experiments have verified the effectiveness and superiority of our method on image desensitization.

Face Recognition Using Raspberry PI for Mentally Retarded Person

The face recognition system described in this paper focuses on addressing the needs of individuals with intellectual disabilities. Intellectual disabilities often affect an individual's cognitive and adaptive functioning, making it challenging for them to perform everyday tasks independently. This system utilizes the capabilities of Raspberry Pi , a low cost single-board computer, to create an accessible and user-friendly solution for face recognition. The main objective of this system is to assist mentally challenged individuals in recognizing and identifying familiar faces. By utilizing face recognition technology, the system aims to enhance their independence, social interactions, and overall quality of life. The Raspberry Pi platform provides a flexible and affordable hardware foundation for implementing this system. The system leverages a pre-trained deep learning model for face recognition, such as OpenCV or dlib, to perform face detection and identification tasks. The model is trained on a dataset containing facial images of familiar individuals to ensure accurate recognition. Raspberry Pi's computational capabilities enable real-time processing of the captured video feed, facilitating prompt recognition and response.

A modified YOLOv5 helmet detection algorithm based on Swin Transformer

For the current stage of helmet detection in complex environments with low accuracy, missed detection and not easy to manage wearing, this paper proposes a YOLOv5 face helmet detection algorithm based on Swin Transformer improvement from the overall semantics of the image. In this paper, experiments are conducted using a self-built dataset to further enhance the performance of the model and improve the accuracy of face helmet detection through Mosaic data enhancement, label smoothing processing, adaptive weighted features combined with Wconcat module and the application of C3TR and C3STR modules to fuse multi-scale information, enhance the feature extraction capability of the network, and improve the generalization and robustness of the model with a self-built dataset . Experiments show that the improved YOLOv5 face helmet detection algorithm mAP based on Swin Transformer improves 5.7% compared with Faster RCNN, 6.1% compared with YOLOV3, 5.3% compared with YOLOV4, and 1.6% compared with the original algorithm. It performs well in helmet face detection tasks in complex environments, achieving real-time detection and higher accuracy, while reducing missed detections.

HeadPose-Softmax: Head pose adaptive curriculum learning loss for deep face recognition

Face recognition has been one of the most popular applications in the field of target detection. Currently, frontal faces can be easily detected, but multi-view face detection remains a difficult task because of various factors such as illumination, various poses, occlusions, and facial expressions. Margin-based loss functions are used to increase the feature margins between different classes, thus enhancing the discriminability of face recognition models, but the performance in face detection in complex scenes (e.g., high pitch angle face detection in surveillance environments) can be significantly degraded. Recently, the idea of a mining-based strategy to emphasize hard samples has been used to achieve good results in multi-view face detection. However, most of the existing methods do not explicitly emphasize samples based on their importance, resulting in the underutilization of hard samples. In this paper, we propose a curriculum learning loss function (HeadPose-Softmax) to classify the difficulty of a sample based on its facial pose, and embed the concept of curriculum learning into the loss function to implement a novel training strategy for deep face recognition. The loss function explicitly emphasizes the importance of the samples according to the different difficulty of each sample, which allows the model to make fuller use of hard samples, focus on learning pose invariant features, and improve the accuracy of the model in multi-view face detection tasks. Specifically, our HeadPose-Softmax dynamically adjusts the relative importance of the hard samples according to the pose angle of the face in the hard samples during the training phase. At each stage, different samples are assigned different importance according to their corresponding difficulty. Extensive experimental results under popular benchmarks show that our HeadPose-Softmax can enhance the accuracy of the model in multi-view face detection and outperform the state-of-the-art competitors.

General figure and face-specific closure ability: predictors of trait-autism?

Little research has examined the association between general figure closure speed, a stratum I ability within the Cattell-Horn-Carroll (CHC) model of intelligence, and face-specific closure ability; an important consideration, as some research suggests face processing is independent of other abilities. Additionally, figure closure ability (general and face) may correlate negatively with trait-autism, due to theorised autism-related difficulties in global processing. Therefore, in addition to developing a psychometrically robust short-form Mooney face detection task, we administered the Gestalt Figure Completion Test and the Autism Spectrum Quotient to a sample of 263 general community adults. We found convergent validity between face-specific and general figure closure ability (r = .44, 95%CI:[.30, .58]). Furthermore, based on a latent variable model, general figure closure ability was directly, and face-specific figure closure ability indirectly, negatively associated with trait-autism (i.e., nonverbal communication). We conclude that face detection ability evidences convergent validity with general figure closure ability and can be measured reliably in less than four minutes. Finally, a general, rather than face-specific, figure closure process may tap more directly into autism-like nonverbal communication ability, supporting the notion that global processing abilities of non-face objects may better reflect the requirements to ‘fill in the gaps’ in social contexts.

Optimizing Face Recognition Inference with a Collaborative Edge–Cloud Network

The rapid development of deep-learning-based edge artificial intelligence applications and their data-driven nature has led to several research issues. One key issue is the collaboration of the edge and cloud to optimize such applications by increasing inference speed and reducing latency. Some researchers have focused on simulations that verify that a collaborative edge-cloud network would be optimal, but the real-world implementation is not considered. Most researchers focus on the accuracy of the detection and recognition algorithm but not the inference speed in actual deployment. Others have implemented such networks with minimal pressure on the cloud node, thus defeating the purpose of an edge-cloud collaboration. In this study, we propose a method to increase inference speed and reduce latency by implementing a real-time face recognition system in which all face detection tasks are handled on the edge device and by forwarding cropped face images that are significantly smaller than the whole video frame, while face recognition tasks are processed at the cloud. In this system, both devices communicate using the TCP/IP protocol of wireless communication. Our experiment is executed using a Jetson Nano GPU board and a PC as the cloud. This framework is studied in terms of the frame-per-second (FPS) rate. We further compare our framework using two scenarios in which face detection and recognition tasks are deployed on the (1) edge and (2) cloud. The experimental results show that combining the edge and cloud is an effective way to accelerate the inferencing process because the maximum FPS achieved by the edge-cloud deployment was 1.91× more than the cloud deployment and 8.5× more than the edge deployment.

Intruder Detection System Using Image Processing

In the modern era, security and surveillance are important issues. Recent acts of theft have highlighted the urgent need for efficient video surveillance and on-the-spotnotification of ongoing thefts to house owners. A number of surveillance solutions are currently available on the market for the populace, such as CCTV cameras and digital video recorders (DVRs) that can record the unauthorized activities of a trespasser, however they demand excessive amounts of investment in terms of funds and timeand cannot distinguish between authorised and non-authorised users in most cases. Also the commonly available systems are often targeted towards medium sized franchises and are out of reach of common people. The task of face detection and the recognition of an intruder become very difficult when the intruder hides their face partially or fully. Hence the following system makes use of a program to perform motion detection by analyzing the relative differences between two consecutive frames. To facilitate this process the frames are further processed using the various image processing methods for ensuring the highest possible precision while distinguishing genuine motion from ‘noise’ disturbance

Evidence for a theta-band behavioural oscillation in rapid face detection.

Theories of rhythmic perception propose that perceptual sampling operates in a periodic way, with alternating moments of high and low responsiveness to sensory inputs. This rhythmic sampling is linked to neural oscillations and thought to produce fluctuations in behavioural outcomes. Previous studies have revealed theta- and alpha-band behavioural oscillations in low-level visual tasks and object categorization. However, less is known about fluctuations in face perception, for which the human brain has developed a highly specialized network. To investigate this, we ran an online study (N = 179) incorporating the dense sampling technique with a dual-target rapid serial visual presentation (RSVP) paradigm. In each trial, a stream of object images was presented at 30 Hz and participants were tasked with detecting whether or not there was a face image in the sequence. On some trials, one or two (identical) face images (the target) were embedded in each stream. On dual-target trials, the targets were separated by an interstimulus interval (ISI) that varied between 0 to 633 ms. The task was to indicate the presence of the target and its gender if present. Performance varied as a function of ISI, with a significant behavioural oscillation in the face detection task at 7.5Hz, driven mainly by the male target faces. This finding is consistent with a high theta-band-based fluctuation in visual processing. Such fluctuations might reflect rhythmic attentional sampling or, alternatively, feedback loops involved in updating top-down predictions.

An Efficient Multiscale Pyramid Attention Network for Face Detection in Surveillance Images

Although a large number of works have been done to explore efficient face detection in various scenes, practical face detection in unconstrained condition of varying lighting, pose, scale, and occlusion remains a challenging task. The primary limitation of existing solutions is that they are vulnerable to influence from the wild environment. Practical features extraction plays a crucial part in the face detection of low-quality images. Based on the EfficientNet, this paper builds a novel pyramid attention network to integrate multilevel features with rich context messages. Firstly, a context model is exploited to increase the receptive fields at the beginning of the network. Secondly, stacked pyramid feature attention modules and feature fusion simultaneously selectively integrate the contextual information and enable spatial details, thus enhancing the capacity to detect faces on hard images. In addition, hard samples augmentation of the training sets is conducted, which is beneficial for improving the accuracy. A thorough study on ablation verifies the effect of the proposed strategies. Moreover, extensive experiments on Wider Face and FDDB datasets, remarkably pushing the accuracy both of 96.3% (2%↑), demonstrate the performance of the proposed deep face detector which is superior and outperforms most of the preexisting methods. The method presented in this paper can perform the task of face detection in surveillance images well.

Face Detection Based on Dynamic Min Size and Dense Connection

Since the face detection technology is widely used, improving the detection speed and accuracy in face detection tasks has become a key challenge. Therefore, this paper takes the MTCNN model as the research object and makes improvements, the purpose of which is to optimize the detection speed and accuracy simultaneously. A dynamic min size algorithm is proposed. According to the size of the input image, it dynamically controls the minimum size of the face to be recognized by the model and reduces the number of iterations of the image pyramid, increasing the number of detection frames per second by 4[Formula: see text]fps. Then, the standard convolution structure in the P-Net and R-Net models is replaced by the depth-wise separable convolution, which effectively reduces the number of parameters and computation of the model. Meanwhile, an O-Net model with a densely connected structure is also developed. Our experiments with well-known public datasets have demonstrated that the proposed network structure can improve the detection frame rate. By reusing the features of different levels of the image, the recall rate and the precision rate of the MTCNN model on the validation set are increased by 2.39% and 1.65%, respectively.

Application Of The Viola Jones Algorithm On Face Detection

The face is the central organ for human expression, recognition, and communication. The face consists of four very important sense organs, namely the nose, eyes, ears, and tongue. The human brain has the privilege of recognizing someone based on their face. Face Detection is used to find out faces or nonfaces in an image and this section is the first step in the face detection process so its presence is very vital. Face detection task is very easy for humans, but this task is very complicated for computers because there are several stages that must be passed to detect faces. Where in this study the author designed and made an application for the application of the viola jones algorithm on face detection by discussing Python programming and using the Mysql database. The application of the viola jones algorithm on face detection goes well. Faces can be recognized by the application and display face owner information. Where the test is divided into 5 (five) including lighting testing with a test percentage level of 66.67% of detected faces, testing distance testing of 100% of faces detected at a distance of 1-100 cm, testing the number of faces there are 6 (six) results detected with 2 (two) faces, where faces must be approximately 10 cm apart so that faces can be detected and no results are detected on 3 (three) faces. So it can be concluded that the application cannot capture all faces detected by the camera. different face testing percentage test 100% the application can detect different people's faces with a single face.