Face Detection Task Research Articles

The effect of early visual deprivation on the development of face detection

The expertise of adults in face perception is facilitated by their ability to rapidly detect that a stimulus is a face. In two experiments, we examined the role of early visual input in the development of face detection by testing patients who had been treated as infants for bilateral congenital cataract. Experiment 1 indicated that, at age 9 to 20, patients' accuracy and response times on a Mooney face detection task were normal. Experiment 2 revealed that the neural mechanisms underlying face detection in a similar group of adult patients are abnormal: the amplitude of both the P100 and N170 event-related potential were larger in patients than in visually normal controls, and the extent of augmentation was related to the duration of deprivation. Thus, early visual experience is necessary for the establishment of normal neural networks for face detection; abnormalities at these early processing stages may contribute to the deficits we previously reported in configural face processing for this patient cohort.

Infant face preferences after binocular visual deprivation

Early visual deprivation impairs some, but not all, aspects of face perception. We investigated the possible developmental roots of later abnormalities by using a face detection task to test infants treated for bilateral congenital cataract within 1 hour of their first focused visual input. The seven patients were between 5 and 12 weeks old (n = 3) or older than 12 weeks (n = 4). Like newborns, but unlike visually normal age-matched controls, the patients looked preferentially toward config (three squares arranged as facial features) over its inverted version and none of the older patients preferred a positive-contrast face over the negative-contrast version. We conclude that postnatal changes in face perception are experience-dependent, and that interference with their typical development may contribute to later deficits in face processing.

A Hybrid Approach for Gender Classification of Web Images

In recent times, gender recognition of facial images has achieved lots of attraction. It can be useful in many places e.g. security, web searching, human computer interaction etc. In this paper, an approach containing both face detection and gender classification tasks has been proposed. In face detection part, Haar features have been chosen to present appearance features along with Ada-Boost technique to target strong and powerful features in cascaded form. For gender classification, Bayesian Classifier has been used where image is analyzed in blocks/patches form. The blocking technique is same as used in DCT approach. Experimental results have shown that proposed approach is effective and robust with changes in pose (some degree), expressions and illumination.

Comparing computer and human performance on identical face detection tasks

Comparing computer and human performance on identical face detection tasks

Absence of Face-specific Cortical Activity in the Complete Absence of Awareness: Converging Evidence from Functional Magnetic Resonance Imaging and Event-related Potentials

In this study, we explored the neural correlates of perceptual awareness during a masked face detection task. To assess awareness more precisely than in previous studies, participants employed a 4-point scale to rate subjective visibility. An event-related fMRI and a high-density ERP study were carried out. Imaging data showed that conscious face detection was linked to activation of fusiform and occipital face areas. Frontal and parietal regions, including the pre-SMA, inferior frontal sulcus, anterior insula/frontal operculum, and intraparietal sulcus, also responded strongly when faces were consciously perceived. In contrast, no brain area showed face-selective activity when participants reported no impression of a face. ERP results showed that conscious face detection was associated with enhanced N170 and also with the presence of a second negativity around 300 msec and a slow positivity around 415 msec. Again, face-related activity was absent when faces were not consciously perceived. We suggest that, under conditions of backward masking, ventral stream and fronto-parietal regions show similar, strong links of face-related activity to conscious perception and stress the importance of a detailed assessment of awareness to examine activity related to unseen stimulus events.

Measuring Internal Representations from Behavioral and Brain Data

The study of internal knowledge representations is a cornerstone of the research agenda in the interdisciplinary study of cognition. An influential proposal assumes that the brain uses its internal knowledge of the external world to constrain, in a top-down manner, high-dimensional sensory data into a lower-dimensional representation that enables perceptual decisions and other higher-level cognitive functions [1-9]. This proposal relies on a precise formulation of the observer-specific internal knowledge (i.e., the internal representations, or models) that guides reduction of the high-dimensional retinal input onto a low-dimensional code. Here, we directly revealed the content of subjective internal representations by instructing five observers to detect a face in the presence of only white noise, to force a pure top-down, knowledge-based task. We used reverse correlation methods to visualize each observer's internal representation that supports detection of an illusory face. Using reverse correlation again, this time applied to observers' electroencephalogram activity, we established where and when in the brain specific internal knowledge conceptually interprets the input white noise as a face. We show that internal representations can be reconstructed experimentally from behavioral and brain data, and that their content drives neural activity first over frontal and then over occipitotemporal cortex.

Context modeling for facial landmark detection based on Non-Adjacent Rectangle (NAR) Haar-like feature

Automatically locating facial landmarks in images is an important task in computer vision. This paper proposes a novel context modeling method for facial landmark detection, which integrates context constraints together with local texture model in the cascaded AdaBoost framework. The motivation of our method lies in the basic human psychology observation that not only the local texture information but also the global context information is used for human to locate facial landmarks in faces. Therefore, in our solution, a novel type of feature, called Non-Adjacent Rectangle (NAR) Haar-like feature, is proposed to characterize the co-occurrence between facial landmarks and its surroundings, i.e., the context information, in terms of low-level features. For the locating task, traditional Haar-like features (characterizing local texture information) and NAR Haar-like features (characterizing context constraints in global sense) are combined together to form more powerful representations. Through Real AdaBoost learning, the most discriminative feature set is selected automatically and used for facial landmark detection. To verify the effectiveness of the proposed method, we evaluate our facial landmark detection algorithm on BioID and Cohn-Kanade face databases. Experimental results convincingly show that the NAR Haar-like feature is effective to model the context and our proposed algorithm impressively outperforms the published state-of-the-art methods. In addition, the generalization capability of the NAR Haar-like feature is further validated by extended applications to face detection task on FDDB face database.

Complexity Reduced Face Detection Using Probability-Based Face Mask Prefiltering and Pixel-Based Hierarchical-Feature Adaboosting

The Adaboosting has attracted attention for its efficient face-detection performance. However, in the training process, the large number of possible Haar-like features in a standard sub-window becomes time consuming, which makes specific environment feature adaptation extremely difficult. This letter presents a two-stage hybrid face detection scheme using Probability-based Face Mask Pre-Filtering (PFMPF) and the Pixel-Based Hierarchical-Feature Adaboosting (PBHFA) method to effectively solve the above-mentioned problems in cascade Adaboosting. The two stages both provide far less training time than that of the cascade Adaboosting and thus reduce the computation complexity in face-detection tasks. In particular, the proposed PFMPF can effectively filter out more than 85% nonface in an image and the remaining few face candidates are then secondly filtered with a single PBHF Adaboost strong classifier. Given a <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">${\rm M}\times{\rm N}$</tex></formula> sub-window, the number of possible PBH features is simplified down to a level less than <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">${\rm M}\times{\rm N}$</tex></formula> , which significantly reduces the length of the training period by a factor of 1500. Moreover, when the two-stage hybrid face detection scheme are employed for practical face-detection tasks, the complexity is still lower than that of the integral-image based approach in the traditional Adaboosting method. Experimental results obtained using the gray feret database show that the proposed two-stage hybrid face detection scheme is significantly more effective than Haar-like features.

Adaptive Gaze Control for Object Detection

We propose a novel gaze-control model for detecting objects in images. The model, named act-detect, uses the information from local image samples in order to shift its gaze towards object locations. The model constitutes two main contributions. The first contribution is that the model’s setup makes it computationally highly efficient in comparison with existing window-sliding methods for object detection, while retaining an acceptable detection performance. act-detect is evaluated on a face-detection task using a publicly available image set. In terms of detection performance, act-detect slightly outperforms the window-sliding methods that have been applied to the face-detection task. In terms of computational efficiency, act-detect clearly outperforms the window-sliding methods: it requires in the order of hundreds fewer samples for detection. The second contribution of the model lies in its more extensive use of local samples than previous models: instead of merely using them for verifying object presence at the gaze location, the model uses them to determine a direction and distance to the object of interest. The simultaneous adaptation of both the model’s visual features and its gaze-control strategy leads to the discovery of features and strategies for exploiting the local context of objects. For example, the model uses the spatial relations between the bodies of the persons in the images and their faces. The resulting gaze control is a temporal process, in which the object’s context is exploited at different scales and at different image locations relative to the object.

Differential impact of attention on the early and late categorization related human brain potentials

Previous human studies have shown that object-selective attention enhances neural activities evoked in response to target stimuli. However, it is not clear whether the magnitude of activity enhancement is constant or varies according to the level of stimulus visibility. To examine the effect of attentional modulation on different event-related potentials (ERPs) and the relationship between attentional modulation and level of stimulus visibility, subjects were instructed to perform face detection and leaf detection tasks in separate blocks while the level of stimulus visibility varied randomly from trial to trial. As a result of object-selective attention, N170 and P400 ERP components were both modulated in response to the task target category compared to when the same category was used as distracter. We found that the magnitude of modulation of the N170 component was independent of the stimulus visibility level, while the P400 ERP component showed increased enhancement as the level of stimulus visibility increased. These findings demonstrate that attention impacts the neural populations, indexed by early (N170) and late (P400) evoked potentials, differentially and that attentional modulation becomes increasingly dependent on the level of stimulus visibility through the course of brain activities.

Particle swarm optimisation based AdaBoost for object detection

This paper proposes a new approach to using particle swarm optimisation (PSO) within an AdaBoost framework for object detection. Instead of using exhaustive search for finding good features to be used for constructing weak classifiers in AdaBoost, we propose two methods based on PSO. The first uses PSO to evolve and select good features only, and the weak classifiers use a simple decision stump. The second uses PSO for both selecting good features and evolving weak classifiers in parallel. These two methods are examined and compared on two challenging object detection tasks in images: detection of individual pasta pieces and detection of a face. The experimental results suggest that both approaches can successfully detect object positions and that using PSO for selecting good individual features and evolving associated weak classifiers in AdaBoost is more effective than for selecting features only. We also show that PSO can evolve and select meaningful features in the face detection task.

In the Eye of the Beholder: A Survey of Models for Eyes and Gaze

Despite active research and significant progress in the last 30 years, eye detection and tracking remains challenging due to the individuality of eyes, occlusion, variability in scale, location, and light conditions. Data on eye location and details of eye movements have numerous applications and are essential in face detection, biometric identification, and particular human-computer interaction tasks. This paper reviews current progress and state of the art in video-based eye detection and tracking in order to identify promising techniques as well as issues to be further addressed. We present a detailed review of recent eye models and techniques for eye detection and tracking. We also survey methods for gaze estimation and compare them based on their geometric properties and reported accuracies. This review shows that, despite their apparent simplicity, the development of a general eye detection technique involves addressing many challenges, requires further theoretical developments, and is consequently of interest to many other domains problems in computer vision and beyond.

Face Detection using Neural Network

paper describes a face detection framework that is capable of processing images extremely rapidly while achieving high detection rates. As continual research is being conducted in the area of computer vision, one of the most practical applications under vigorous development is in the construction of a robust real-time face detection system. Successfully constructing a real-time face detection system not only implies a system capable of analyzing video streams, but also naturally leads onto the solution to the problems of extremely constraint testing environments. Analyzing a video sequence is the current challenge since faces are constantly in dynamic motion, presenting many different possible rotational and illumination conditions. While solutions to the task of face detection have been presented, detection performances of many systems are heavily dependent upon a strictly constrained environment. The problem of detecting faces under gross variations remains largely uncovered. This paper gives a face detection system which uses an image based neural network to detect face images.

A Hybrid Approach to Human Face Detection

Face detection is the problem of determining whether a subwindow of an image contains a face or not. The rapidly expanding research in face processing is based on the premise that information about a user’s identity, state, and intent can be extracted from images, and that computers can then react accordingly. This hybrid face detection system is combination of two methods i.e. Feature Extraction method and Neural Networks method. This method works in two stages. The first stage involves extraction of pertinent features from the localized facial image using Gabor filters. Second stage requires classification of facial images based on the derived feature vector obtained in the previous stage. And so a Neural Network Based classifier is used which examines an incremental small window of an image to decide if there is a face contained in each window. The neural network is trained to choose between two classes „faces‟ and „non-faces‟ images. To decrease the amount of time needed for detection and to enhance image’s quality, processing the image before training the neural network on these two classes enhances the algorithm. Training a neural network for the face detection task is a challenging job due to the difficulty in characterizing „non-face‟ images. It is easy to get a representative sample of images that contain faces but it is much harder to get a representative sample of those images that do not contain faces.

Complementary roles of systems representing sensory evidence and systems detecting task difficulty during perceptual decision making.

Perceptual decision making is a multi-stage process where incoming sensory information is used to select one option from several alternatives. Researchers typically have adopted one of two conceptual frameworks to define the criteria for determining whether a brain region is involved in decision computations. One framework, building on single-unit recordings in monkeys, posits that activity in a region involved in decision making reflects the accumulation of evidence toward a decision threshold, thus showing the lowest level of BOLD signal during the hardest decisions. The other framework instead posits that activity in a decision-making region reflects the difficulty of a decision, thus showing the highest level of BOLD signal during the hardest decisions. We had subjects perform a face detection task on degraded face images while we simultaneously recorded BOLD activity. We searched for brain regions where changes in BOLD activity during this task supported either of these frameworks by calculating the correlation of BOLD activity with reaction time – a measure of task difficulty. We found that the right supplementary eye field, right frontal eye field, and right inferior frontal gyrus had increased activity relative to baseline that positively correlated with reaction time, while the left superior frontal sulcus and left middle temporal gyrus had decreased activity relative to baseline that negatively correlated with reaction time. We propose that a simple mechanism that scales a region's activity based on task demands can explain our results.

A study of N250 event-related brain potential during face and non-face detection tasks

Face perception relies on activation of a complex set of different neural modules. In this study, we assessed the stimulus selectivity of the occipitotemporal N250 ERP component and the possible link between its neural substrates and modules underlying preceding (N170/VPP) and following (P400) category selective ERPs. We recorded N250 during face and leaf detection tasks while we varied stimulus visibility from trial to trial by using a backward masking paradigm. Our results revealed that N250, but not the other tested potentials, was exclusively sensitive to the visibility of faces even when the non-face stimuli served as the task target. We also found a correlation between evoked N170 and N250, in response to face stimuli and to a lesser extent in response to other non-face objects, irrespective of the subjects' task. Besides N250, P400 also showed a strong correlation with N170, but here, the amount of correlation was not affected by stimulus category. Interestingly, despite N250 and N400 correlation with N170, we did not find any correlation between N250 and P400, suggesting that modules underlying these ERP components belong to two different face-processing pathways. We suggest that N250 is initiated by N170 and indexes processes exclusively responsible for encoding faces.

Data preparation for sample-based face detection

Face recognition is one of the most active biometric traits. Sample-based face detection is an active research topic in this area. It is of great interest to enhance the performance of the current face detection methods. In this paper, we propose new algorithms of sample selection and active samples generation to solve the problem of imbalanced training samples in the face detection task. Experimental results show that our proposed approaches enhance the performance of face detection, and the accuracy gain of the proposed approaches is significant when the available training samples are imbalanced or insufficient.

Fast and Robust Face Detection Using Evolutionary Pruning

Face detection task can be considered as a classifier training problem. Finding the parameters of the classifier model by using training data is a complex process. To solve such a complex problem, evolutionary algorithms can be employed in cascade structure of classifiers. This paper proposes evolutionary pruning to reduce the number of weak classifiers in AdaBoost-based cascade detector, while maintaining the detection accuracy. The computation time is proportional to the number of weak classifiers and, therefore, a reduction in the number of weak classifiers results in an increased detection speed. Three kinds of cascade structures are compared by the number of weak classifiers. The efficiency in computation time of the proposed cascade structure is shown experimentally. It is also compared with the state-of-the-art face detectors, and the results show that the proposed method outperforms the previous studies. A multiview face detector is constructed by incorporating the three face detectors: frontal, left profile, and right profile.

Brain response to complex visual stimuli in Parkinson's patients with hallucinations: A functional magnetic resonance imaging study

Visual hallucinations (VH) in Parkinson's disease (PD) have been associated with gray matter reductions in visual associative areas and with abnormal patterns of brain activation in posterior and frontal regions. However, all previous fMRI studies have used simple visual stimuli. The objective of our study was, therefore, to compare the pattern of brain activation during a one-back face detection task. We examined 10 PD patients with VH, 10 PD patients without VH, and 10 controls matched for age and education. The fMRI task consisted in three blocks of 21-face stimuli (activation condition) and three blocks of 21-colored mosaics (control condition). Subjects were asked to press a key when two identical stimuli were presented consecutively. During the face condition, compared with patients without VH, hallucinating PD patients showed significant reductions in the activation of several right prefrontal areas, such as the inferior (BA 10,47), superior (BA 6/8), middle frontal (BA 8), and anterior cingulate gyrus (BA 31/32). In the control condition, we found a hyperactivation in the hallucinating PD sample compared with the nonVH patients in the right inferior frontal gyrus. A dysfunction of the frontal areas associated with the control of attention could predispose to VH through an abnormal processing of relevant and irrelevant visual stimuli.

Electrophysiological evidence for early non-conscious processing of fearful facial expressions

Non-conscious processing of emotionally expressive faces has been found in patients with damage to visual brain areas and has been demonstrated experimentally in healthy controls using visual masking procedures. The time at which this subliminal processing occurs is not known. To address this question, a group of healthy participants performed a fearful face detection task in which backward masked fearful and non-fearful faces were presented at durations ranging from 16 to 266 ms. On the basis of the group's behavioural results, high-density event-related potentials were analysed for subliminal, intermediate and supraliminal presentations. Subliminally presented fearful faces were found to produce a stronger posterior negativity at 170 ms (N170) than non-fearful faces. This increase was also observed for intermediate and supraliminal conditions. A later component, the N2 occurring between 260 and 300 ms, was the earliest component related to stimulus detectability, increasing with target duration and differentiating fearful from non-fearful faces at longer durations of presentation. Source localisation performed on the N170 component showed that fear produced a greater activation of extrastriate visual areas, particularly on the right. Whether they are presented subliminally or supraliminally, fearful faces are processed at an early stage in the stream of visual processing, giving rise to enhanced activation of right extrastriate temporal cortex as early as 170 ms post-stimulus onset.