Single Face Research Articles

Learning Low-Rank Regularized Generic Representation With Block-Sparse Structure for Single Sample Face Recognition

In this paper, we propose a novel low-rank regularized generic representation method to address the single sample per person problem in face recognition, which simultaneously employs the structure information from the probe dataset and the generic variation information. Each face image is divided into overlapped patches, whose classification results are aggregated to produce the final result. We generate a subspace for each patch by extracting its eight nearest neighbor patches and explore the relationships between subspaces by imposing a low-rank constraint on the reconstruction coefficients. Moreover, a block-sparsity constraint on the coefficient matrix is imposed to further promote the discrimination of representations. We also propose a dictionary learning method to learn the intra-class facial variations from the generic face datasets, which separates the face contour noise from the variation dictionary by an incoherence regularization item. The experimental results on four public face databases not only show the robustness of our approach to expression, illumination, occlusion, and time variation but also demonstrate the effectiveness of face sketch recognition.

You Only Move Once: An Efficient Convolutional Neural Network for Face Detection

Our “You Only Move Once” (YOMO) detector based on depthwise separable convolutions is a single stage face detector that balances accuracy and latency. YOMO performs scale-invariantly by utilizing top-down architecture with feature agglomeration, and multiple detection modules instead of in an image pyramid approach. At the same time, we propose a semi-soft random cropping algorithm that enables different detection module adequately trained by different scales of samples. Several experiments are conducted on the FDDB dataset with discrete and continuous measures, to demonstrate that the methods have strongly competitiveness results. After using an ellipses regressor, the recall rates reached a satisfactory 97.59% and 83.66%, respectively. Surprisingly, YOMO has only 21 million parameters and achieves superior performance with 51 frames per second (FPS) for a $544\times544$ input image on a GPU.

Face Depth Estimation With Conditional Generative Adversarial Networks

Depth map estimation and 3-D reconstruction from a single or a few face images is an important research field in computer vision. Many approaches have been proposed and developed over the last decade. However, issues like robustness are still to be resolved through additional research. With the advent of the GPU computational methods, convolutional neural networks are being applied to many computer vision problems. Later, conditional generative adversarial networks (CGAN) have attracted attention for its easy adaptation for many picture-to-picture problems. CGANs have been applied for a wide variety of tasks, such as background masking, segmentation, medical image processing, and superresolution. In this work, we developed a GAN-based method for depth map estimation from any given single face image. Many variants of GANs have been tested for the depth estimation task for this work. We conclude that conditional Wasserstein GAN structure offers the most robust approach. We have also compared the method with other two state-of-the-art methods based on deep learning and traditional approaches and experimentally shown that the proposed method offers great opportunities for estimation of face depth maps from face images.

Single Sample Face Recognition under Varying Illumination via QRCP Decomposition.

In this paper, we present a novel high-frequency facial feature and a high-frequency based sparse representation classification to tackle single sample face recognition (SSFR) under varying illumination. Firstly, we propose the assumption that QRCP bases can represent intrinsic face surface features with different frequencies, and their corresponding energy coefficients describe illumination intensities. Based on this assumption, we take QRCP bases with corresponding weighting coefficients (i.e. the major components of energy coefficients) to develop the high-frequency facial feature of the face image, which is named as QRCP-face. The normalized QRCP-face (NQRCPface) is constructed to further constraint illumination effects by normalizing the weighting coefficients of QRCP-face. Moreover, we propose the adaptive QRCP-face (AQRCP-face) that assigns a special parameter to NQRCP-face via the illumination level estimated by the weighting coefficients. Secondly, we consider that the differences of pixel images cannot model the intraclass variations of generic faces with illumination variations, and the specific identification information of the generic face is redundant for the current SSFR with generic learning. To tackle above two issues, we develop a general high-frequency based sparse representation (GHSP) model. Two practical approaches separated high-frequency based sparse representation (SHSP) and unified high-frequency based sparse representation (UHSP) are developed. Finally, the performances of the proposed methods are verified on the Extended Yale B, CMU PIE, AR, LFW and our self-built Driver face databases. The experimental results indicate that the proposed methods outperform previous approaches for SSFR under varying illumination.

Mechanical properties of a node-interlocking pyramidal welded tube lattice sandwich structure

A novel node interlocking assembly method is proposed for manufacturing a new type of pyramidal welded tube lattice sandwich structures with enhanced mechanical properties, consisting of millimeter-scale tubular truss cores and single layer solid face sheets. Sandwich panel specimens with relative densities ranging from 1.99–5.47% are fabricated from a stainless steel alloy. Mechanical responses and energy absorption are tested under through-thickness compressive and transverse shear loading. A three-stage nominal compressive stress-strain model for as-brazed stainless steel welded tubes with different wall thicknesses is proposed to accurately predict the inelastic buckling strength of the lattice cores as well as the nonlinear material behaviors up to a nominal strain level on the order of 0.23. In general, the tested compressive and shear stiffnesses and strengths are in good agreement with analytical predictions. By comparison with numerous competing periodic cellular topologies, the node-interlocking pyramidal welded tube lattices exhibit significantly better overall mechanical properties, especially for specific strengths and specific energy absorptions.

A video prediction approach for animating single face image

Generating dynamic 2D image-based facial expressions is a challenging task for facial animation. Much research work focused on performance-driven facial animation from given videos or images of a target face, while animating a single face image driven by emotion labels is a less explored problem. In this work, we treat the task of animating single face image from emotion labels as a conditional video prediction problem, and propose a novel framework by combining factored conditional restricted boltzmann machines (FCRBM) and reconstruction contractive auto-encoder (RCAE). A modified RCAE with an associated efficient training strategy is used to extract low dimensional features and reconstruct face images. FCRBM is used as animator to predict facial expression sequence in the feature space given discrete emotion labels and a frontal neutral face image as input. Both quantitative and qualitative evaluations on two facial expression databases, and comparison to state-of-the-art showed the effectiveness of our proposed framework for animating frontal neutral face image from given emotion labels.

Nonrepetitive colorings of graphs excluding a fixed immersion or topological minor

AbstractWe prove that graphs excluding a fixed immersion have bounded nonrepetitive chromatic number. More generally, we prove that if is a fixed planar graph that has a planar embedding with all the vertices with degree at least 4 on a single face, then graphs excluding as a topological minor have bounded nonrepetitive chromatic number. This is the largest class of graphs known to have bounded nonrepetitive chromatic number.

Cooperation between Two Micro-Grids Considering Power Exchange: An Optimal Sizing Approach Based on Collaborative Operation

Optimal sizing of single micro-grid faces problems such as high life cycle cost, low self-consumption of power generated by renewable energy, and disturbances of intermittent renewable energy. Interconnecting single micro-grids as a cooperative system to reach a proper size of renewable energy generations and batteries is a credible method to promote performance in reliability and economy. However, to guarantee the optimal collaborative sizing of two micro-grids is a challenging task, particularly with power exchange. In this paper, the optimal sizing of economic and collaborative for two micro-grids and the tie line is modelled as a unit commitment problem to express the influence of power exchange between micro-grids on each life cycle cost, meanwhile guaranteeing certain degree of power supply reliability, which is calculated by Loss of Power Supply Probability in the simulation. A specified collaborative operation of power exchange between two micro-grids is constructed as the scheduling scheme to optimize the life cycle cost of two micro-grids using genetic algorithm. The case study verifies the validity of the method proposed and reveal the advantages of power exchange in the two micro-grids system. The results demonstrate that the proposed optimal sizing means based on collaborative operation can minimize the life cycle cost of two micro-grids respectively considering different renewable energy sources. Compared to the sizing of single micro-grid, the suggested method can not only improve the economic performance for each micro-grid but also form a strong support between interconnected micro-grids. In addition, a proper price of power exchanges will balance the cost saving between micro-grids, making the corresponding stake-holders prefer to be interconnected.

A hybrid approach to building face shape classifier for hairstyle recommender system

Identifying human face shape is the first and the most vital process prior to choosing the right hairstyle to wear on according to guidelines from hairstyle experts, especially for women. This work presents a novel framework for a hairstyle recommender system that is based on face shape classifier. This framework enables an automatic hairstyle recommendation with a single face image. This has a direct impact on beauty industry service providers. It can simulate how the user looks like when she is wearing the chosen hairstyle recommended by the expert system. The model used in this framework is based on Support Vector Machine. The framework is evaluated on hand-crafted, deep-learned (VGG-face) features and VGG-face fine-tuned version for the face shape classification task. In addition to evaluating these individual features by a well-designed framework, we attempted to fuse these three descriptors together in order to improve the performance of the classification task. Two combination techniques were employed, namely: Vector Concatenation and Multiple Kernel Learning (MKL) techniques. All the hyper-parameters of the model were optimised by using Particle Swarm Optimisation. The results show that combining hand-crafted and VGG-face descriptors with MKL yielded the best results at 70.3% of accuracy which was statistically significantly better than using individual features. Thus, combining multiple representations of the data with MKL can improve the overall performance of the expert system. In addition, this proves that hand-crafted descriptor can be complementary to deep-learned descriptor.

Hard negative generation for identity-disentangled facial expression recognition

Various factors such as identity-specific attributes, pose, illumination and expression affect the appearance of face images. Disentangling the identity-specific factors is potentially beneficial for facial expression recognition (FER). Existing image-based FER systems either use hand-crafted or learned features to represent a single face image. In this paper, we propose a novel FER framework, named identity-disentangled facial expression recognition machine (IDFERM), in which we untangle the identity from a query sample by exploiting its difference from its references (e.g., its mined or generated frontal and neutral normalized faces). We demonstrate a possible ‘recognition via generation’ scheme which consists of a novel hard negative generation (HNG) network and a generalized radial metric learning (RML) network. For FER, generated normalized faces are used as hard negative samples for metric learning. The difficulty of threshold validation and anchor selection are alleviated in RML and its distance comparisons are fewer than those of traditional deep metric learning methods. The expression representations of RML achieve superior performance on the CK + , MMI and Oulu-CASIA datasets, given a single query image for testing.

Monofacet-Selective Cavitation within Solid-State Silica-Nanoconfinement toward Janus Iron Oxide Nanocube.

Here, a highly selective solid-state nanocrystal conversion strategy is developed toward concave iron oxide (Fe3O4) nanocube with an open-mouthed cavity engraved exclusively on a single face. The strategy is based on a novel heat-induced nanospace-confined domino-type migration of Fe2+ ions from the SiO2-Fe3O4 interface toward the surrounding silica shell and concomitant self-limiting nanoscale phase-transition to the Fe-silicate form. Equipped with the chemically unique cavity, the produced Janus-type concave iron oxide nanocube was further functionalized with controllable density of catalytic Pt-nanocrystals exclusively on concave sites and utilized as a highly diffusive catalytic Janus nanoswimmer for the efficient degradation of pollutant-dyes in water.

High Resolution Human Eye Tracking During Continuous Visual Search.

While several studies have shown human subjects’ impressive ability to detect faces in individual images in paced settings (Crouzet et al., 2010), we here report the details of an eye movement dataset in which subjects rapidly and continuously targeted single faces embedded in different scenes at rates approaching six face targets each second (including blinks and eye movement times). In this paper, we describe details of a large publicly available eye movement dataset of this new psychophysical paradigm (Martin et al., 2018). The paradigm produced high-resolution eye-tracking data from an experiment on continuous upright and inverted 3° sized face detection in both background and no-background conditions. The new “Zapping” paradigm allowed large amounts of trials to be completed in a short amount of time. For example, our three studies encompassed a total of 288,000 trials done in 72 separate experiments, and yet only took approximately 40 hours of recording for the three experimental cohorts. Each subject did 4000 trials split into eight blocks of 500 consecutive trials in one of the four different experimental conditions: {upright, inverted} × {scene, no scene}. For each condition, there are several covariates of interest, including: temporal eye positions sampled at 1250 hz, saccades, saccade reaction times, microsaccades, pupil dynamics, target luminances, and global contrasts.

Teleface: Serial reproduction of faces reveals a whiteward bias in race memory.

How is race encoded into memory when viewing faces? Here we demonstrate a novel systematic bias in which our memories of faces converge on certain prioritized regions in our underlying "face space," as they relate to perceived race. This convergence was made especially salient using a new visual variant of the method of serial reproduction: "TeleFace." A single face was briefly presented, with its race selected from a smooth continuum between White and Black (matched for mean luminance). The observer then reproduced that face, using a slider to morph a test face along this continuum. Their response was then used as the face initially presented to the next observer, and so on down the line in each reproduction chain. White observers' chains consistently and steadily converged onto faces significantly Whiter than they had initially encountered-Whiter than both the original face in the chain and the continuum's midpoint-regardless of where chains began. Indeed, even chains beginning near the Black end of the continuum inevitably ended up well into White space. Very different patterns resulted when the same method was applied to other arbitrary face stimuli. These results highlight a systematic bias in memory for race in White observers, perhaps contributing to the more general notion in social cognition research of a 'White default.' (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Investigation of various magnetic features of spinel type cobalt ferrite (CoFe2O4) nanoparticles tuned by annealing temperature

Cobalt ferrite (CoFe2O4) nanoparticles synthesized by chemical co-precipitation method were studied to see the effect of annealing temperatures on the structural parameters and magnetic properties of the sample. X-ray powder diffraction (XRD), transmission electron microscope (TEM) and SAD pattern demonstrated that single face (fcc) spinel structure of CoFe2O4 has been formed showing particles in completely crystalline state which was further confirmed by the lattice parameter and the unmixed hkl values. The particle size is in the range of about 5-10 nm depending on the annealing temperature from 200°C to 400°C. The grain growth occurred monotonically with the increase of annealing temperature. The magnetic properties demonstrated a strong dependence on particle size. The saturation magnetization Ms, remanent ratio Mr/Ms and coercivity Hc increased with the increase of particle size. The hysteresis curves for samples of different grain size showed the ferrimagnetic behavior which is completely analogous to the slow relaxation sextet peaks of Mössbauer spectroscopy. Other parameters such as chemical shift, quadruple splitting and hyperfine field and site occupancy of Fe3+ were determined by Mössbauer spectroscopy. Ferrimagnetic to superparamagnetic transition temperature known as blocking temperature TB was determined from the temperature dependent magnetization curves. With the increase of grain size, Blocking temperature also increases. Maximum entropy changes due to magnetic phase transition were also observed for mentioned annealing temperatures in the context of Magnetocaloric effect.

Environmental tobacco smoke exposure among electronic cigarette users

IntroductionExposure to environmental tobacco smoke (ETS) from combustible tobacco products causes various diseases and makes quitting smoking more difficult. However, little is known about exposure of e-cigarette users to ETS from combustible tobacco products. This study aimed to investigate e-cigarette users' exposure to ETS from tobacco smokers. MethodsThe association between ETS exposure frequency and different types of smokers including e-cigarette users was examined using ordered logistic regression analysis and nationally representative survey data on 28,765 individuals who were interviewed in the Canadian Tobacco, Alcohol and Drugs Surveys conducted during 2013 and 2015. Survey respondents were classified into one of five smoker types: smokers of tobacco only, dual users of tobacco and e-cigarettes, users of e-cigarette only, former smokers and never smokers. The analyses were conducted using the entire sample and by age group. ResultsYoung to mid-age (15–54) dual users of both regular cigarettes and e-cigarettes have higher ETS exposure than even tobacco smokers. Young to mid-age single users of e-cigarettes are less exposed to ETS than tobacco smokers, but still have higher ETS than never smokers. At older age (55+), both dual and single e-cigarette users face similar risks of ETS exposure as tobacco smokers. ConclusionsE-cigarette users are at high risk of ETS exposure. Policies that target the behaviour of e-cigarette users as well as the environments surrounding them to address their high ETS exposure risk would be beneficial.

Single Image 3D Face Reconstruction Based on Statistical Model

Single image 3D face reconstruction technology has a broad application in the life, also has a high technical difficulties. This paper proposes a single image 3D face reconstruction technology based on statistical model. The method established 3D face statistical model based on 3D face database in advance, and trained 2D facial feature points parameter model based on the 2D face database using the algorithm of SDM. when reconstruct 3D face model from single image, first we use 2D facial feature points parameter model to extract face feature points; Then, according to 3D face model, we use adaptive learning factor gradient descent method to iteratively optimize the energy function, to get statistical model parametric vector, namely got the 3D face model corresponding to the 2D face images. The results show that use the proposed method to rebuild the faces have high similarity.

Remembering who was where: A happy expression advantage for face identity-location binding in working memory.

It is well established that visual working memory (WM) for face identity is enhanced when faces display threatening versus nonthreatening expressions. During social interaction, it is also important to bind person identity with location information in WM to remember who was where, but we lack a clear understanding of how emotional expression influences this. Here, we conducted two touchscreen experiments to investigate how angry versus happy expressions displayed at encoding influenced the precision with which participants relocated a single neutral test face to its original position. Maintenance interval was manipulated (Experiment 2; 1 s, 3 s, 6 s) to assess durability of binding. In both experiments, relocation accuracy was enhanced when faces were happy versus angry, and this happy benefit endured from 1-s to 6-s maintenance interval. Eye movement measures during encoding showed no convincing effects of oculomotor behavior that could readily explain the happy benefit. However, accuracy in general was improved, and the happy benefit was strongest for the last, most recent face fixated at encoding. Improved, durable binding of who was where in the presence of a happy expression may reflect the importance of prosocial navigation.

Insights into the key determinants of membrane protein topology enable the identification of new monotopic folds.

Monotopic membrane proteins integrate into the lipid bilayer via reentrant hydrophobic domains that enter and exit on a single face of the membrane. Whereas many membrane-spanning proteins have been structurally characterized and transmembrane topologies can be predicted computationally, relatively little is known about the determinants of membrane topology in monotopic proteins. Recently, we reported the X-ray structure determination of PglC, a full-length monotopic membrane protein with phosphoglycosyl transferase (PGT) activity. The definition of this unique structure has prompted in vivo, biochemical, and computational analyses to understand and define key motifs that contribute to the membrane topology and to provide insight into the dynamics of the enzyme in a lipid bilayer environment. Using the new information gained from studies on the PGT superfamily we demonstrate that two motifs exemplify principles of topology determination that can be applied to the identification of reentrant domains among diverse monotopic proteins of interest.

Diffraction of electromagnetic waves by a wedge with perfectly absorbing and impedance faces

The scattering process of electromagnetic waves by a wedge with perfectly absorbing and impedance faces is investigated. The scattered geometric optics waves are obtained for two cases, namely the illumination of single and double faces. The geometric optic field is expressed as a sum of two fields, which represent two wedges with different boundary conditions. The diffracted waves are determined for both of the wedge configurations. The behaviours of the total, total geometric optics and diffracted fields are studied numerically.

Zapping 500 faces in less than 100\u2009seconds: Evidence for extremely fast and sustained continuous visual search

A number of studies have shown human subjects’ impressive ability to detect faces in individual images, with saccade reaction times starting as fast as 100 ms after stimulus onset. Here, we report evidence that humans can rapidly and continuously saccade towards single faces embedded in different scenes at rates approaching 6 faces/scenes each second (including blinks and eye movement times). These observations are impressive, given that humans usually make no more than 2 to 5 saccades per second when searching a single scene with eye movements. Surprisingly, attempts to hide the faces by blending them into a large background scene had little effect on targeting rates, saccade reaction times, or targeting accuracy. Upright faces were found more quickly and more accurately than inverted faces; both with and without a cluttered background scene, and over a large range of eccentricities (4°–16°). The fastest subject in our study made continuous saccades to 500 small 3° upright faces at 4° eccentricities in only 96 seconds. The maximum face targeting rate ever achieved by any subject during any sequence of 7 faces during Experiment 3 for the no scene and upright face condition was 6.5 faces targeted/second. Our data provide evidence that the human visual system includes an ultra-rapid and continuous object localization system for upright faces. Furthermore, these observations indicate that continuous paradigms such as the one we have used can push humans to make remarkably fast reaction times that impose strong constraints and challenges on models of how, where, and when visual processing occurs in the human brain.