State-of-the-art in visual attention modeling

A unified computational framework for visual attention dynamics

Eye movements are an essential part of human vision as they drive the fovea and, consequently, selective visual attention toward a region of interest in space. Free visual exploration is an inherently stochastic process depending on image statistics but also individual variability of cognitive and attentive state. We propose a theory of free visual exploration entirely formulated within the framework of physics and based on the general Principle of Least Action. Within this framework, differential laws describing eye movements emerge in accordance with bottom-up functional principles. In addition, we integrate top-down semantic information captured by deep convolutional neural networks pre-trained for the classification of common objects. To stress the model, we used a wide collection of images including basic features as well as high level semantic content. Results in a task of saliency prediction validate the theory.

Disentangling bottom-up versus top-down and low-level versus high-level influences on eye movements over time

Bottom-up and top-down as well as low-level and high-level factors influence where we fixate when viewing natural scenes. However, the importance of each of these factors and how they interact remains a matter of debate. Here, we disentangle these factors by analyzing their influence over time. For this purpose, we develop a saliency model that is based on the internal representation of a recent early spatial vision model to measure the low-level, bottom-up factor. To measure the influence of high-level, bottom-up features, we use a recent deep neural network-based saliency model. To account for top-down influences, we evaluate the models on two large data sets with different tasks: first, a memorization task and, second, a search task. Our results lend support to a separation of visual scene exploration into three phases: the first saccade, an initial guided exploration characterized by a gradual broadening of the fixation density, and a steady state that is reached after roughly 10 fixations. Saccade-target selection during the initial exploration and in the steady state is related to similar areas of interest, which are better predicted when including high-level features. In the search data set, fixation locations are determined predominantly by top-down processes. In contrast, the first fixation follows a different fixation density and contains a strong central fixation bias. Nonetheless, first fixations are guided strongly by image properties, and as early as 200 ms after image onset, fixations are better predicted by high-level information. We conclude that any low-level, bottom-up factors are mainly limited to the generation of the first saccade. All saccades are better explained when high-level features are considered, and later, this high-level, bottom-up control can be overruled by top-down influences.

Predicting the Driver's Focus of Attention: The DR(eye)VE Project

In this work we aim to predict the driver's focus of attention. The goal is to estimate what a person would pay attention to while driving, and which part of the scene around the vehicle is more critical for the task. To this end we propose a new computer vision model based on a multi-branch deep architecture that integrates three sources of information: raw video, motion and scene semantics. We also introduce DR(eye)VE, the largest dataset of driving scenes for which eye-tracking annotations are available. This dataset features more than 500,000 registered frames, matching ego-centric views (from glasses worn by drivers) and car-centric views (from roof-mounted camera), further enriched by other sensors measurements. Results highlight that several attention patterns are shared across drivers and can be reproduced to some extent. The indication of which elements in the scene are likely to capture the driver's attention may benefit several applications in the context of human-vehicle interaction and driver attention analysis.

A Deep Network Solution for Attention and Aesthetics Aware Photo Cropping

We study the problem of photo cropping, which aims to find a cropping window of an input image to preserve as much as possible its important parts while being aesthetically pleasant. Seeking a deep learning-based solution, we design a neural network that has two branches for attention box prediction (ABP) and aesthetics assessment (AA), respectively. Given the input image, the ABP network predicts an attention bounding box as an initial minimum cropping window, around which a set of cropping candidates are generated with little loss of important information. Then, the AA network is employed to select the final cropping window with the best aesthetic quality among the candidates. The two sub-networks are designed to share the same full-image convolutional feature map, and thus are computationally efficient. By leveraging attention prediction and aesthetics assessment, the cropping model produces high-quality cropping results, even with the limited availability of training data for photo cropping. The experimental results on benchmark datasets clearly validate the effectiveness of the proposed approach. In addition, our approach runs at 5 fps, outperforming most previous solutions. The code and results are available at: https://github.com/shenjianbing/DeepCropping.

Modeling of Human Visual Attention in Multiparty Open-World Dialogues

This study proposes, develops, and evaluates methods for modeling the eye-gaze direction and head orientation of a person in multiparty open-world dialogues, as a function of low-level communicative signals generated by his/hers interlocutors. These signals include speech activity, eye-gaze direction, and head orientation, all of which can be estimated in real time during the interaction. By utilizing these signals and novel data representations suitable for the task and context, the developed methods can generate plausible candidate gaze targets in real time. The methods are based on Feedforward Neural Networks and Long Short-Term Memory Networks. The proposed methods are developed using several hours of unrestricted interaction data and their performance is compared with a heuristic baseline method. The study offers an extensive evaluation of the proposed methods that investigates the contribution of different predictors to the accurate generation of candidate gaze targets. The results show that the methods can accurately generate candidate gaze targets when the person being modeled is in a listening state. However, when the person being modeled is in a speaking state, the proposed methods yield significantly lower performance.

Do target detection and target localization always go together? Extracting information from briefly presented displays

The human visual system is capable of processing an enormous amount of information in a short time. Although rapid target detection has been explored extensively, less is known about target localization. Here we used natural scenes and explored the relationship between being able to detect a target (present vs. absent) and being able to localize it. Across four presentation durations (~ 33-199 ms), participants viewed scenes taken from two superordinate categories (natural and manmade), each containing exemplars from four basic scene categories. In a two-interval forced choice task, observers were asked to detect a Gabor target inserted in one of the two scenes. This was followed by one of two different localization tasks. Participants were asked either to discriminate whether the target was on the left or the right side of the display or to click on the exact location where they had seen the target. Targets could be detected and localized at our shortest exposure duration (~ 33 ms), with a predictable improvement in performance with increasing exposure duration. We saw some evidence at this shortest duration of detection without localization, but further analyses demonstrated that these trials typically reflected coarse or imprecise localization information, rather than its complete absence. Experiment 2 replicated our main findings while exploring the effect of the level of "openness" in the scene. Our results are consistent with the notion that when we are able to extract what objects are present in a scene, we also have information about where each object is, which provides crucial guidance for our goal-directed actions.

Superior colliculus encodes visual saliency during smooth pursuit eye movements

The saliency map has played a long-standing role in models and theories of visual attention, and it is now supported by neurobiological evidence from several cortical and subcortical brain areas. While visual saliency is computed during moments of active fixation, it is not known whether the same is true while engaged in smooth pursuit of a moving stimulus, which is very common in real-world vision. Here, we examined extrafoveal saliency coding in the superior colliculus, a midbrain area associated with attention and gaze, during smooth pursuit eye movements. We found that SC neurons from the superficial visual layers showed a robust representation of peripheral saliency evoked by a conspicuous stimulus embedded in a wide-field array of goal-irrelevant stimuli. In contrast, visuomotor neurons from the intermediate saccade-related layers showed a poor saliency representation, even though most of these neurons were visually responsive during smooth pursuit. These results confirm and extend previous findings that place the SCs in a unique role as a saliency map that monitors peripheral vision during foveation of stationary and now moving objects.

Meaning and Attentional Guidance in Scenes: A Review of the Meaning Map Approach

Perception of a complex visual scene requires that important regions be prioritized and attentionally selected for processing. What is the basis for this selection? Although much research has focused on image salience as an important factor guiding attention, relatively little work has focused on semantic salience. To address this imbalance, we have recently developed a new method for measuring, representing, and evaluating the role of meaning in scenes. In this method, the spatial distribution of semantic features in a scene is represented as a meaning map. Meaning maps are generated from crowd-sourced responses given by naïve subjects who rate the meaningfulness of a large number of scene patches drawn from each scene. Meaning maps are coded in the same format as traditional image saliency maps, and therefore both types of maps can be directly evaluated against each other and against maps of the spatial distribution of attention derived from viewers' eye fixations. In this review we describe our work focusing on comparing the influences of meaning and image salience on attentional guidance in real-world scenes across a variety of viewing tasks that we have investigated, including memorization, aesthetic judgment, scene description, and saliency search and judgment. Overall, we have found that both meaning and salience predict the spatial distribution of attention in a scene, but that when the correlation between meaning and salience is statistically controlled, only meaning uniquely accounts for variance in attention.

Unique objects attract attention even when faint

Locally contrasting objects, e.g. a red apple surrounded by green apples, attract attention. Does this generalize to differences in feature space? That is, do unique objects-regardless of their location-stand out from a collection of objects that are similar to one another, even when the unique object has lower local contrast with the background than the other objects? Behavioral data show indeed a preference for unique items but previous experiments enabled viewers to anticipate what response they were "supposed" to give. We developed a new experimental paradigm that minimizes such top-down effects. Pitting local contrast against global uniqueness, we show that unique stimuli attract attention even in not-anticipated, never-seen images, and even when the unique stimuli are faint (low contrast). A computational model explains how competition between objects in feature space favors dissimilar objects over those with similar features. The model explains how humans select unique objects, without a loss of performance on natural scenes.

Online Data Thinning via Multi-Subspace Tracking

In an era of ubiquitous large-scale streaming data, the availability of data far exceeds the capacity of expert human analysts. In many settings, such data is either discarded or stored unprocessed in data centers. This paper proposes a method of online data thinning, in which large-scale streaming datasets are winnowed to preserve unique, anomalous, or salient elements for timely expert analysis. At the heart of this proposed approach is an online anomaly detection method based on dynamic, low-rank Gaussian mixture models. Specifically, the high-dimensional covariance matrices associated with the Gaussian components are associated with low-rank models. According to this model, most observations lie near a union of subspaces. The low-rank modeling mitigates the curse of dimensionality associated with anomaly detection for high-dimensional data, and recent advances in subspace clustering and subspace tracking allow the proposed method to adapt to dynamic environments. Furthermore, the proposed method allows subsampling, is robust to missing data, and uses a mini-batch online optimization approach. The resulting algorithms are scalable, efficient, and are capable of operating in real time. Experiments on wide-area motion imagery and e-mail databases illustrate the efficacy of the proposed approach.

SGaze: A Data-Driven Eye-Head Coordination Model for Realtime Gaze Prediction

We present a novel, data-driven eye-head coordination model that can be used for realtime gaze prediction for immersive HMD-based applications without any external hardware or eye tracker. Our model (SGaze) is computed by generating a large dataset that corresponds to different users navigating in virtual worlds with different lighting conditions. We perform statistical analysis on the recorded data and observe a linear correlation between gaze positions and head rotation angular velocities. We also find that there exists a latency between eye movements and head movements. SGaze can work as a software-based realtime gaze predictor and we formulate a time related function between head movement and eye movement and use that for realtime gaze position prediction. We demonstrate the benefits of SGaze for gaze-contingent rendering and evaluate the results with a user study.

Machine learning accurately classifies age of toddlers based on eye tracking

How people extract visual information from complex scenes provides important information about cognitive processes. Eye tracking studies that have used naturalistic, rather than highly controlled experimental stimuli, reveal that variability in looking behavior is determined by bottom-up image properties such as intensity, color, and orientation, top-down factors such as task instructions and semantic information, and individual differences in genetics, cognitive function and social functioning. These differences are often revealed using areas of interest that are chosen by the experimenter or other human observers. In contrast, we adopted a data-driven approach by using machine learning (Support Vector Machine (SVM) and Deep Learning (DL)) to elucidate factors that contribute to age-related variability in gaze patterns. These models classified the infants by age with a high degree of accuracy, and identified meaningful features distinguishing the age groups. Our results demonstrate that machine learning is an effective tool for understanding how looking patterns vary according to age, providing insight into how toddlers allocate attention and how that changes with development. This sensitivity for detecting differences in exploratory gaze behavior in toddlers highlights the utility of machine learning for characterizing a variety of developmental capacities.

Mid-level feature contributions to category-specific gaze guidance

Our research has previously shown that scene categories can be predicted from observers' eye movements when they view photographs of real-world scenes. The time course of category predictions reveals the differential influences of bottom-up and top-down information. Here we used these known differences to determine to what extent image features at different representational levels contribute toward guiding gaze in a category-specific manner. Participants viewed grayscale photographs and line drawings of real-world scenes while their gaze was tracked. Scene categories could be predicted from fixation density at all times over a 2-s time course in both photographs and line drawings. We replicated the shape of the prediction curve found previously, with an initial steep decrease in prediction accuracy from 300 to 500 ms, representing the contribution of bottom-up information, followed by a steady increase, representing top-down knowledge of category-specific information. We then computed the low-level features (luminance contrasts and orientation statistics), mid-level features (local symmetry and contour junctions), and Deep Gaze II output from the images, and used that information as a reference in our category predictions in order to assess their respective contributions to category-specific guidance of gaze. We observed that, as expected, low-level salience contributes mostly to the initial bottom-up peak of gaze guidance. Conversely, the mid-level features that describe scene structure (i.e., local symmetry and junctions) split their contributions between bottom-up and top-down attentional guidance, with symmetry contributing to both bottom-up and top-down guidance, while junctions play a more prominent role in the top-down guidance of gaze.

A New Maritime Moving Target Detection and Tracking Method for Airborne Forward-looking Scanning Radar

Maritime moving target detection and tracking through particle filter based track-before-detect (PF-TBD) has significant practical value for airborne forward-looking scanning radar. However, villainous weather and surging of ocean waves make it extremely difficult to accurately obtain a statistical model for sea clutter. As the likelihood ratio calculation in PF-TBD is dependent on the distribution of the clutter, the performance of traditional distribution-based PF-TBD seriously declines. To resolve these difficulties, this paper proposes a new target detection and tracking method, named spectral-residual-binary-entropy-based PF-TBD (SRBE-PF-TBD), which is independent from the prior knowledge of sea clutter. In the proposed method, the likelihood ratio calculation is implemented by first extracting the spectral residual of the input image to obtain the saliency map, and then constructing likelihood ratio through a binarization processing and information entropy calculation. Simulation results show that the proposed method had superior performance of maritime moving target detection and tracking.

Deeply Supervised Salient Object Detection with Short Connections

Recent progress on salient object detection is substantial, benefiting mostly from the explosive development of Convolutional Neural Networks (CNNs). Semantic segmentation and salient object detection algorithms developed lately have been mostly based on Fully Convolutional Neural Networks (FCNs). There is still a large room for improvement over the generic FCN models that do not explicitly deal with the scale-space problem. The Holistically-Nested Edge Detector (HED) provides a skip-layer structure with deep supervision for edge and boundary detection, but the performance gain of HED on saliency detection is not obvious. In this paper, we propose a new salient object detection method by introducing short connections to the skip-layer structures within the HED architecture. Our framework takes full advantage of multi-level and multi-scale features extracted from FCNs, providing more advanced representations at each layer, a property that is critically needed to perform segment detection. Our method produces state-of-the-art results on 5 widely tested salient object detection benchmarks, with advantages in terms of efficiency (0.08 seconds per image), effectiveness, and simplicity over the existing algorithms. Beyond that, we conduct an exhaustive analysis of the role of training data on performance. We provide a training set for future research and fair comparisons.

Gender-based eye movement differences in passive indoor picture viewing: An eye-tracking study

Male and female brains have different structures, which can make genders produce various eye- movement patterns. This study presents the results of an eye tracking experiment in which we analyzed the eye movements of 25 male and 20 female participants during passive indoor picture viewing. We examined eye-movement parameters, namely fixation duration, scan path length, number of saccades, spatial density, saccade amplitude, and the ratio of total fixation duration to total saccade duration so as to investigate gender differences in eye-movement patterns while indoor picture viewing. We found significant differences in eye-movement patterns between genders. The results of eye-movement analysis also indicated that females showed more explorative gaze behavior, indicated by larger saccade amplitudes, and by longer scan paths. Furthermore, owing to shorter ratio of fixation durations to saccade duration in females as compared to male, we speculate that females inspect the images faster than males. In addition, we classified the genders into two subgroups-males and females-based on their eye-movement parameters by using a support vector machine classifier achieving an accuracy of 70%. We have come to the result males and females - with same culture - see the environment differently. Our findings have profound implications for researches employing gaze-based models.

Guidance in Cinematic Virtual Reality-Taxonomy, Research Status and Challenges

In Cinematic Virtual Reality (CVR), the viewer of an omnidirectional movie can freely choose the viewing direction when watching a movie. Therefore, traditional techniques in filmmaking for guiding the viewers’ attention cannot be adapted directly to CVR. Practices such as panning or changing the frame are no longer defined by the filmmaker; rather it is the viewer who decides where to look. In some stories, it is necessary to show certain details to the viewer, which should not be missed. At the same time, the freedom of the viewer to look around in the scene should not be destroyed. Therefore, techniques are needed which guide the attention of the spectator to visual information in the scene. Attention guiding also has the potential to improve the general viewing experience, since viewers will be less afraid to miss something when watching an omnidirectional movie where attention-guiding techniques have been applied. In recent years, there has been a lot of research about attention guiding in images, movies, virtual reality, augmented reality and also in CVR. We classify these methods and offer a taxonomy for attention-guiding methods. Discussing the different characteristics, we elaborate the advantages and disadvantages, give recommendations for use cases and apply the taxonomy to several examples of guiding methods.

Infrared Small Target Detection Based on Non-Convex Optimization with Lp-Norm Constraint

The infrared search and track (IRST) system has been widely used, and the field of infrared small target detection has also received much attention. Based on this background, this paper proposes a novel infrared small target detection method based on non-convex optimization with Lp-norm constraint (NOLC). The NOLC method strengthens the sparse item constraint with Lp-norm while appropriately scaling the constraints on low-rank item, so the NP-hard problem is transformed into a non-convex optimization problem. First, the infrared image is converted into a patch image and is secondly solved by the alternating direction method of multipliers (ADMM). In this paper, an efficient solver is given by improving the convergence strategy. The experiment shows that NOLC can accurately detect the target and greatly suppress the background, and the advantages of the NOLC method in detection efficiency and computational efficiency are verified.

Saccades predict and synchronize to visual rhythms irrespective of musical beats

Music has been shown to entrain movement. One of the body’s most frequent movements, saccades, are arguably subject to a timer that may also be susceptible to musical entrainment. We developed a continuous and highly-controlled visual search task and varied the timing of the search target presentation, it was either gaze-contingent, tap-contingent, or visually-timed. We found: (1) explicit control of saccadic timing is limited to gross duration variations and imprecisely synchronized; (2) saccadic timing does not implicitly entrain to musical beats, even when closely aligned in phase; (3) eye movements predict visual onsets produced by motor-movements (finger-taps) and externally-timed sequences, beginning fixation prior to visual onset; (4) eye movement timing can be rhythmic, synchronizing to both motor-produced and externally timed visual sequences; each unaffected by musical beats. These results provide evidence that saccadic timing is sensitive to the temporal demands of visual tasks and impervious to influence from musical beats.

What Do Different Evaluation Metrics Tell Us About Saliency Models?

How best to evaluate a saliency model's ability to predict where humans look in images is an open research question. The choice of evaluation metric depends on how saliency is defined and how the ground truth is represented. Metrics differ in how they rank saliency models, and this results from how false positives and false negatives are treated, whether viewing biases are accounted for, whether spatial deviations are factored in, and how the saliency maps are pre-processed. In this paper, we provide an analysis of 8 different evaluation metrics and their properties. With the help of systematic experiments and visualizations of metric computations, we add interpretability to saliency scores and more transparency to the evaluation of saliency models. Building off the differences in metric properties and behaviors, we make recommendations for metric selections under specific assumptions and for specific applications.

Task instructions can accelerate the early preference for social features in naturalistic scenes

Previous research demonstrated that humans rapidly and reflexively prioritize social features (especially heads and faces) irrespective of their physical saliency when freely viewing naturalistic scenes. In the current study, we investigated whether this preference for social elements already occurs maximally fast during free exploration or whether it is possible to additionally accelerate it by means of top-down instructions. To examine this question, we presented participants with colour photographs of naturalistic scenes containing social features (e.g. heads and bodies) while recording their eye movements. For half of the stimuli, observers were instructed to freely view the images; for the other half of the stimuli, their task was to spot depicted people as fast as possible. We replicated that social elements (especially heads) were rapidly preferred over physically salient image parts. Moreover, we found the orienting towards social elements to be additionally enhanced and accelerated when participants were instructed to detect people quickly. Importantly, this effect was strongest for heads and already evident at the very first fixation. Thus, the present study not only corroborates that the prioritization of social features in naturalistic scenes partially relies on reflexive processes, but also demonstrates that these mechanisms can be additionally accelerated by top-down instructions.

Exploiting Color Volume and Color Difference for Salient Region Detection

Foreground and background cues can assist humans in quickly understanding visual scenes. In computer vision, however, it is difficult to detect salient objects when they touch the image boundary. Hence, detecting salient objects robustly under such circumstances without sacrificing precision and recall can be challenging. In this paper, we propose a novel model for salient region detection, namely, the foreground-center-background (FCB) saliency model. Its main highlights as follows. First, we use regional color volume as the foreground, together with perceptually uniform color differences within regions to detect salient regions. This can highlight salient objects robustly, even when they touched the image boundary, without greatly sacrificing precision and recall. Second, we employ center saliency to detect salient regions together with foreground and background cues, which improves saliency detection performance. Finally, we propose a novel and simple yet efficient method that combines foreground, center, and background saliency. Experimental validation with three well-known benchmark data sets indicates that the FCB model outperforms several state-of-the-art methods in terms of precision, recall, F-measure, and particularly, the mean absolute error. Salient regions are brighter than those of some existing state-of-the-art methods.

High Precision Detection of Salient Objects Based on Deep Convolutional Networks with Proper Combinations of Shallow and Deep Connections

In this paper, a high precision detection method of salient objects is presented based on deep convolutional networks with proper combinations of shallow and deep connections. In order to achieve better performance in the extraction of deep semantic features of salient objects, based on a symmetric encoder and decoder architecture, an upgrade of backbone networks is carried out with a transferable model on the ImageNet pre-trained ResNet50. Moreover, by introducing shallow and deep connections on multiple side outputs, feature maps generated from various layers of the deep neural network (DNN) model are well fused so as to describe salient objects from local and global aspects comprehensively. Afterwards, based on a holistically nested edge detector (HED) architecture, multiple fused side outputs with various sizes of receptive fields are integrated to form detection results of salient objects accordingly. A series of experiments and assessments on extensive benchmark datasets demonstrate the dominant performance of our DNN model for the detection of salient objects in accuracy, which has outperformed those of other published works.

Predicting eye movement patterns from fMRI responses to natural scenes

Eye tracking has long been used to measure overt spatial attention, and computational models of spatial attention reliably predict eye movements to natural images. However, researchers lack techniques to noninvasively access spatial representations in the human brain that guide eye movements. Here, we use functional magnetic resonance imaging (fMRI) to predict eye movement patterns from reconstructed spatial representations evoked by natural scenes. First, we reconstruct fixation maps to directly predict eye movement patterns from fMRI activity. Next, we use a model-based decoding pipeline that aligns fMRI activity to deep convolutional neural network activity to reconstruct spatial priority maps and predict eye movements in a zero-shot fashion. We predict human eye movement patterns from fMRI responses to natural scenes, provide evidence that visual representations of scenes and objects map onto neural representations that predict eye movements, and find a novel three-way link between brain activity, deep neural network models, and behavior.

Attention and Information Acquisition: Comparison of Mouse-Click with Eye-Movement Attention Tracking

Attention is crucial as a fundamental prerequisite for perception. The measurement of attention in viewing and recognizing the images that surround us constitutes an important part of eye movement research, particularly in advertising-effectiveness research. Recording eye and gaze (i.e. eye and head) movements is considered the standard procedure for measuring attention. However, alternative measurement methods have been developed in recent years, one of which is mouse-click attention tracking (mcAT) by means of an on-line based procedure that measures gaze motion via a mouse-click (i.e. a hand and finger positioning maneuver) on a computer screen. Here we compared the validity of mcAT with eye movement attention tracking (emAT). We recorded data in a between subject design via emAT and mcAT and analyzed and compared 20 subjects for correlations. The test stimuli consisted of 64 images that were assigned to eight categories. Our main results demonstrated a highly significant correlation (p < 0.001) between mcAT and emAT data. We also found significant differences in correlations between different image categories. For simply structured pictures of humans or animals in particular, mcAT provided highly valid and more consistent results compared to emAT. We concluded that mcAT is a suitable method for measuring the attention we give to the images that surround us, such as photographs, graphics, art or digital and print advertisements.

Psychophysical evaluation of individual low-level feature influences on visual attention

In this study we provide the analysis of eye movement behavior elicited by low-level feature distinctiveness with a dataset of synthetically-generated image patterns. Design of visual stimuli was inspired by the ones used in previous psychophysical experiments, namely in free-viewing and visual searching tasks, to provide a total of 15 types of stimuli, divided according to the task and feature to be analyzed. Our interest is to analyze the influences of low-level feature contrast between a salient region and the rest of distractors, providing fixation localization characteristics and reaction time of landing inside the salient region. Eye-tracking data was collected from 34 participants during the viewing of a 230 images dataset. Results show that saliency is predominantly and distinctively influenced by: 1. feature type, 2. feature contrast, 3. temporality of fixations, 4. task difficulty and 5. center bias. This experimentation proposes a new psychophysical basis for saliency model evaluation using synthetic images.

Investigating overt and covert shifts of attention within social naturalistic scenes

Eye-tracking studies on social attention have consistently shown that humans prefer to attend to other human beings. Much less is known about whether a similar preference is also evident in covert attentional processes. To enable a direct comparison, this study examined covert and overt attentional guidance within two different experimental setups using complex naturalistic scenes instead of isolated single features. In the first experiment, a modified version of the dot-probe paradigm served as a measure of covert reflexive attention toward briefly presented scenes containing a social feature in one half of the visual field compared to nonsocial elements in the other while controlling for low-level visual saliency. Participants showed a stable congruency effect with faster reaction times and fewer errors for probes presented on the social side of the scene. In a second experiment, we tracked eye movements for the same set of stimuli while manipulating the presentation time to allow for differentiating reflexive and more sustained aspects of overt attention. Supportive of the first results, analyses revealed a robust preference for social features concerning initial saccade direction as well as fixation allocation. Collectively, these experiments imply preferential processing of social features over visually salient aspects for automatic allocation of covert as well as overt attention.

Systematic assessment of intrinsic factors influencing visual attention performances in air traffic control via clustering algorithm and statistical inference

The intrinsic factors (IF) influencing visual attention performance (VAP) might cause potential human errors, such as “error/mistake”, “forgetting” and “omission”. It is a key issue to develop a systematic assessment of IF in order to distinguish the levels of VAP. Motivated by the Stimulus-Response (S-R) model, we take an interactive cancellation test—Neuron Type Test (NTT)—to explore the IF and present the corresponding systematic assessment. The main contributions of this work include three elements: a) modeling the IF on account of attention span, attention stability, distribution-shift of attention with measurable parameters by combining the psychological and statistical concepts; b) proposing quantitative analysis methods for assessing the IF via its computational representation—intrinsic qualities (IQ)—in the sense of computational model; and c) clustering the IQ of air traffic control (ATC) students in the feature space of interest. The response sequences of participants collected with the NTT system are characterized by three parameters: Hurst exponent, normalized number of decisions (NNoD) and error rate of decisions (ERD). The K-means clustering is applied to partition the feature space constructed from practical data of VAP. For the distinguishable clusters, the statistical inference is utilized to refine the assessment of IF. Our comprehensive analysis shows that the IQ can be classified into four levels, i.e., excellent, good, moderate and unqualified, which has a potential application in selecting air traffic controllers subject to reducing the risk of the inadequacy of attention performances in aviation safety management.

A video-driven model of response statistics in the primate middle temporal area

Neurons in the primate middle temporal area (MT) encode information about visual motion and binocular disparity. MT has been studied intensively for decades, so there is a great deal of information in the literature about MT neuron tuning. In this study, our goal is to consolidate some of this information into a statistical model of the MT population response. The model accepts arbitrary stereo video as input. It uses computer-vision methods to calculate known correlates of the responses (such as motion velocity), and then predicts activity using a combination of tuning functions that have previously been used to describe data in various experiments. To construct the population response, we also estimate the distributions of many model parameters from data in the electrophysiology literature. We show that the model accounts well for a separate dataset of MT speed tuning that was not used in developing the model. The model may be useful for studying relationships between MT activity and behavior in ethologically relevant tasks. As an example, we show that the model can provide regression targets for internal activity in a deep convolutional network that performs a visual odometry task, so that its representations become more physiologically realistic.

Synthetic Aperture Radar Processing Approach for Simultaneous Target Detection and Image Formation

Finding out interested targets from synthetic aperture radar (SAR) imagery is an attractive but challenging problem in SAR application. Traditional target detection is independent on SAR imaging process, which is purposeless and unnecessary. Hence, a new SAR processing approach for simultaneous target detection and image formation is proposed in this paper. This approach is based on SAR imagery formation in time domain and human visual saliency detection. First, a series of sub-aperture SAR images with resolutions from low to high are generated by the time domain SAR imaging method. Then, those multiresolution SAR images are detected by the visual saliency processing, and the corresponding intermediate saliency maps are obtained. The saliency maps are accumulated until the result with a sufficient confidence level. After some screening operations, the target regions on the imaging scene are located, and only these regions are focused with full aperture integration. Finally, we can get the SAR imagery with high-resolution detected target regions but low-resolution clutter background. Experimental results have shown the superiority of the proposed approach for simultaneous target detection and image formation.

Neurophysiological Profile of Antismoking Campaigns

Over the past few decades, antismoking public service announcements (PSAs) have been used by governments to promote healthy behaviours in citizens, for instance, against drinking before the drive and against smoke. Effectiveness of such PSAs has been suggested especially for young persons. By now, PSAs efficacy is still mainly assessed through traditional methods (questionnaires and metrics) and could be performed only after the PSAs broadcasting, leading to waste of economic resources and time in the case of Ineffective PSAs. One possible countermeasure to such ineffective use of PSAs could be promoted by the evaluation of the cerebral reaction to the PSA of particular segments of population (e.g., old, young, and heavy smokers). In addition, it is crucial to gather such cerebral activity in front of PSAs that have been assessed to be effective against smoke (Effective PSAs), comparing results to the cerebral reactions to PSAs that have been certified to be not effective (Ineffective PSAs). The eventual differences between the cerebral responses toward the two PSA groups will provide crucial information about the possible outcome of new PSAs before to its broadcasting. This study focused on adult population, by investigating the cerebral reaction to the vision of different PSA images, which have already been shown to be Effective and Ineffective for the promotion of an antismoking behaviour. Results showed how variables as gender and smoking habits can influence the perception of PSA images, and how different communication styles of the antismoking campaigns could facilitate the comprehension of PSA's message and then enhance the related impact.

Finding any Waldo with zero-shot invariant and efficient visual search

Searching for a target object in a cluttered scene constitutes a fundamental challenge in daily vision. Visual search must be selective enough to discriminate the target from distractors, invariant to changes in the appearance of the target, efficient to avoid exhaustive exploration of the image, and must generalize to locate novel target objects with zero-shot training. Previous work on visual search has focused on searching for perfect matches of a target after extensive category-specific training. Here, we show for the first time that humans can efficiently and invariantly search for natural objects in complex scenes. To gain insight into the mechanisms that guide visual search, we propose a biologically inspired computational model that can locate targets without exhaustive sampling and which can generalize to novel objects. The model provides an approximation to the mechanisms integrating bottom-up and top-down signals during search in natural scenes.

Meaning Guides Attention during Real-World Scene Description

Intelligent analysis of a visual scene requires that important regions be prioritized and attentionally selected for preferential processing. What is the basis for this selection? Here we compared the influence of meaning and image salience on attentional guidance in real-world scenes during two free-viewing scene description tasks. Meaning was represented by meaning maps capturing the spatial distribution of semantic features. Image salience was represented by saliency maps capturing the spatial distribution of image features. Both types of maps were coded in a format that could be directly compared to maps of the spatial distribution of attention derived from viewers' eye fixations in the scene description tasks. The results showed that both meaning and salience predicted the spatial distribution of attention in these tasks, but that when the correlation between meaning and salience was statistically controlled, only meaning accounted for unique variance in attention. The results support theories in which cognitive relevance plays the dominant functional role in controlling human attentional guidance in scenes. The results also have practical implications for current artificial intelligence approaches to labeling real-world images.

Find who to look at: Turning from action to saliency

The past decade has witnessed the use of highlevel features in saliency prediction for both videos and images. Unfortunately, the existing saliency prediction methods only handle high-level static features, such as face. In fact, high-level dynamic features (also called actions), such as speaking or head turning, are also extremely attractive to visual attention in videos. Thus, in this paper, we propose a data-driven method for learning to predict the saliency of multiple-face videos, by leveraging both static and dynamic features at high-level. Specifically, we introduce an eye-tracking database, collecting the fixations of 39 subjects viewing 65 multiple-face videos. Through analysis on our database, we find a set of high-level features that cause a face to receive extensive visual attention. These high-level features include the static features of face size, center-bias and head pose, as well as the dynamic features of speaking and head turning. Then, we present the techniques for extracting these high-level features. Afterwards, a novel model, namely multiple hidden Markov model (M-HMM), is developed in our method to enable the transition of saliency among faces. In our MHMM, the saliency transition takes into account both the state of saliency at previous frames and the observed high-level features at the current frame. The experimental results show that the proposed method is superior to other state-of-the-art methods in predicting visual attention on multiple-face videos. Finally, we shed light on a promising implementation of our saliency prediction method in locating the region-of-interest (ROI), for video conference compression with high efficiency video coding (HEVC).

Visual attention and its intimate links to spatial cognition

It is almost universal to regard attention as the facility that permits an agent, human or machine, to give priority processing resources to relevant stimuli while ignoring the irrelevant. The reality of how this might manifest itself throughout all the forms of perceptual and cognitive processes possessed by humans, however, is not as clear. Here, we examine this reality with a broad perspective in order to highlight the myriad ways that attentional processes impact both perception and cognition. The paper concludes by showing two real-world problems that exhibit sufficient complexity to illustrate the ways in which attention and cognition connect. These then point to new avenues of research that might illuminate the overall cognitive architecture of spatial cognition.

Integrating Gaze Tracking and Head-Motion Prediction for Mobile Device Authentication: A Proof of Concept

We introduce a two-stream model to use reflexive eye movements for smart mobile device authentication. Our model is based on two pre-trained neural networks, iTracker and PredNet, targeting two independent tasks: (i) gaze tracking and (ii) future frame prediction. We design a procedure to randomly generate the visual stimulus on the screen of mobile device, and the frontal camera will simultaneously capture head motions of the user as one watches it. Then, iTracker calculates the gaze-coordinates error which is treated as a static feature. To solve the imprecise gaze-coordinates caused by the low resolution of the frontal camera, we further take advantage of PredNet to extract the dynamic features between consecutive frames. In order to resist traditional attacks (shoulder surfing and impersonation attacks) during the procedure of mobile device authentication, we innovatively combine static features and dynamic features to train a 2-class support vector machine (SVM) classifier. The experiment results show that the classifier achieves accuracy of 98.6% to authenticate the user identity of mobile devices.

Attentional Selection of Social Features Persists Despite Restricted Bottom-Up Information and Affects Temporal Viewing Dynamics

Previous studies have shown an attentional bias towards social features during free-viewing of naturalistic scenes. This social attention seems to be reflexive and able to defy top-down demands in form of explicit search tasks. However, the question remains whether social features continue to be prioritized when peripheral information is limited, thereby reducing the influence of bottom-up image information on gaze orienting. Therefore, we established a gaze-contingent viewing paradigm, in which the visual field was constrained and updated in response to the viewer's eye movements. Participants viewed social and non-social images that were randomly allocated to a free and a gaze-contingent viewing condition while their eye movements were tracked. Our results revealed a strong attentional bias towards social features in both conditions. However, gaze-contingent viewing altered temporal and spatial dynamics of viewing behavior. Additionally, recurrent fixations were more frequent and closer together in time for social compared to non-social stimuli in both viewing conditions. Taken together, this study implies a predominant selection of social features when bottom-up influences are diminished and a general influence of social content on visual exploratory behavior, thus highlighting mechanisms of social attention.

Deadeye: A Novel Preattentive Visualization Technique Based on Dichoptic Presentation

Preattentive visual features such as hue or flickering can effectively draw attention to an object of interest - for instance, an important feature in a scientific visualization. These features appear to pop out and can be recognized by our visual system, independently from the number of distractors. Most cues do not take advantage of the fact that most humans have two eyes. In cases where binocular vision is applied, it is almost exclusively used to convey depth by exposing stereo pairs. We present Deadeye, a novel preattentive visualization technique based on presenting different stimuli to each eye. The target object is rendered for one eye only and is instantly detected by our visual system. In contrast to existing cues, Deadeye does not modify any visual properties of the target and, thus, is particularly suited for visualization applications. Our evaluation confirms that Deadeye is indeed perceived preattentively. We also explore a conjunction search based on our technique and show that, in contrast to 3D depth, the task cannot be processed in parallel.