Synthetic Face Research Articles

Contributions of individual face features to face discrimination

Faces are highly complex stimuli that contain a host of information. Such complexity poses the following questions: (a) do observers exhibit preferences for specific information? (b) how does sensitivity to individual face parts compare? These questions were addressed by quantifying sensitivity to different face features. Discrimination thresholds were determined for synthetic faces under the following conditions: (i) ‘full face’: all face features visible; (ii) ‘isolated feature’: single feature presented in isolation; (iii) ‘embedded feature’: all features visible, but only one feature modified. Mean threshold elevations for isolated features, relative to full-faces, were 0.84x, 1.08, 2.12, 3.34, 4.07 and 4.47 for head-shape, hairline, nose, mouth, eyes and eyebrows respectively. Hence, when two full faces can be discriminated at threshold, the difference between the eyes is about four times less than what is required when discriminating between isolated eyes. In all cases, sensitivity was higher when features were presented in isolation than when they were embedded within a face context (threshold elevations of 0.94x, 1.74, 2.67, 2.90, 5.94 and 9.94). This reveals a specific pattern of sensitivity to face information. Observers are between two and four times more sensitive to external than internal features. The pattern for internal features (higher sensitivity for the nose, compared to mouth, eyes and eyebrows) is consistent with lower sensitivity for those parts affected by facial dynamics (e.g. facial expressions). That isolated features are easier to discriminate than embedded features supports a holistic face processing mechanism which impedes extraction of information about individual features from full faces.

Dense 3D Face Correspondence.

We present an algorithm that automatically establishes dense correspondences between a large number of 3D faces. Starting from automatically detected sparse correspondences on the outer boundary of 3D faces, the algorithm triangulates existing correspondences and expands them iteratively by matching points of distinctive surface curvature along the triangle edges. After exhausting keypoint matches, further correspondences are established by generating evenly distributed points within triangles by evolving level set geodesic curves from the centroids of large triangles. A deformable model (K3DM) is constructed from the dense corresponded faces and an algorithm is proposed for morphing the K3DM to fit unseen faces. This algorithm iterates between rigid alignment of an unseen face followed by regularized morphing of the deformable model. We have extensively evaluated the proposed algorithms on synthetic data and real 3D faces from the FRGCv2, Bosphorus, BU3DFE and UND Ear databases using quantitative and qualitative benchmarks. Our algorithm achieved dense correspondences with a mean localisation error of 1.28 mm on synthetic faces and detected 14 anthropometric landmarks on unseen real faces from the FRGCv2 database with 3 mm precision. Furthermore, our deformable model fitting algorithm achieved 98.5 percent face recognition accuracy on the FRGCv2 and 98.6 percent on Bosphorus database. Our dense model is also able to generalize to unseen datasets.

Nonlinear dimensionality reduction methods for synthetic biology biobricks\u2019 visualization

BackgroundVisualizing data by dimensionality reduction is an important strategy in Bioinformatics, which could help to discover hidden data properties and detect data quality issues, e.g. data noise, inappropriately labeled data, etc. As crowdsourcing-based synthetic biology databases face similar data quality issues, we propose to visualize biobricks to tackle them. However, existing dimensionality reduction methods could not be directly applied on biobricks datasets. Hereby, we use normalized edit distance to enhance dimensionality reduction methods, including Isomap and Laplacian Eigenmaps.ResultsBy extracting biobricks from synthetic biology database Registry of Standard Biological Parts, six combinations of various types of biobricks are tested. The visualization graphs illustrate discriminated biobricks and inappropriately labeled biobricks. Clustering algorithm K-means is adopted to quantify the reduction results. The average clustering accuracy for Isomap and Laplacian Eigenmaps are 0.857 and 0.844, respectively. Besides, Laplacian Eigenmaps is 5 times faster than Isomap, and its visualization graph is more concentrated to discriminate biobricks.ConclusionsBy combining normalized edit distance with Isomap and Laplacian Eigenmaps, synthetic biology biobircks are successfully visualized in two dimensional space. Various types of biobricks could be discriminated and inappropriately labeled biobricks could be determined, which could help to assess crowdsourcing-based synthetic biology databases’ quality, and make biobricks selection.

Position selectivity in face-sensitive visual cortex to facial and nonfacial stimuli: an fMRI study.

BackgroundEvidence for position sensitivity in object‐selective visual areas has been building. On one hand, most of the relevant studies have utilized stimuli for which the areas are optimally selective and examine small sections of cortex. On the other hand, visual field maps established with nonspecific stimuli have been found in increasingly large areas of visual cortex, though generally not in areas primarily responsive to faces.Methods fMRI was used to study the position sensitivity of the occipital face area (OFA) and the fusiform face area (FFA) to both standard rotating wedge retinotopic mapping stimuli and quadrant presentations of synthetic facial stimuli. Analysis methods utilized were both typical, that is, mean univariate BOLD signals and multivoxel pattern analysis (MVPA), and novel, that is, distribution of voxels to pattern classifiers and use of responses to nonfacial retinotopic mapping stimuli to classify responses to facial stimuli.ResultsPolar angle sensitivity was exhibited to standard retinotopic mapping stimuli with a stronger contralateral bias in OFA than in FFA, a stronger bias toward the vertical meridian in FFA than in OFA, and a bias across both areas toward the inferior visual field. Contralateral hemispheric lateralization of both areas was again shown using synthetic face stimuli based on univariate BOLD signals, MVPA, and the biased contribution of voxels toward multivariate classifiers discriminating the contralateral visual field. Classifiers based on polar angle responsivity were used to classify the patterns of activation above chance levels to face stimuli in the OFA but not in the FFA.ConclusionsBoth the OFA and FFA exhibit quadrant sensitivity to face stimuli, though the OFA exhibits greater position responsivity across stimuli than the FFA and includes overlap in the response pattern to the disparate stimulus types. Such biases are consistent with varying position sensitivity along different surfaces of occipito‐temporal cortex.

Gaze direction detection in autism spectrum disorder.

Detecting where our partners direct their gaze is an important aspect of social interaction. An atypical gaze processing has been reported in autism. However, it remains controversial whether children and adults with autism spectrum disorder interpret indirect gaze direction with typical accuracy. This study investigated whether the detection of gaze direction toward an object is less accurate in autism spectrum disorder. Individuals with autism spectrum disorder (n = 33) and intelligence quotients-matched and age-matched controls (n = 38) were asked to watch a series of synthetic faces looking at objects, and decide which of two objects was looked at. The angle formed by the two possible targets and the face varied following an adaptive procedure, in order to determine individual thresholds. We found that gaze direction detection was less accurate in autism spectrum disorder than in control participants. Our results suggest that the precision of gaze following may be one of the altered processes underlying social interaction difficulties in autism spectrum disorder.

A Novel Approach to Mugshot Based Arbitrary View Face Recognition

National Key Laboratory of Fundamental Science on Synthetic Vision,College of Computer Science, Sichuan University, Chengdu 610065, P. R. China(Received November 18, 2015 : revised February 11, 2016 : accepted March 4, 2016)Mugshot face images, routinely collected by police, usually contain both frontal and profile views. Existing automated face recognition methods exploited mugshot databases by enlarging the gallery with synthetic multi-view face images generated from the mugshot face images. This paper, instead, proposes to match the query arbitrary view face image directly to the enrolled frontal and profile face images. During matching, the 3D face shape model reconstructed from the mugshot face images is used to establish corresponding semantic parts between query and gallery face images, based on which comparison is done. The final recognition result is obtained by fusing the matching results with frontal and profile face images. Compared with previous methods, the proposed method better utilizes mugshot databases without using synthetic face images that may have artifacts. Its effectiveness has been demonstrated on the Color FERET and CMU PIE databases. Keywords : Mugshot-based face recognition, Arbitrary view face recognition, Three-dimensional face reconstructionOCIS codes : (100.5010) Pattern recognition; (100.0100) Image processing; (100.3008) Image recognition, algorithms and filters; (100.6890) Three-dimensional image processing

Intact priors for gaze direction in adults with high-functioning autism spectrum conditions.

BackgroundAutism Spectrum Conditions (ASC) are associated with a range of perceptual atypicalities, including abnormalities in gaze processing. Pellicano and Burr (Trends Cogn Sci 16(10):504-10, 2012) have argued that these atypicalities might be explained within a Bayesian framework, in which perception represents the combination of sensory information with prior knowledge. They propose that the Bayesian priors of individuals with ASC might be attenuated, such that their perception is less reliant on prior knowledge than neurotypical individuals. An important tenet of Bayesian decision theory is that increased uncertainty about incoming sensory information will lead to a greater influence of the prior on perception. Consistent with this, Mareschal et al. (Curr Biol 23(8):717-21, 2013) showed that when noise is added to the eyes of a face (increasing uncertainty about gaze direction), gaze is more likely to be perceived as direct.MethodsWe adopted the same paradigm as Mareschal et al. to determine whether the influence of a prior on gaze perception is reduced in neurotypical participants with high numbers of autistic traits (experiment 1) and in individuals with a clinical diagnosis of ASC (experiment 2). Participants were presented with synthetic faces and asked to make a judgement about the relative gaze directions of the faces. Uncertainty about gaze direction was manipulated by adding noise to the eyes of a face.ResultsConsistent with previous work, in both experiment 1 and experiment 2, participants showed a bias towards perceiving gaze as direct under conditions of uncertainty. However, there was no evidence that the magnitude of this bias was reduced either in the ASC group or in neurotypical controls with a high number of autistic traits.ConclusionsOur findings challenge the attenuated priors theory of perception in ASC (Trends Cogn Sci 16(10):504-10, 2012) and related proposals (Trends Cogn Sci 17(1):1, 2013, Front Hum Neurosci 8:302, 2014), and suggest priors for gaze direction are intact in high-functioning ASC.

The Caledonian face test: A new test of face discrimination

This study aimed to develop a clinical test of face perception which is applicable to a wide range of patients and can capture normal variability. The Caledonian face test utilises synthetic faces which combine simplicity with sufficient realism to permit individual identification. Face discrimination thresholds (i.e. minimum difference between faces required for accurate discrimination) were determined in an “odd-one-out” task. The difference between faces was controlled by an adaptive QUEST procedure. A broad range of face discrimination sensitivity was determined from a group (N=52) of young adults (mean 5.75%; SD 1.18; range 3.33–8.84%). The test is fast (3–4min), repeatable (test–re-test r2=0.795) and demonstrates a significant inversion effect. The potential to identify impairments of face discrimination was evaluated by testing LM who reported a lifelong difficulty with face perception. While LM’s impairment for two established face tests was close to the criterion for significance (Z-scores of −2.20 and −2.27) for the Caledonian face test, her Z-score was −7.26, implying a more than threefold higher sensitivity. The new face test provides a quantifiable and repeatable assessment of face discrimination ability. The enhanced sensitivity suggests that the Caledonian face test may be capable of detecting more subtle impairments of face perception than available tests.

Masking with faces in central visual field under a variety of temporal schedules

With a few exceptions, previous studies have explored masking using either a backward mask or a common onset trailing mask, but not both. In a series of experiments, we demonstrate the use of faces in central visual field as a viable method to study the relationship between these two types of mask schedule. We tested observers in a two alternative forced choice face identification task, where both target and mask comprised synthetic faces, and show that a simple model can successfully predict masking across a variety of masking schedules ranging from a backward mask to a common onset trailing mask and a number of intermediate variations. Our data are well accounted for by a window of sensitivity to mask interference that is centered at around 100ms.

A robust framework for tracking simultaneously rigid and non-rigid face using synthesized data

This paper presents a robust framework for simultaneously tracking rigid pose and non-rigid animation of a single face with a monocular camera. Our proposed method consists of two phases: training and tracking. In the training phase, using automatically detected landmarks and the three-dimensional face model Candide-3, we built a cohort of synthetic face examples with a large range of the three axial rotations. The representation of a face’s appearance is a set of local patches of landmarks that are characterized by Scale Invariant Feature Transform (SIFT) descriptors. In the tracking phase, we propose an original approach combining geometric and appearance models. The purpose of the geometric model is to provide a SIFT baseline matching between the current frame and an adaptive set of keyframes for rigid parameter estimation. The appearance model uses nearest synthetic examples of the training set to re-estimate rigid and non-rigid parameters. We found a tracking capability up to 90° of vertical axial rotation, and our method is robust even in the presence of fast movements, illumination changes and tracking losses. Numerical results on the rigid and non-rigid parameter sets are reported using several annotated public databases. Compared to other published algorithms, our method provides an excellent compromise between rigid and non-rigid parameter accuracies. The approach has some potential, providing good pose estimation (average error less than 4° on the Boston University Face Tracking dataset) and landmark tracking precision (6.3 pixel error compared to 6.8 of one of state-of-the-art methods on Talking Face video).

Artificial faces are harder to remember

Observers interact with artificial faces in a range of different settings and in many cases must remember and identify computer-generated faces. In general, however, most adults have heavily biased experience favoring real faces over synthetic faces. It is well known that face recognition abilities are affected by experience such that faces belonging to “out-groups” defined by race or age are more poorly remembered and harder to discriminate from one another than faces belonging to the “in-group”. Here, we examine the extent to which artificial faces form an “out-group” in this sense when other perceptual categories are matched. We rendered synthetic faces using photographs of real human faces and compared performance in a memory task and a discrimination task across real and artificial versions of the same faces. We found that real faces were easier to remember, but only slightly more discriminable than artificial faces. Artificial faces were also equally susceptible to the well-known face inversion effect, suggesting that while these patterns are still processed by the human visual system in a face-like manner, artificial appearance does compromise the efficiency of face processing.

Morphometric facial analysis: a methodology to create lateral facial images

Currently, the majority of research in facial analysis using computational morphing methods focuses exclusively on analysis of frontal facial projections. Lateral facial morphing analysis has not been extensively investigated, and landmark features critical to specify registry points are unknown. This study aims to (1) determine the quantity of registry points (RP) required to create realistic lateral faces and (2) determine key facial registry point landmarks required to create synthetic lateral faces. 36 synthetic lateral faces with a 50 to 250 RP were created to determine the ideal quantity of RP to create a realistic lateral image; ear, eyebrow, eye, nose, lips, hairline, facial outline, and overall outline were evaluated by an expert panel of seven evaluators using a 1 to 5 point Lickert scale rating system. ANOVA single-variable analyses revealed significant differences when comparing templates of 200 and 250 RP with 50 and 100 RP templates (p < 0.05). Furthermore, analysis of all key landmark areas of the face indicated significant differences between different registry points except for 200 and 250 registry point markers. Kruskal-Wallis statistical analysis revealed the landmarks varied significantly from 50 to 200 RP,but had no significance with 200 and 250 RP. The most ideal quantity of RP used for the creation of realistic lateral faces was in the range of 200 RP. Defined lateral facial registry point landmarks generated successful realistic faces.

Big Five personality traits contribute to prosocial responses to others’ pain

ObjectivesTwo studies examined whether observers’ personality traits contribute to prosocial responses to others’ facial expression of pain. Experiment 1 examined the personality traits that could account for observers’ variability in estimating others’ pain intensity. Experiment 2 questioned to what extent the contribution of personality traits on inclination to help people in pain depend on observers’ beliefs about pain’ characteristics. Method59 (experiment 1) and 76 (experiment 2) participants observed to 3-D realistic synthetic face movements mobilizing action units of pain, in order to estimate others’ pain. In experiment 2, painful localizations (e.g., chest, hand) were also manipulated. In each experiment, Big Five personality traits were assessed. ResultsExperiment 1 revealed that agreeableness and conscientiousness contributed to observers’ pain estimates across the increase of facial expression intensity. Experiment 2 showed that conscientiousness contributed to observers’ judgments whatever pain’ characteristics. Neuroticism was only salient for pain referring to life-threatening pain. ConclusionProsocial response to others’ pain depends on agreeableness, conscientiousness and neuroticism. However, these links are modulated by the pain behavior elicited and observers’ belief about the characteristic of pain.

Atypical Face Perception in Autism: A Point of View?

Face perception is the most commonly used visual metric of social perception in autism. However, when found to be atypical, the origin of face perception differences in autism is contentious. One hypothesis proposes that a locally oriented visual analysis, characteristic of individuals with autism, ultimately affects performance on face tasks where a global analysis is optimal. The objective of this study was to evaluate this hypothesis by assessing face identity discrimination with synthetic faces presented with and without changes in viewpoint, with the former condition minimizing access to local face attributes used for identity discrimination. Twenty-eight individuals with autism and 30 neurotypical participants performed a face identity discrimination task. Stimuli were synthetic faces extracted from traditional face photographs in both front and 20° side viewpoints, digitized from 37 points to provide a continuous measure of facial geometry. Face identity discrimination thresholds were obtained using a two-alternative, temporal forced choice match-to-sample paradigm. Analyses revealed an interaction between group and condition, with group differences found only for the viewpoint change condition, where performance in the autism group was decreased compared to that of neurotypical participants. The selective decrease in performance for the viewpoint change condition suggests that face identity discrimination in autism is more difficult when access to local cues is minimized, and/or when dependence on integrative analysis is increased. These results lend support to a perceptual contribution of atypical face perception in autism.

Evaluating the effect of interactions between appearance-related product designs and facial characteristics on social affectivity

For products that can improve the appearance of the user, such as facial accessories, both the characteristics of the product user and design features must be considered in design evaluation. This paper proposes an experimental evaluation scheme that investigates the interactions between the design features of 3D eyeglasses frames and user facial characteristics. Face models of users containing both geometric and image data were constructed using 3D scanning. A face deformation method was developed to manipulate individual facial features without changing the other features on the face models. In the evaluation scheme, participants judged synthetic faces, which had varied eye distances and orientations and were wearing factorized eyeglasses frames, according to three affective measures related to the personality attributes of confidence, friendliness, and attractiveness. The experimental results show that changing certain design features influences the impressions of the face models with varied facial characteristics. The proposed scheme facilitates designing products that strengthen the impression of specific personality traits by accommodating individual differences in facial features. Relevance to industryThe evaluation scheme proposed in this paper facilitates designing products that strengthen the impression of specific personality traits by accommodating individual differences in facial features. By the scheme, companies can create products and services that satisfy individual customer requirements of personalized design.

Atypical Social Judgment and Sensitivity to Perceptual Cues in Autism Spectrum Disorders.

Evaluation of faces is an important dimension of social relationships. A degraded sensitivity to facial perceptual cues might contribute to atypical social interactions in autism spectrum disorder (ASD). The current study investigated whether face based social judgment is atypical in ASD and if so, whether it could be related to a degraded sensitivity to facial perceptual cues. Individuals with ASD (n=33) and IQ- and age-matched controls (n=38) were enrolled in this study. Watching a series of photographic or synthetic faces, they had to judge them for "kindness". In synthetic stimuli, the amount of perceptual cues available could be either large or small. We observed that social judgment was atypical in the ASD group on photographic stimuli, but, contrarily to the prediction based on the degraded sensitivity hypothesis, analyses on synthetic stimuli found a similar performance and a similar effect of the amount of perceptual cues in both groups. Further studies on perceptual differences between photographs and synthetic pictures of faces might help understand atypical social judgment in ASD.

Investigating Macroexpressions and Microexpressions in Computer Graphics Animated Faces

Due to varied personal, social, or even cultural situations, people sometimes conceal or mask their true emotions. These suppressed emotions can be expressed in a very subtle way by brief movements called microexpressions. We investigate human subjects' perception of hidden emotions in virtual faces, inspired by recent psychological experiments. We created animations with virtual faces showing some facial expressions and inserted brief secondary expressions in some sequences, in order to try to convey a subtle second emotion in the character. Our evaluation methodology consists of two sets of experiments, with three different sets of questions. The first experiment verifies that the accuracy and concordance of the participant's responses with synthetic faces matches the empirical results done with photos of real people in the paper by X.-b. Shen, Q. Wu, and X.-l. Fu, 2012, “Effects of the duration of expressions on the recognition of microexpressions,” Journal of Zhejiang University Science B, 13(3), 221–230. The second experiment verifies whether participants could perceive and identify primary and secondary emotions in virtual faces. The third experiment tries to evaluate the participant's perception of realism, deceit, and valence of the emotions. Our results show that most of the participants recognized the foreground (macro) emotion and most of the time they perceived the presence of the second (micro) emotion in the animations, although they did not identify it correctly in some samples. This experiment exposes the benefits of conveying microexpressions in computer graphics characters, as they may visually enhance a character's emotional depth through subliminal microexpression cues, and consequently increase the perceived social complexity and believability.

3D Assisted Face Recognition: Dealing With Expression Variations

One of the most critical sources of variation in face recognition is facial expressions, especially in the frequent case where only a single sample per person is available for enrollment. Methods that improve the accuracy in the presence of such variations are still required for a reliable authentication system. In this paper, we address this problem with an analysis-by-synthesis-based scheme, in which a number of synthetic face images with different expressions are produced. For this purpose, an animatable 3D model is generated for each user based on 17 automatically located landmark points. The contribution of these additional images in terms of the recognition performance is evaluated with three different techniques (principal component analysis, linear discriminant analysis, and local binary patterns) on face recognition grand challenge and Bosphorus 3D face databases. Significant improvements are achieved in face recognition accuracies, for each database and algorithm.

Masquerade attack on transform-based binary-template protection based on perceptron learning

With the increasing deployment of biometric systems, security of the biometric systems has become an essential issue to which serious attention has to be given. To prevent unauthorized access to a biometric system, protection has to be provided to the enrolled biometric templates so that if the database is compromised, the stored information will not enable any adversary to impersonate the victim in gaining an illegal access. In the past decade, transform-based template protection that stores binary one-way-transformed templates (e.g. Biohash) has appeared being one of the benchmark template protection techniques. While the security of such approach lies in the non-invertibility of the transform (e.g. given a transformed binary template, deriving the corresponding face image is infeasible), we will prove in this paper that, irrespective of whether the algorithm of transform-based approach is revealed, a synthetic face image can be constructed from the binary template and the stolen token (storing projection and discretization parameters) to obtain a highly-probable positive authentication response. Our proposed masquerade attack algorithms are mainly composed of a combination of perceptron learning and customized hill climbing algorithms. Experimental results show that our attack algorithms achieve very promising results where the best setting of our attack achieves 100% and 98.3% rank one recognition rates for the CMU PIE and FRGC databases correspondingly when the binarization algorithm (transformation plus discretization) is known; and 85.29% and 46.57% rank one recognition rates for the CMU PIE and FRGC databases correspondingly when the binarization algorithm is unknown.

A Learner-Independent Knowledge Transfer Approach to Multi-task Learning

This paper proposes a learner-independent multi-task learning (MTL) scheme in which knowledge transfer (KT) is running beyond the learner. In the proposed KT approach, we use minimum enclosing balls (MEBs) as knowledge carriers to extract and transfer knowledge from one task to another. Since the knowledge presented in MEB can be decomposed as raw data, it can be incorporated into any learner as additional training data for a new learning task to improve the learning rate. The effectiveness and robustness of the proposed KT is evaluated, respectively, on multi-task pattern recognition problems derived from synthetic datasets, UCI datasets, and real face image datasets, using classifiers from different disciplines for MTL. The experimental results show that multi-task learners using KT via MEB carriers perform better than learners without-KT, and this has been successfully applied to different classifiers such as k nearest neighbor and support vector machines.