Remote Eye Tracker Research Articles

Pupillary dilation response to the auditory food words in adolescents with obesity without binge eating disorder

Childhood obesity is a growing global public health problem. Studies suggest that environmental cues contribute to developing and maintaining obesity. We aimed to evaluate pupillary changes to auditory food words vs. nonfood words and to conduct a dynamic temporal analysis of pupil size changes in adolescents with obesity without binge eating disorder by comparing healthy-weight adolescents. In this study, a total of 63 adolescents aged 12–18 years (n = 32, obesity group (OG); n = 31, control group (CG)) were included. In an auditory paradigm, participants were presented with a series of high and low-calorie food and nonfood words. A binocular remote eye-tracking device was used to measure pupil diameter. Generalized additive mixed models (GAMMs) were used for dynamic temporal analysis of pupillometry data. The results of GAMM analysis indicated that CG had larger pupil dilation than the OG while listening to auditory food words. CG had larger pupil dilation in food words than in nonfood words. However, the OG had a similar pupillary response in food and nonfood words. Pupil dilation response to higher-calorie foods was extended over the later stages of the time period (after 2000 ms) in the OG. In summary, our findings indicated that individuals with obesity had lower pupil dilation to auditory food words compared to normal-weight peers. Adolescents with obesity had prolonged pupillary dilation in higher calories of food words. The individual psychological factors affecting the dynamic changes of pupil responses to food cues in adolescents with obesity should be examined in further studies.

Eye Movement Differences in Contact Versus Non-Contact Olympic Athletes

The purpose of this study was to investigate the difference in oculomotor functioning between Olympic-level contact and non-contact sports participants. In total, 67 male and female Olympic-level contact (n = 27) and non-contact (n = 40) athletes completed oculomotor tasks, including Horizontal Saccade (HS), Circular Smooth Pursuit (CSP), Horizontal Smooth Pursuit (HSP), and Vertical Smooth Pursuit (VSP) using a remote eye tracker. No significant differences for sex or age occurred. Each variable indicated higher scores for contact compared to non-contact athletes (p < .05) except for VSP Pathway differences and CSP Synchronization. A logistic regression was performed to determine the degree that HS measures, CSP synchronization, and VSP pathway predicted sport type. The model was significant, χ2 (6) = 37.08, p < .001, explaining 57.4% of the variance and correctly classified 88.1% of cases. The sensitivity was 87.5% and specificity was 88.9%. CSP synchronization did not increase the likelihood of participating in a contact sport. This was the first study to identify oculomotor differences between Olympic athletes of contact and non-contact sports, which adds to the growing evidence that oculomotor functioning may be a reliable, quick, real-time tool to help detect mTBI in sport.

Visual Attention to Evolutionarily Relevant Information by Heterosexual Men and Women While Viewing Mock Online Dating Profiles

The way people create social connections and access information has been altered greatly by technology in recent decades. Online browsing of visual profiles has become a common means for seeking potential partners for both short- and long-term relationships. Little is known, however, about how people prioritize mate quality information while viewing online profiles. Using eye-tracking methods and self-report, this study investigated how people evaluated profile-based facial attractiveness and text-based financial resources information, represented by income and occupation. Heterosexual male and female participants, aged between 18 and 27 years, viewed opposite-sex profiles while their eye-movements were recorded using a remote eye-tracking camera. In line with current theory, resources information had little effect on men’s overall attention to women’s faces, whereas women’s overall attention to men’s faces varied depending on the level of income and occupation. Women evaluated men’s faces more when income and occupation were low, regardless of attractiveness. Unexpectedly, however, men marginally increased their attention toward unattractive women who showed a high-level of income and more esteemed occupation. Men self-reported a higher interest in women for a short-term relationship and women self-reported a higher interest in men for a long-term relationship. This work provides a foundation to further examine how people browse profile-based information and to investigate the mate selection process, with real-world implications for online dating app users, profile design, and content.

Strategies for enhancing automatic fixation detection in head-mounted eye tracking

Moving through a dynamic world, humans need to intermittently stabilize gaze targets on their retina to process visual information. Overt attention being thus split into discrete intervals, the automatic detection of such fixation events is paramount to downstream analysis in many eye-tracking studies. Standard algorithms tackle this challenge in the limiting case of little to no head motion. In this static scenario, which is approximately realized for most remote eye-tracking systems, it amounts to detecting periods of relative eye stillness. In contrast, head-mounted eye trackers allow for experiments with subjects moving naturally in everyday environments. Detecting fixations in these dynamic scenarios is more challenging, since gaze-stabilizing eye movements need to be reliably distinguished from non-fixational gaze shifts. Here, we propose several strategies for enhancing existing algorithms developed for fixation detection in the static case to allow for robust fixation detection in dynamic real-world scenarios recorded with head-mounted eye trackers. Specifically, we consider (i) an optic-flow-based compensation stage explicitly accounting for stabilizing eye movements during head motion, (ii) an adaptive adjustment of algorithm sensitivity according to head-motion intensity, and (iii) a coherent tuning of all algorithm parameters. Introducing a new hand-labeled dataset, recorded with the Pupil Invisible glasses by Pupil Labs, we investigate their individual contributions. The dataset comprises both static and dynamic scenarios and is made publicly available. We show that a combination of all proposed strategies improves standard thresholding algorithms and outperforms previous approaches to fixation detection in head-mounted eye tracking.

Pre-service and in-service teachers' professional vision depending on the video perspective—What teacher gaze and verbal reports can tell us

Teachers are involved in complex teaching situations every day; thus, they must understand what to pay attention to in the classroom, how this information is to be interpreted, and which teaching decisions become necessary as a result. In educational research, these competencies are known as “professional vision.” The purpose of this exploratory study was to examine the professional vision of pre-service teachers (PTs) and in-service teachers (ITs) by investigating whether the groups differ in what they notice and how they reason about videotaped classroom events; whether the perspective of the video viewed influences their noticing and reasoning; and to what extent their gaze behavior differs from their verbal statements. Thirty-one PTs and twenty ITs watched a video clip of authentic teaching, shot from different perspectives, and their visual focus of attention was recorded using a remote eye-tracker. Subsequently, participants reported in an interview what they had noticed. The triangulated data show that the gaze behavior of the PTs and ITs did not differ, but the content of their verbal statements did. Depending on the video perspective, participants focused on different subjects, but this difference was not reflected in the verbal data. Thus, the gaze behavior and verbal statements are not consistent. The findings indicate that considering multiple sources and types of data is beneficial to explore professional vision and that further research is needed to understand the concept in depth.

Deep learning models for webcam eye tracking in online experiments

Eye tracking is prevalent in scientific and commercial applications. Recent computer vision and deep learning methods enable eye tracking with off-the-shelf webcams and reduce dependence on expensive, restrictive hardware. However, such deep learning methods have not yet been applied and evaluated for remote, online psychological experiments. In this study, we tackle critical challenges faced in remote eye tracking setups and systematically evaluate appearance-based deep learning methods of gaze tracking and blink detection. From their own homes and laptops, 65 participants performed a battery of eye tracking tasks including (i) fixation, (ii) zone classification, (iii) free viewing, (iv) smooth pursuit, and (v) blink detection. Webcam recordings of the participants performing these tasks were processed offline through appearance-based models of gaze and blink detection. The task battery required different eye movements that characterized gaze and blink prediction accuracy over a comprehensive list of measures. We find the best gaze accuracy to be 2.4° and precision of 0.47°, which outperforms previous online eye tracking studies and reduces the gap between laboratory-based and online eye tracking performance. We release the experiment template, recorded data, and analysis code with the motivation to escalate affordable, accessible, and scalable eye tracking that has the potential to accelerate research in the fields of psychological science, cognitive neuroscience, user experience design, and human–computer interfaces.

Analysis of eye movements to study drawing in the context of vector fields

Research has shown that visual representations can substantially enhance the learning and understanding of STEM concepts; despite this, students tend to struggle in using them fluently and consistently. Consequently, educators advocate for explicit instructions that support the coordination of multiple representations, especially when concepts become more abstract and complex. For recent years, the drawing (or sketching) technique has received increasing attention. Theoretical considerations and prior research suggest that drawing has the potential to support knowledge construction and to provide cognitive relief. In this article, we present two studies that investigate the impact of drawing activities in a multi-representational, instruction-based learning scenario from physics, more precisely, in the context of vector fields. Further, mobile and remote eye tracking was used to record students' gaze behavior in addition to monitoring indicators of performance and cognitive load. Here, eye movements provide information about cognitive processes during the completion of the instruction, on the one hand, and during subsequent problem solving, on the other hand. Comparisons of a treatment group instructed with drawing activities and a control group instructed without drawing activities revealed significant differences in students' perceived cognitive load (p = 0.02, d = 0.47 and p = 0.0045, d = 0.37), as well as their response accuracy (p = 0.02, d = 0.51) and their response confidence (p = 0.02, d = 0.55 and p = 0.004, d = 0.64) during assessment after instruction (N = 84). Moreover, students instructed with drawing activities were found to distribute more visual attention to important parts of the instruction (vector field diagram and instructional text, N = 32) compared to the control group and, further, showed effective, expert-like behaviors during subsequent problem solving (N = 53). Finally, as a contribution to current trends in eye-tracking research, the application of mobile and remote eye-tracking in drawing-based learning and assessment scenarios is compared and critically discussed.

Let's get it started: Eye Tracking in VR with the Pupil Labs Eye Tracking Add-On for the HTC Vive.

Combining eye tracking and virtual reality (VR) is a promising approach to tackle various applied research questions. As this approach is relatively new, routines are not established yet and the first steps can be full of potential pitfalls. The present paper gives a practice example to lower the boundaries for getting started. More specifically, I focus on an affordable add-on technology, the Pupil Labs eye tracking add-on for the HTC Vive. As add-on technology with all relevant source code available on GitHub, a high degree of freedom in preprocessing, visualizing, and analyzing eye tracking data in VR can be achieved. At the same time, some extra preparatory steps for the setup of hardware and software are necessary. Therefore, specifics of eye tracking in VR from unboxing, software integration, and procedures to analyzing the data and maintaining the hardware will be addressed. The Pupil Labs eye tracking add-on for the HTC Vive represents a highly transparent approach to existing alternatives. Characteristics of eye tracking in VR in contrast to other headmounded and remote eye trackers applied in the physical world will be discussed. In conclusion, the paper contributes to the idea of open science in two ways: First, by making the necessary routines transparent and therefore reproducible. Second, by stressing the benefits of using open source software.

Examining Visual Attention to Tobacco Marketing Materials Among Young Adult Smokers: Protocol for a Remote Webcam-Based Eye-Tracking Experiment.

Eye tracking provides an objective way to measure attention, which can advance researchers' and policy makers' understanding of tobacco marketing influences. The development of remote webcam-based eye-tracking technology, integrated with web-based crowdsourcing studies, may be a cost-effective and time-efficient alternative to laboratory-based eye-tracking methods. However, research is needed to evaluate the utility of remote eye-tracking methods. This study aimed to detail the process of designing a remote webcam-based eye-tracking experiment and provide data on associations between participant characteristics and the outcomes of experiment completion. A total of 2023 young adult (aged 18-34 years) cigarette smokers in the United States were recruited to complete a web-based survey that included a 90-second remote eye-tracking experiment that examined attention to e-cigarette marketing materials. Primary outcome measures assessed the completion of the remote eye-tracking experiment-specifically, experiment initiated versus not initiated, experiment completed versus not completed, and usable versus nonusable eye-tracking data generated. Multivariable logistic regressions examined the associations between outcome measures and participants' sociodemographic backgrounds, tobacco use history, and electronic devices (mobile vs desktop) used during the experiment. Study recruitment began on April 14, 2022, and ended on May 3, 2022. Of the 2023 survey participants, 1887 (93.28%) initiated the experiment, and 777 (38.41%) completed the experiment. Of the 777 participants who completed the experiment, 381 (49%) generated usable data. Results from the full regression models show that non-Hispanic Black participants (adjusted odds ratio [AOR] 0.64, 95% CI 0.45-0.91) were less likely to complete the eye-tracking experiment than non-Hispanic White participants. In addition, female (vs male) participants (AOR 1.46, 95% CI 1.01-2.11), those currently using (vs not using) e-cigarettes (AOR 2.08, 95% CI 1.13-3.82), and those who used mobile (vs desktop) devices (AOR 5.10, 95% CI 3.05-8.52) were more likely to generate usable eye-tracking data. Young adult participants were willing to try remote eye-tracking technology, and nearly half of those who completed the experiment generated usable eye-tracking data (381/777, 49%). Thus, we believe that the use of remote eye-tracking tools, integrated with crowdsourcing recruitment, can be a useful approach for the tobacco regulatory science research community to collect high-quality, large-scale eye-tracking data in a timely fashion and thereby address research questions related to the ever-evolving tobacco marketing landscape. It would be useful to investigate techniques to enhance completion rates and data usability. RR1-10.2196/43512.

Maximizing valid eye-tracking data in human and macaque infants by optimizing calibration and adjusting areas of interest.

Remote eye tracking with automated corneal reflection provides insights into the emergence and development of cognitive, social, and emotional functions in human infants and non-human primates. However, because most eye-tracking systems were designed for use in human adults, the accuracy of eye-tracking data collected in other populations is unclear, as are potential approaches to minimize measurement error. For instance, data quality may differ across species or ages, which are necessary considerations for comparative and developmental studies. Here we examined how the calibration method and adjustments to areas of interest (AOIs) of the Tobii TX300 changed the mapping of fixations to AOIs in a cross-species longitudinal study. We tested humans (N = 119) at 2, 4, 6, 8, and 14 months of age and macaques (Macaca mulatta; N = 21) at 2 weeks, 3 weeks, and 6 months of age. In all groups, we found improvement in the proportion of AOI hits detected as the number of successful calibration points increased, suggesting calibration approaches with more points may be advantageous. Spatially enlarging and temporally prolonging AOIs increased the number of fixation-AOI mappings, suggesting improvements in capturing infants' gaze behaviors; however, these benefits varied across age groups and species, suggesting different parameters may be ideal, depending on the population studied. In sum, to maximize usable sessions and minimize measurement error, eye-tracking data collection and extraction approaches may need adjustments for the age groups and species studied. Doing so may make it easier to standardize and replicate eye-tracking research findings.

Reduced eye gaze during facial emotion recognition in chronic depression: Effects of intranasal oxytocin

Chronic depression disorders (CDD) are characterized by impaired social cognitive functioning. Visual attention during social perception is altered in clinical depression and is known to be sensitive to intranasal treatment with oxytocin (OT). The present study thus investigated potential alterations in gaze preferences during a standardized facial emotion recognition (FER) task using remote eye tracking in patients with CDD and the effect of a single dose of intranasal OT (compared to placebo). In emotion recognition, CDD patients were not more impaired than healthy controls, and there was no OT effect. However, CDD patients (with placebo) demonstrated less attentional preference for the eye region during FER than healthy controls, which was not apparent in the CDD group after OT treatment. Our results suggest that despite largely preserved basic facial emotions recognition, attention in social perception may be altered in CDD, and that this bias may be sensitive to OT treatment. These findings highlight OTs potential as a means of augmenting psychotherapy.

Eye-Tracking Image Encoding: Autoencoders for the Crossing of Language Boundaries in Developmental Dyslexia Detection

The negative influence of developmental dyslexia on academic performance is a well-documented and researched topic. Although the research focused on developmental dyslexia detection and evaluation is plentiful, the study designs vary to a great degree, making the exchange of obtained knowledge often difficult. This paper focuses on bridging the gap between different study designs by developing a machine learning based pipeline that was evaluated on two completely different eye-tracking datasets (training on one, testing on the other, and vice versa). One dataset included 30 subjects who read text written in Serbian on different color configurations and were tracked with a remote eye-tracker. The second dataset included 185 subjects who read text written in Swedish and recorded eye-tracking data using a goggle-based system. The data from both datasets were converted to grayscale images, using various time window configurations to parse the signals, and to plot the data in a 2D plane. The train images were used to train an Autoencoder neural network, and the images’ reconstruction error was used to create features that describe each instance of both the training and test sets. The train feature set was used to train various machine learning algorithms, which were then evaluated on the testing feature dataset. A classification accuracy of 85.6% was obtained when testing on Serbian readers’ data and 82.9% when testing on Swedish readers. The proposed pipeline was shown to be transferable between the datasets, despite many differences in the experiment design, showing potential in combining various eye-tracking dyslexia studies.

Gaze-contingent display technology can help to reduce the ipsilesional attention bias in hemispatial neglect following stroke

BackgroundHemispatial neglect results from unilateral brain damage and represents a disabling unawareness for objects in the hemispace opposite the brain lesion (contralesional). The patients’ attentional bias for ipsilesional hemispace represents a hallmark of neglect, which results from an imbalanced attentional priority map in the brain. The aim of this study was to investigate whether gaze-contingent display (GCD) technology, reducing the visual salience of objects in ipsilesional hemispace, is able to rebalance this map and increase awareness and exploration of objects in the neglected contralesional hemispace.MethodsUsing remote eye-tracking, we recorded gaze positions in 19 patients with left hemispatial neglect following right-hemisphere stroke and 22 healthy control subjects, while they were watching static naturalistic scenes. There were two task conditions, free viewing (FV) or goal-directed visual search (VS), and four modification conditions including the unmodified original picture, a purely static modification and two differently strong modifications with an additional gaze-contingent mask (GC-LOW, GC-HIGH), that continuously reduced color saturation and contrast of objects in the right hemispace.ResultsThe patients’ median gaze position (Center of Fixation) in the original pictures was markedly deviated to the right in both tasks (FV: 6.8° ± 0.8; VS: 5.5° ± 0.7), reflecting the neglect-typical ipsilesional attention bias. GC modification significantly reduced this bias in FV (GC-HIGH: d = − 3.2 ± 0.4°; p < 0.001). Furthermore, in FV and VS, GC modification increased the likelihood to start visual exploration in the (neglected) left hemifield by about 20%. This alleviation of the ipsilesional fixation bias was not associated with an improvement in detecting left-side targets, in contrast, the GC mask even decreased and slowed the detection of right-side targets. Subjectively, patients found the intervention pleasant and most of the patients did not notice any modification.ConclusionsGCD technology can be used to positively influence visual exploration patterns in patients with hemispatial neglect. Despite an alleviation of the neglect-related ipsilesional fixation bias, a concomitant functional benefit (improved detection of contralesional targets) was not achieved. Future studies may investigate individualized GCD-based modifications as augmented reality applications during the activities of daily living.

A toolkit for wide-screen dynamic area of interest measurements using the Pupil Labs Core Eye Tracker

Eye tracking measurements taken while watching a wide field screen are challenging to perform. Commercially available remote eye trackers typically do not measure more than 35 degrees in eccentricity. Analysis software was developed using the Pupil Core Eye Tracking data to analyze viewing behavior under circumstances as natural as possible, on a 1.55-m-wide screen allowing free head movements. Additionally, dynamic area of interest (AOI) analyses were performed on data of participants viewing traffic scenes. A toolkit was created including software for simple allocation of dynamic AOIs (semi-automatically and manually), measurement of parameters such as dwell times and time to first entry, and overlaying gaze and AOIs on video. Participants (n =11) were asked to look at 13 dynamic AOIs in traffic scenes from appearance to disappearance in order to validate the setup and software. Different AOI margins were explored for the included objects. The median ratio between total appearance time and dwell time was about 90% for most objects when appropriate margins were chosen. This validated open-source toolkit is readily available for researchers who want to perform dynamic AOI analyses with the Pupil Core eye tracker, especially when measurements are desired on a wide screen, in various fields such as psychology, transportation, and low vision research.

Comparison of image-based quantification methods in evaluating fixation stability using a remote eye tracker in abnormal phoria.

ObjectiveThis study was performed to establish a quantitative evaluation and comparison of fixation stability, as measured by an eye tracker, using image-based areas determined by the bivariate contour ellipse area (BCEA), kernel density estimation (KDE), and Scanpath methods.MethodsThis prospective cross-sectional study included 45 and 20 participants with abnormal and normal phoria, respectively. Eye movements were recorded using a remote eye tracker and were plotted using RStudio software. Image-based areas were evaluated using ImageJ software.ResultsThe image-based areas used to evaluate fixation stability exhibited decreasing stability in the abnormal phoria group in the following order: KDE with ±1 standard deviation (SD), BCEA with ±1 SD, KDE with ±2 SD or Scanpath, and BCEA with ±2 SD. The BCEA tended to be overestimated, and the KDE tended to be underestimated at high density. The Scanpath method had a very high probability area because the area spans all gaze points.ConclusionsFixation stability could be quantified as image-based areas by the KDE, BCEA, and Scanpath methods. Our findings suggest that fixation stability may be evaluated using one or more methods.

Quantitative Evaluation of the Association Between Fixation Stability and Phoria During Short-Term Binocular Viewing.

PurposeFixation stability for binocular anomalies with a phoria cannot be detected by direct observations. This study aimed to quantitatively evaluate fixation stability using an eye tracker rather than direct directions in binocular vision with abnormal and normal phorias.MethodsThirty-five and 25 participants with abnormal and normal phoria, respectively, were included in the study. The horizontal and vertical gaze points and convergence were recorded for 10 s using a remote eye tracker while binocularly viewing a target on a display screen 550 mm away. Fixation stability was quantified using bivariate contour ellipse areas (BCEA).ResultsThe fixation stability for all participants-based evaluations as a single cluster in the abnormal phoria group was lower than that in the normal phoria group (p = 0.005). There was no difference between the two groups in the evaluation based on the BCEA for each participant-based evaluation (p = 0.66). Fixation stability was also more related to convergence for the abnormal phoria group than for the normal phoria group (r = 0.769, p < 0.001; r = 0.417, p = 0.038, respectively).ConclusionAs the first study to evaluate fixation stability using an eye-tracker to differentiate between abnormal and normal phoria for non-strabismus, these findings may provide evidence for improving the evaluation of binocular vision not detected with clinical diagnostic tests.

Signatures of functional visuospatial asymmetries in early infancy

Adults present a large number of asymmetries in visuospatial behavior that are known to be supported by functional brain lateralization. Although there is evidence of lateralization for motor behavior and language processing in infancy, no study has explored visuospatial attention biases in the early stages of development. In this study, we tested for the presence of a leftward visuospatial bias (i.e., pseudoneglect) in 4- and 5-month-old infants using an adapted version of the line bisection task. Infants were trained to identify the center of a horizontal line (Experiment 1) while their eye gazes were monitored using a remote eye-tracking procedure to measure their potential gazing error. Infants exhibited a robust pseudoneglect, gazing leftward with respect to the veridical midpoint of the horizontal line. To investigate whether infants’ pseudoneglect generalizes to any given object or is dependent on the horizontal dimension, in Experiment 2 we assessed infants’ gaze deployment in vertically oriented lines. No leftward bias was found, suggesting that early visuospatial attention biases in infancy are constrained by the orientation of the visual plane in which the information is organized. The interplay between biological and cultural factors that might contribute to the early establishment of the observed leftward bias in the allocation of visuospatial attention is discussed.

Mobile Eye-Tracking Data Analysis Using Object Detection via YOLO v4.

Remote eye tracking has become an important tool for the online analysis of learning processes. Mobile eye trackers can even extend the range of opportunities (in comparison to stationary eye trackers) to real settings, such as classrooms or experimental lab courses. However, the complex and sometimes manual analysis of mobile eye-tracking data often hinders the realization of extensive studies, as this is a very time-consuming process and usually not feasible for real-world situations in which participants move or manipulate objects. In this work, we explore the opportunities to use object recognition models to assign mobile eye-tracking data for real objects during an authentic students’ lab course. In a comparison of three different Convolutional Neural Networks (CNN), a Faster Region-Based-CNN, you only look once (YOLO) v3, and YOLO v4, we found that YOLO v4, together with an optical flow estimation, provides the fastest results with the highest accuracy for object detection in this setting. The automatic assignment of the gaze data to real objects simplifies the time-consuming analysis of mobile eye-tracking data and offers an opportunity for real-time system responses to the user’s gaze. Additionally, we identify and discuss several problems in using object detection for mobile eye-tracking data that need to be considered.

Quiet Eye and Computerized Precision Tasks in First-Person Shooter Perspective Esport Games

The gaze behavior in sports and other applied settings has been studied for more than 20 years. A common finding is related to the “quiet eye” (QE), predicting that the duration of the last fixation before a critical event is associated with higher performance. Unlike previous studies conducted in applied settings with mobile eye trackers, we investigate the QE in a context similar to esport, in which participants click the mouse to hit targets presented on a computer screen under different levels of cognitive load. Simultaneously, eye and mouse movements were tracked using a high-end remote eye tracker at 300 Hz. Consistent with previous studies, we found that longer QE fixations were associated with higher performance. Increasing the cognitive load delayed the onset of the QE fixation, but had no significant influence on the QE duration. We discuss the implications of our results in the context of how the QE is defined, the quality of the eye-tracker data, and the type of analysis applied to QE data.

Accurate Pupil Center Detection in Off-the-Shelf Eye Tracking Systems Using Convolutional Neural Networks.

Remote eye tracking technology has suffered an increasing growth in recent years due to its applicability in many research areas. In this paper, a video-oculography method based on convolutional neural networks (CNNs) for pupil center detection over webcam images is proposed. As the first contribution of this work and in order to train the model, a pupil center manual labeling procedure of a facial landmark dataset has been performed. The model has been tested over both real and synthetic databases and outperforms state-of-the-art methods, achieving pupil center estimation errors below the size of a constricted pupil in more than 95% of the images, while reducing computing time by a 8 factor. Results show the importance of use high quality training data and well-known architectures to achieve an outstanding performance.