Face Matching Research Articles

WebFace260M: A Benchmark for Million-Scale Deep Face Recognition.

Face benchmarks empower the research community to train and evaluate high-performance face recognition systems. In this paper, we contribute a new million-scale recognition benchmark, containing uncurated 4M identities/260M faces (WebFace260M) and cleaned 2M identities/42M faces (WebFace42M) training data, as well as an elaborately designed time-constrained evaluation protocol. First, we collect 4M name lists and download 260M faces from the Internet. Then, a Cleaning Automatically utilizing Self-Training (CAST) pipeline is devised to purify the tremendous WebFace260M, which is efficient and scalable. To the best of our knowledge, the cleaned WebFace42M is the largest public face recognition training set and we expect to close the data gap between academia and industry. Referring to practical deployments, Face Recognition Under Inference Time conStraint (FRUITS) protocol and a new test set with rich attributes are constructed. Besides, we gather a large-scale masked face sub-set for biometrics assessment under COVID-19. For a comprehensive evaluation of face matchers, three recognition tasks are performed under standard, masked and unbiased settings, respectively. Equipped with this benchmark, we delve into million-scale face recognition problems. A distributed framework is developed to train face recognition models efficiently without tampering with the performance. Enabled by WebFace42M, we reduce 40% failure rate on the challenging IJB-C set and rank 3rd among 430 entries on NIST-FRVT. Even 10% data (WebFace4M) shows superior performance compared with the public training sets. Furthermore, comprehensive baselines are established under the FRUITS-100/500/1000 milliseconds protocols. The proposed benchmark shows enormous potential on standard, masked and unbiased face recognition scenarios. Our WebFace260M website is https://www.face-benchmark.org.

Face matching and metacognition: investigating individual differences and a training intervention.

Although researchers have begun to consider metacognitive insight during face matching, little is known about the underlying mechanism. Here, I investigated whether objective ability, as well as self-assessed ability, were able to predict metacognitive performance, that is, the ability to differentiate correct and incorrect responses in terms of confidence. In addition, I considered whether a training intervention resulted in improvements to both face matching performance and metacognitive insight. In this experiment (N = 220), participants completed a face matching task, with either a diagnostic feature training course or a control course presented at the halfway point. In addition, a second face matching task, as well as a self-report questionnaire regarding ability, were completed to provide measures of objective and self-assessed ability respectively. Higher self-assessed ability with faces, as well as higher objective ability with face matching, predicted better metacognitive performance, i.e., greater confidence in correct, in comparison with incorrect, responses. This pattern of results was evident both when objective ability was measured through performance on the same task used to measure metacognitive insight and when a different task was used. Finally, the training intervention failed to produce improvements in face matching performance and showed no evidence of altering metacognitive ability. The current work begins to address the mechanism underlying individual differences in metacognitive insight during face matching. Although support was provided for a competence-based account, where better face matchers showed greater performance on the task and were more successful in monitoring their performance, further work might focus on decoupling task performance and competence in order to more conclusively explain why some people are more insightful than others.

Prepended Domain Transformer: Heterogeneous Face Recognition Without Bells and Whistles

Heterogeneous Face Recognition (HFR) refers to matching face images captured in different domains, such as thermal to visible images (VIS), sketches to visible images, near-infrared to visible, and so on. This is particularly useful in matching visible spectrum images to images captured from other modalities. Though highly useful, HFR is challenging because of the domain gap between the source and target domain. Often, large-scale paired heterogeneous face image datasets are absent, preventing training models specifically for the heterogeneous task. In this work, we propose a surprisingly simple, yet, very effective method for matching face images across different sensing modalities. The core idea of the proposed approach is to add a novel neural network block called Prepended Domain Transformer (PDT) in front of a pre-trained face recognition (FR) model to address the domain gap. Retraining this new block with few paired samples in a contrastive learning setup was enough to achieve state-of-the-art performance in many HFR benchmarks. The PDT blocks can be retrained for several source-target combinations using the proposed general framework. The proposed approach is architecture agnostic, meaning they can be added to any pre-trained FR models. Further, the approach is modular and the new block can be trained with a minimal set of paired samples, making it much easier for practical deployment. The source code and protocols will be made available publicly.

Identification of Facial Emotions Using Reinforcement model under Deep Learning

This paper addresses the identification of facial emotions using a reinforcement model under deep learning. Close-to-perception ability presents a more exhaustive recommendation on human-machine interaction (HMI). Because of the Transfer Self-training (TST), and the Representation Reinforcement Network (RRN), this study offers an active FER arrangement. Two modules are considered for depiction support arranging such as Surface Representation Reinforcement (SurRR) and Semantic Representation Reinforcement (SemaRR). SurRR highlights are detracting component communication centers in feature maps and match face attributes in different facets. Worldwide face settings are semantically sent in channel and dimensional facets of a piece. RRN has a limit concerning involved origin when the edges and computational complication are considerably belittled. Our technique was tried on informational indexes from CK+, RaFD, FERPLUS, and RAFDB, and it was viewed as 100 percent, 98.62 percent, 89.64 percent, and 88.72 percent, individually. Also, the early application exploration shows the way that our strategy can be utilized in HMI.

Cattle Facial Matching Recognition Algorithm Based on Multi-View Feature Fusion

In the process of collecting facial images of cattle in the field, some features of the collected images end up going missing due to the changeable posture of the cattle, which makes the recognition accuracy decrease or impossible to recognize. This paper verifies the practical effects of the classical matching algorithms ORB, SURF, and SIFT in bull face matching recognition. The experimental results show that the traditional matching algorithms perform poorly in terms of matching accuracy and matching time. In this paper, a new matching recognition model is constructed. The model inputs the target cattle facial data from different angles into the feature extraction channel, combined with GMS (grid-based motion statistics) algorithm and random sampling consistent algorithm, to achieve accurate recognition of individual cattle, and the recognition process is simple and fast. The recognition accuracy of the model was 85.56% for the Holstein cow face dataset, 82.58% for the Simmental beef cattle, and 80.73% for the mixed Holstein and Simmental beef cattle dataset. The recognition model constructed in the study can achieve individual recognition of cattle in complex environments, has good robustness to matching data, and can effectively reduce the effects of data angle changes and partial features missing in cattle facial recognition.

Facial comparison behaviour of forensic facial examiners

AbstractFacial examiners make visual comparisons of face images to establish the identities of persons in police investigations. This study utilised eye‐tracking and an individual differences approach to investigate whether these experts exhibit specialist viewing behaviours during identification, by comparing facial examiners with forensic fingerprint analysts and untrained novices across three tasks. These comprised of face matching under unlimited (Experiment 1) and time‐restricted viewing (Experiment 2), and with a feature‐comparison protocol derived from examiner casework procedures (Experiment 3). Facial examiners exhibited individual differences in facial comparison accuracy and did not consistently outperform fingerprint analysts and novices. Their behaviour was also marked by similarities to the comparison groups in terms of how faces were viewed, as evidenced from eye movements, and how faces were perceived, based on the made feature judgements and identification decisions. These findings further understanding of how facial comparisons are performed and clarify the nature of examiner expertise.

Contributions of face processing, social anhedonia and mentalizing to the expression of social autistic-like traits.

Quantitative autistic-like traits (QATs) are a constellation of traits that mirror those of clinical autism and are thought to share the same mechanisms as the condition. There is great interest in identifying the genetic and neurobiological basis of QATs, but progress is hindered by the composite nature of these clinically based constructs. Social QATs are defined according to the diagnostic criteria for autism, comprising multiple potential neural mechanisms that may contribute to varying degrees. The objective of this study was to decompose social QATs into more specific constructs, in line with the Research Domain Criteria (RDoC). We chose constructs with trait-like properties and known or suggested significance for autistic social function: (1) social anhedonia, (2) prosopagnosia (face blindness), and (3) mentalizing (attributing mental states to images of eyes). We hypothesized that these constructs may all contribute to observed variance in social QATs. We recruited 148 adults with a broad range of QATs (mean age 37.9 years, range 18-69; 50% female; 5.4% autistic) to an experimental behavioral study conducted online. We estimated social QATs using the social factor of the Comprehensive Autistic Traits Inventory. We used the Oxford Face Matching Task and the Reading the Mind in the Eyes Test to measure face matching ability and mentalizing, respectively. Social anhedonia traits were measured with the Anticipatory and Consummatory Interpersonal Pleasure Scale, and prosopagnosic traits with the 20-item Prosopagnosia Index. A combination of frequentist and Bayesian statistics was used to test the social constructs as predictors of social QATs. We found that social anhedonic traits, prosopagnosic traits, and face matching performance were likely predictors of social QATs, whereas mentalizing showed limited contribution. The findings support prosopagnosic and anhedonic traits, but not mentalizing deficits, as dimensional predictors of individual differences in social function across the autistic spectrum. Further, the study strongly suggests that social reward systems and face processing networks play significant and independent roles in autistic-like social function.

Face Recognition Based Attendance System Using OpenCV Python

When it comes to the most productive image processing application, face recognition considered to be the most reliable one and its role in technical field is incredible. Attendance marking system using face recognition is a procedure of marking attendance of student via matching face of a student with their stored biometric facial measurements. This system is developed to remove the pen paper marking attendance to save time, energy and increase accuracy. What method we are using nowadays to mark attendance is tedious and time-energy consuming. Attendance records can be easily manipulated because of lenient security but this system store attendance record in excel sheet in database of computer which is more secure than normal method of marking attendance. Teachers won't be carrying attendance record with them daily; extra burden will be reduced. The system will be tested under various parameter like illumination, variation of the distance between student and camera, head movement etc. After all this testing, accuracy and productivity of system can be finalized. The proposed system provides efficient way of marking attendance. This system can be used mark attendance with mask too. Also, multiple student attendance can be marked, hence reducing time and efforts.

Individuals with developmental prosopagnosia show independent impairments in face perception, face memory and face matching.

Individuals with developmental prosopagnosia (DP) all exhibit impairments in face memory, but the specificity of these face memory impairments is debated. One problem is that standard behavioural tasks are not able to provide independent measurement of face perception, face memory, and face matching (the decision process required to judge whether two instances of a face are of the same individual or different individuals). The present study utilised a new test of face matching, the Oxford Face Matching Test (OFMT), and a novel analysis strategy to derive these independent indices. Twenty-nine individuals with DP and the same number of matched neurotypical controls completed the OFMT, the Glasgow Face Matching Test, and the Cambridge Face Memory Test. Results revealed individuals with DP exhibit impairments in face perception, face memory and face matching. Collectively, these results suggest that face processing impairments in DP are more comprehensive than has previously been suggested.

High-frequency transcranial random noise stimulation enhances unfamiliar face matching of high resolution and pixelated faces

Face identification is useful for social interactions and its impairment can lead to severe social and mental problems. This ability is also remarkably important in applied settings, including eyewitness identification and ID verification. Several studies have demonstrated the potential of Transcranial Random Noise Stimulation (tRNS) to enhance different cognitive skills. However, research has produced inconclusive results about the effectiveness of tRNS to improve face identification. The present study aims to further explore the effect of tRNS on face identification using an unfamiliar face matching task. Observers firstly received either high-frequency bilateral tRNS or sham stimulation for 20 min. The stimulation targeted occipitotemporal areas, which have been previously involved in face processing. In a subsequent stage, observers were asked to perform an unfamiliar face matching task consisting of unaltered and pixelated face pictures. Compared to the sham stimulation group, the high-frequency tRNS group showed better unfamiliar face matching performance with both unaltered and pixelated faces. Our results show that a single high-frequency tRNS session might suffice to improve face identification abilities. These results have important consequences for the treatment of face recognition disorders, and potential applications in those scenarios whereby the identification of faces is primordial.

Unfamiliar faces might as well be another species: Evidence from a face matching task with human and monkey faces

ABSTRACT Humans are good at recognizing familiar faces, but are more error-prone at recognizing an unfamiliar person across different images. It has been suggested that familiar and unfamiliar faces are processed qualitatively differently. But are unfamiliar faces at least processed differently from monkey faces? Here we tested 366 volunteers on a face matching test – two images presented side-by-side with participants judging whether the images show the same identity or two different identities – comparing performance with familiar and unfamiliar human faces, and monkey faces. The results showed that performance was most accurate for familiar faces, and was above chance for monkey faces. Although accuracy was higher for unfamiliar humans than monkeys on different identity trials, there was no unfamiliar human advantage over monkeys on same identity trials. The results give new insights into unfamiliar face processing, showing that in some ways unfamiliar faces might as well be another species.

COVID-19 masks increase the influence of face recognition algorithm decisions on human decisions in unfamiliar face matching.

Face masks, recently adopted to reduce the spread of COVID-19, have had the unintended consequence of increasing the difficulty of face recognition. In security applications, face recognition algorithms are used to identify individuals and present results for human review. This combination of human and algorithm capabilities, known as human-algorithm teaming, is intended to improve total system performance. However, prior work has shown that human judgments of face pair similarity-confidence can be biased by an algorithm's decision even in the case of an error by that algorithm. This can reduce team effectiveness, particularly for difficult face pairs. We conducted two studies to examine whether face masks, now routinely present in security applications, impact the degree to which this cognitive bias is experienced by humans. We first compared the influence of algorithm's decisions on human similarity-confidence ratings in the presence and absence of face masks and found that face masks more than doubled the influence of algorithm decisions on human similarity-confidence ratings. We then investigated if this increase in cognitive bias was dependent on perceived algorithm accuracy by also presenting algorithm accuracy rates in the presence of face masks. We found that making humans aware of the potential for algorithm errors mitigated the increase in cognitive bias due to face masks. Our findings suggest that humans reviewing face recognition algorithm decisions should be made aware of the potential for algorithm errors to improve human-algorithm team performance.

Perceptual similarity modulates effects of learning from variability on face recognition

Face recognition is a challenging classification task that humans perform effortlessly for familiar faces. Recent studies have emphasized the importance of exposure to high variability appearances of the same identity to perform this task. However, these studies did not explicitly measure the perceptual similarity between the learned images and the images presented at test, which may account for the advantage of learning from high variability. Particularly, randomly selected test images are more likely to be perceptually similar to learned high variability images, and dissimilar to learned low variability images. Here we dissociated effects of learning from variability and study-test perceptual similarity, by collecting human similarity ratings for the study and test images. Using these measures, we independently manipulated the variability between the learning images and their perceptual similarity to the test images. Different groups of participants learned face identities from a low or high variability set of images. The learning phase was followed by a face matching test (Experiment 1) or a face recognition task (Experiment 2) that presented novel images of the learned identities that were perceptually dissimilar or similar to the learned images. Results of both experiments show that perceptual similarity between study and test, rather than image variability at learning per se, predicts face recognition. We conclude that learning from high variability improves face recognition for perceptually similar but not for perceptually dissimilar images. These findings may not be specific to faces and should be similarly evaluated for other domains.

Forensic facial examiners versus super-recognizers: Evaluating behavior beyond accuracy.

We evaluated the detailed, behavioral properties of face matching performance in two specialist groups: forensic facial examiners and super-recognizers. Both groups compare faces to determine identity with high accuracy and outperform the general population. Typically, facial examiners are highly trained; super-recognizers rely on natural ability. We found distinct behaviors between these two groups. Examiners used the full 7-point identity judgment scale (-3: "different"; +3: "same"). Super-recognizers' judgments clustered toward highly confident decisions. Examiners' judgments for same- and different-identities were symmetric across the scale midpoint (0); super-recognizers' judgments were not. Examiners showed higher identity judgment agreement than super-recognizers. Despite these qualitative differences, both groups showed insight into their own accuracy: more confident people and those who rated the task to be easier tended to be more accurate. Altogether, we show to better understand and interpret judgments according to the nature of someone's facial expertise, evaluations should assess more than accuracy.

Application of Face Recognition in E-commerce Security Authentication in the Era of Big Data

Although Internet technology brings invaluable benefits to all walks of life, the security of network and information is becoming more and more prominent. The leakage of personnel information caused by Internet security incidents causes irreparable harm and loss to individuals or enterprises. E-commerce is a virtual transaction mode based on Internet technology. Its security requirements for transactions are more stringent than those for traditional transaction modes. Traditional identity authentication technology can no longer meet its security needs. People urgently need a reliable identity authentication system, meaning to ensure the security of e-commerce. According to the actual application scenarios of the algorithm in e-commerce, this paper, based on the research on face recognition technology, focuses on the in-depth research on face detection technology under the background of big data in order to introduce face recognition technology in this paper. This paper proposes a feature-based face matching algorithm. The face image is preprocessed to improve the accuracy of real-time face detection and reduce the false detection rate. Based on the research of traditional facial feature extraction technology, a virtual sample set that can effectively support a traditional facial feature extraction algorithm is constructed to solve the problem of insufficient training samples in practical applications. The experimental results showed that the accuracy of the method in this paper can reach up to 79.5% and that the minimum time consumption is only 0.142 s. Compared with the traditional method, the accuracy rate is higher and the time consumption is shorter.

Effect of surgical mask on fMRI signals during task and rest

Wearing a face mask has become essential to contain the spread of COVID-19 and has become mandatory when collecting fMRI data at most research institutions. Here, we investigate the effects of wearing a surgical mask on fMRI data in n = 37 healthy participants. Activations during finger tapping, emotional face matching, working memory tasks, and rest were examined. Preliminary fMRI analyses show that despite the different mask states, resting-state signals and task activations were relatively similar. Resting-state functional connectivity showed negligible attenuation patterns in mask-on compared with mask-off. Task-based ROI analysis also demonstrated no significant difference between the two mask states under each contrast investigated. Notwithstanding the overall insignificant effects, these results indicate that wearing a face mask during fMRI has little to no significant effect on resting-state and task activations.

Gender-Adversarial Networks for Face Privacy Preserving

Privacy concerns over face recognition systems have attracted extensive attention in various fields. For gender privacy-preserving work, there are two key challenges: 1) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">privacy</i> , i.e., confusing gender classifiers and 2) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">utility</i> , i.e., maintaining its face verification performance. To address both issues, this article develops a novel gender-adversarial network, referred to as Gender-AN, to impart gender privacy to face images. Gender-AN employs an attribute-independent encoder–decoder GAN-based network to perturb the input face image, training with the assistance of the proper facial attributes. The perturbed image is then able to obfuscate gender classifiers while maintaining identity discriminability. To optimize the generator, a multitask-based loss function is utilized, which includes attribute manipulation loss, face matcher loss, adversarial loss, and reconstruction loss functions. This optimization facilitates our model to achieve the generalization, verification preserve, and natural appearance, simultaneously. Extensive experiments confirm the effectiveness of the proposed model in enhancing gender privacy and preserving face verification utility.

Reading emotional faces in deaf and hard-of-hearing and typically hearing children.

Reading emotions from other people's facial expressions is an important skill that guides social interactions. With limited auditory input and atypical emotion socialization, deaf and hard-of-hearing (DHH) children may develop atypical processing patterns when reading emotional faces. The current study aimed at understanding whether and how DHH and typically hearing (TH) children differed at 3 emotion processing levels: gaze patterns, physiological arousal, and interpretation. Fifty-five DHH children and 72 TH children completed an emotional face matching task in which they were presented with happy, angry, fearful, and emotionally neutral faces. During the task participants' eye gazes and pupil diameter were measured by an eye-tracking device. The DHH and TH children both paid most attention to the eye region when reading emotional faces. Yet, a contrast between happy faces and nonhappy faces was observed in physiological arousal and interpretation tendency in the DHH children only: Nonhappy facial expressions were more arousing and were confused more often than happy expressions, which may reflect the DHH children being less experienced in processing nonhappy expressions due to limited access to the social environment. The results highlighted the importance of looking into the qualitative differences between typical and atypical development. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Telepathic Improvisation (2017) by Pauline Boudry and Renate Lorenz (Boudry/Lorenz)

Reviewed by: Telepathic Improvisation (2017) by Pauline Boudry and Renate Lorenz (Boudry/Lorenz) Farrah O’Shea TELEPATHIC IMPROVISATION (2017). By Pauline Boudry and Renate Lorenz (Boudry/Lorenz). Witch Hunt, Hammer Museum and the Institute of Contemporary Art, Los Angeles. October 22, 2021. Entering a small, unlit room with black walls and carpet, audience members hovered around the edges of a short black stage at the Hammer Museum to view the film installation, Telepathic Improvisation (2017) by artist duo Boudry/Lorenz, known for installations that reimagine normative approaches to historical narratives and spectatorship. With its own dimly lit stage, the world of the film mimics the dark walls and floors of the installation space, thinning the boundary between the audience and the film and setting the stage for an experiment in telepathy. To begin the film, which is looped, a performer (Marwa Arsanios) dressed in a vermilion jumpsuit walks to the center of the stage, which is bare and set to look like the backstage of a production. Arsanios—an artist, filmmaker, and researcher whose work concerns gender relations, collectivism, urbanism, and industrialization—is framed by a line of downstage lights, a guitar, and a white motorized unit. Looking into the camera, she addresses the installation’s audience: “You are attempting inter-group, or interstellar telepathic transmission, following the 1974 score, Telepathic Improvisation by Pauline Oliveros.” While she is speaking, performance artist MPA, whose work explores how the social and the political impact the body; Ginger Brooks Takahashi, who creates collaborative project-based feminist and queer works; and Werner Hirsch, a freelance artist, dancer, and drag performer (sometimes also known as Prince Greenhorn or Antonia Baehr) walk on set, dressed in shades of red and white, with Arsanios and MPA in jumpsuits. Audience members are instructed to close their eyes and telepathically send actions to the performers, which include the four human beings as well as the guitar, lights, and motorized units. When the performers receive the action, they enact it. To illustrate that actions are being sent to and received by nonhuman subjects, onscreen stage lights flicker [End Page 370] as though receiving messages, and two motorized platforms suggest independence, at one point pushing a performer onto the screen. As audience members recognize the action they sent onstage, they are to raise an arm, completing the exchange. With this choreography, present-day audience members interact with performers filmed five years ago, activating the installation through a reification of the film’s promise of telepathic exchange. Click for larger view View full resolution Witch Hunt, installation view, Hammer Museum, Los Angeles, October 10, 2021–January 9, 2022: Marwa Arsanios, Werner Hirsch, MPA, and Ginger Brooks Takahashi in Telepathic Improvisation (2017). (Photo: Jeff McClane.) On several occasions during the piece’s run at the Hammer, although not when I attended, MPA appeared in person wearing a red jumpsuit and matching face mask. This visual reference to the COVID-19 pandemic contemporized the film, further connecting it to the installation audience. Moving around, sitting, or lying on a short lighted stage, MPA interacted with the film and a new sound-scape written by Rashad Becker for the Hammer installation. In MPA’s absence, the stage was both a reminder of past performances and those yet to happen. Somehow incomplete, the stage served to connect to the visual, darkened aesthetic of the film while suggesting a reality that was partial and not yet realized. Thus telepathy was construed as an exercise that invites engagement with the past meant to inspire action in the future. The stasis of the present moment, indicated by the vacant, darkened stage, served as a call to action, a reminder of political work yet to be done. In the 1974 performance score from which Telepathic Improvisation takes its name and cue, the late composer/performer Pauline Oliveros deconstructs the boundaries between audience and performer. Through the collective practice of group listening, what she calls “deep listening,” a practice meant to connect its participants to one another and the environment, Oliveros asks the audience to imagine sounds and manifest them as part of a group. Boudry/Lorenz take this practice a step further, engaging multiple dimensions by inviting an intergroup, interstellar...

Image variability and face matching.

This study investigates whether variability in perceived trait judgements disrupts our ability to match unfamiliar faces. In this preregistered study, 174 participants completed a face matching task where they were asked to indicate whether two ambient face images belonged to the same person or different people (17,748 total data points). Participants completed 51 match trials consisting of images of the same person that differed substantially on one trait (either trustworthiness, dominance or attractiveness) with minimal differences in the alternate traits. Participants also completed 51 mismatch trials which contained two photos of similar-looking individuals. We hypothesised that participants would make more errors on match trials when images differed in terms of attractiveness ratings than when they differed on trustworthiness or dominance. Contrary to expectations, images that differed in terms of attractiveness were matched most accurately, and there was no relationship between the extent of differences in attractiveness ratings and accuracy. There was some evidence that differences in perceived dominance and, to a lesser extent, trustworthiness were associated with lower face matching performance. However, these relationships were not significant when alternate traits were accounted for. The findings of our study suggest that face matching performance is largely robust against variation in trait judgements.