Gaze Mapping Research Articles

Domain adaptation for driver's gaze mapping for different drivers and new environments

Distracted driving is a leading cause of traffic accidents, and often arises from a lack of visual attention on the road. To enhance road safety, monitoring a driver's visual attention is crucial. Appearance-based gaze estimation using deep learning and Convolutional Neural Networks (CNN) has shown promising results, but it faces challenges when applied to different drivers and environments. In this paper, we propose a domain adaptation-based solution for gaze mapping, which aims to accurately estimate a driver's gaze in diverse drivers and new environments. Our method consists of three steps: pre-processing, facial feature extraction, and gaze region classification. We explore two strategies for input feature extraction, one utilizing the full appearance of the driver and environment and the other focusing on the driver's face. Through unsupervised domain adaptation, we align the feature distributions of the source and target domains using a conditional Generative Adversarial Network (GAN). We conduct experiments on the Driver Gaze Mapping (DGM) dataset and the Columbia Cave-DB dataset to evaluate the performance of our method. The results demonstrate that our proposed method reduces the gaze mapping error, achieves better performance on different drivers and camera positions, and outperforms existing methods. We achieved an average Strictly Correct Estimation Rate (SCER) accuracy of 81.38% and 93.53% and Loosely Correct Estimation Rate (LCER) accuracy of 96.69% and 98.9% for the two strategies, respectively, indicating the effectiveness of our approach in adapting to different domains and camera positions. Our study contributes to the advancement of gaze mapping techniques and provides insights for improving driver safety in various driving scenarios.

Estimating 3D spatiotemporal point of regard: a device evaluation.

This paper presents and evaluates a system and method that record spatiotemporal scene information and location of the center of visual attention, i.e., spatiotemporal point of regard (PoR) in ecological environments. A primary research application of the proposed system and method is for enhancing current 2D visual attention models. Current eye-tracking approaches collapse a scene's depth structures to a 2D image, omitting visual cues that trigger important functions of the human visual system (e.g., accommodation and vergence). We combined head-mounted eye-tracking with a miniature time-of-flight camera to produce a system that could be used to estimate the spatiotemporal location of the PoR-the point of highest visual attention-within 3D scene layouts. Maintaining calibration accuracy is a primary challenge for gaze mapping; hence, we measured accuracy repeatedly by matching the PoR to fixated targets arranged within a range of working distances in depth. Accuracy was estimated as the deviation from estimated PoR relative to known locations of scene targets. We found that estimates of 3D PoR had an overall accuracy of approximately 2° omnidirectional mean average error (OMAE) with variation over a 1 h recording maintained within 3.6° OMAE. This method can be used to determine accommodation and vergence cues of the human visual system continuously within habitual environments, including everyday applications (e.g., use of hand-held devices).

Gaze Mapping for Immersive Virtual Environments Based on Image Retrieval

In this paper, we introduce a novel gaze mapping approach for free viewing conditions in dynamic immersive virtual environments (VEs), which projects recorded eye fixation data of users, who viewed the VE from different perspectives, to the current view. This generates eye fixation maps, which can serve as ground truth for training machine learning (ML) models to predict saliency and the user’s gaze in immersive virtual reality (VR) environments. We use a flexible image retrieval approach based on SIFT features, which can also map the gaze under strong viewpoint changes and dynamic changes. A vocabulary tree enables to scale to the large amounts of data with typically several hundred thousand frames and a homography transform re-projects the fixations to the current view. To evaluate our approach, we measure the predictive quality of our eye fixation maps to model the gaze of the current user and compare our maps to computer-generated saliency maps on the DGaze and the Saliency in VR datasets. The results show that our method often outperform these saliency predictors. However, in contrast to these methods, our approach collects real fixations from human observers, and can thus serve to estimate ground truth fixation maps in dynamic VR environments, which can be used to train and evaluate gaze predictors.

Dual-Cameras-Based Driver's Eye Gaze Tracking System with Non-Linear Gaze Point Refinement.

The human eye gaze plays a vital role in monitoring people’s attention, and various efforts have been made to improve in-vehicle driver gaze tracking systems. Most of them build the specific gaze estimation model by pre-annotated data training in an offline way. These systems usually tend to have poor generalization performance during the online gaze prediction, which is caused by the estimation bias between the training domain and the deployment domain, making the predicted gaze points shift from their correct location. To solve this problem, a novel driver’s eye gaze tracking method with non-linear gaze point refinement is proposed in a monitoring system using two cameras, which eliminates the estimation bias and implicitly fine-tunes the gaze points. Supported by the two-stage gaze point clustering algorithm, the non-linear gaze point refinement method can gradually extract the representative gaze points of the forward and mirror gaze zone and establish the non-linear gaze point re-mapping relationship. In addition, the Unscented Kalman filter is utilized to track the driver’s continuous status features. Experimental results show that the non-linear gaze point refinement method outperforms several previous gaze calibration and gaze mapping methods, and improves the gaze estimation accuracy even on the cross-subject evaluation. The system can be used for predicting the driver’s attention.

Development and validation of a new methodological platform to measure behavioral, cognitive, and physiological responses to food interventions in real time

To fully understand the causes and mechanisms involved in overeating and obesity, measures of both cognitive and physiological determinants of eating behavior need to be integrated. Effectively synchronizing behavioral measures such as meal micro-structure (e.g., eating speed), cognitive processing of sensory stimuli, and metabolic parameters, can be complex. However, this step is central to understanding the impact of food interventions on body weight. In this paper, we provide an overview of the existing gaps in eating behavior research and describe the development and validation of a new methodological platform to address some of these issues. As part of a controlled trial, 76 men and women self-served and consumed food from a buffet, using a portion-control plate with visual stimuli for appropriate amounts of main food groups, or a conventional plate, on two different days, in a random order. In both sessions participants completed behavioral and cognitive tests using a novel methodological platform that measured gaze movement (as a proxy for visual attention), eating rate and bite size, memory for portion sizes, subjective appetite and portion-size perceptions. In a sub-sample of women, hormonal secretion in response to the meal was also measured. The novel platform showed a significant improvement in meal micro-structure measures from published data (13 vs. 33% failure rate) and high comparability between an automated gaze mapping protocol vs. manual coding for eye-tracking studies involving an eating test (ICC between methods 0.85; 90% CI 0.74, 0.92). This trial was registered at Clinical Trials.gov with Identifier NCT03610776.

Optimizing Fixation Filters for Eye-Tracking on Small Screens.

The study of consumer responses to advertising has recently expanded to include the use of eye-tracking to track the gaze of consumers. The calibration and validation of eye-gaze have typically been measured on large screens in static, controlled settings. However, little is known about how precise gaze localizations and eye fixations are on smaller screens, such as smartphones, and in moving feed-based conditions, such as those found on social media websites. We tested the precision of eye-tracking fixation detection algorithms relative to raw gaze mapping in natural scrolling conditions. Our results demonstrate that default fixation detection algorithms normally employed by hardware providers exhibit suboptimal performance on mobile phones. In this paper, we provide a detailed account of how different parameters in eye-tracking software can affect the validity and reliability of critical metrics, such as Percent Seen and Total Fixation Duration. We provide recommendations for producing improved eye-tracking metrics for content on small screens, such as smartphones, and vertically moving environments, such as a social media feed. The adjustments to the fixation detection algorithm we propose improves the accuracy of Percent Seen by 19% compared to a leading eye-tracking provider’s default fixation filter settings. The methodological approach provided in this paper could additionally serve as a framework for assessing the validity of applied neuroscience methods and metrics beyond mobile eye-tracking.

4D Attention: Comprehensive Framework for Spatio-Temporal Gaze Mapping

This study presents a framework for capturing human attention in the spatio-temporal domain using eye-tracking glasses. Attention mapping is a key technology for human perceptual activity analysis or Human-Robot Interaction (HRI) to support human visual cognition; however, measuring human attention in dynamic environments is challenging owing to the difficulty in localizing the subject and dealing with moving objects. To address this, we present a comprehensive framework, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">4D Attention</i> , for unified gaze mapping onto static and dynamic objects. Specifically, we estimate the glasses pose by leveraging a loose coupling of direct visual localization and Inertial Measurement Unit (IMU) values. Further, by installing reconstruction components into our framework, dynamic objects not captured in the 3D environment map are instantiated based on the input images. Finally, a scene rendering component synthesizes a first-person view with identification (ID) textures and performs direct 2D-3D gaze association. Quantitative evaluations showed the effectiveness of our framework. Additionally, we demonstrated the applications of 4D Attention through experiments in real situations. <xref ref-type="fn" rid="fn1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">¹</xref>

Gaze Focalization System for Driving Applications Using OpenFace 2.0 Toolkit with NARMAX Algorithm in Accidental Scenarios.

Monitoring driver attention using the gaze estimation is a typical approach used on road scenes. This indicator is of great importance for safe driving, specially on Level 3 and Level 4 automation systems, where the take over request control strategy could be based on the driver’s gaze estimation. Nowadays, gaze estimation techniques used in the state-of-the-art are intrusive and costly, and these two aspects are limiting the usage of these techniques on real vehicles. To test this kind of application, there are some databases focused on critical situations in simulation, but they do not show real accidents because of the complexity and the danger to record them. Within this context, this paper presents a low-cost and non-intrusive camera-based gaze mapping system integrating the open-source state-of-the-art OpenFace 2.0 Toolkit to visualize the driver focalization on a database composed of recorded real traffic scenes through a heat map using NARMAX (Nonlinear AutoRegressive Moving Average model with eXogenous inputs) to establish the correspondence between the OpenFace 2.0 parameters and the screen region the user is looking at. This proposal is an improvement of our previous work, which was based on a linear approximation using a projection matrix. The proposal has been validated using the recent and challenging public database DADA2000, which has 2000 video sequences with annotated driving scenarios based on real accidents. We compare our proposal with our previous one and with an expensive desktop-mounted eye-tracker, obtaining on par results. We proved that this method can be used to record driver attention databases.

Eye Tracking Interaction on Unmodified Mobile VR Headsets Using the Selfie Camera

Input methods for interaction in smartphone-based virtual and mixed reality (VR/MR) are currently based on uncomfortable head tracking controlling a pointer on the screen. User fixations are a fast and natural input method for VR/MR interaction. Previously, eye tracking in mobile VR suffered from low accuracy, long processing time, and the need for hardware add-ons such as anti-reflective lens coating and infrared emitters. We present an innovative mobile VR eye tracking methodology utilizing only the eye images from the front-facing (selfie) camera through the headset’s lens, without any modifications. Our system first enhances the low-contrast, poorly lit eye images by applying a pipeline of customised low-level image enhancements suppressing obtrusive lens reflections. We then propose an iris region-of-interest detection algorithm that is run only once. This increases the iris tracking speed by reducing the iris search space in mobile devices. We iteratively fit a customised geometric model to the iris to refine its coordinates. We display a thin bezel of light at the top edge of the screen for constant illumination. A confidence metric calculates the probability of successful iris detection. Calibration and linear gaze mapping between the estimated iris centroid and physical pixels on the screen results in low latency, real-time iris tracking. A formal study confirmed that our system’s accuracy is similar to eye trackers in commercial VR headsets in the central part of the headset’s field-of-view. In a VR game, gaze-driven user completion time was as fast as with head-tracked interaction, without the need for consecutive head motions. In a VR panorama viewer, users could successfully switch between panoramas using gaze.

Artificial Intelligence-Powered Smartphone App to Facilitate Medication Adherence: Protocol for a Human Factors Design Study.

BackgroundMedication Guides consisting of crucial interactions and side effects are extensive and complex. Due to the exhaustive information, patients do not retain the necessary medication information, which can result in hospitalizations and medication nonadherence. A gap exists in understanding patients’ cognition of managing complex medication information. However, advancements in technology and artificial intelligence (AI) allow us to understand patient cognitive processes to design an app to better provide important medication information to patients.ObjectiveOur objective is to improve the design of an innovative AI- and human factor–based interface that supports patients’ medication information comprehension that could potentially improve medication adherence.MethodsThis study has three aims. Aim 1 has three phases: (1) an observational study to understand patient perception of fear and biases regarding medication information, (2) an eye-tracking study to understand the attention locus for medication information, and (3) a psychological refractory period (PRP) paradigm study to understand functionalities. Observational data will be collected, such as audio and video recordings, gaze mapping, and time from PRP. A total of 50 patients, aged 18-65 years, who started at least one new medication, for which we developed visualization information, and who have a cognitive status of 34 during cognitive screening using the TICS-M test and health literacy level will be included in this aim of the study. In Aim 2, we will iteratively design and evaluate an AI-powered medication information visualization interface as a smartphone app with the knowledge gained from each component of Aim 1. The interface will be assessed through two usability surveys. A total of 300 patients, aged 18-65 years, with diabetes, cardiovascular diseases, or mental health disorders, will be recruited for the surveys. Data from the surveys will be analyzed through exploratory factor analysis. In Aim 3, in order to test the prototype, there will be a two-arm study design. This aim will include 900 patients, aged 18-65 years, with internet access, without any cognitive impairment, and with at least two medications. Patients will be sequentially randomized. Three surveys will be used to assess the primary outcome of medication information comprehension and the secondary outcome of medication adherence at 12 weeks.ResultsPreliminary data collection will be conducted in 2021, and results are expected to be published in 2022.ConclusionsThis study will lead the future of AI-based, innovative, digital interface design and aid in improving medication comprehension, which may improve medication adherence. The results from this study will also open up future research opportunities in understanding how patients manage complex medication information and will inform the format and design for innovative, AI-powered digital interfaces for Medication Guides.International Registered Report Identifier (IRRID)PRR1-10.2196/21659

Automated areas of interest analysis for usability studies of tangible screen-based user interfaces using mobile eye tracking

AbstractThe user's gaze can provide important information for human–machine interaction, but the analysis of manual gaze data is extremely time-consuming, inhibiting wide adoption in usability studies. Existing methods for automated areas of interest (AOI) analysis cannot be applied to tangible products with a screen-based user interface (UI), which have become ubiquitous in everyday life. The objective of this paper is to present and evaluate a method to automatically map the user's gaze to dynamic AOIs on tangible screen-based UIs based on computer vision and deep learning. This paper presents an algorithm forautomated Dynamic AOI Mapping(aDAM), which allows the automated mapping of gaze data recorded with mobile eye tracking to the predefined AOIs on tangible screen-based UIs. The evaluation of the algorithm is performed using two medical devices, which represent two extreme examples of tangible screen-based UIs. The different elements of aDAM are examined for accuracy and robustness, as well as the time saved compared to manual mapping. The break-even point for an analyst's effort for aDAM compared to manual analysis is found to be 8.9 min gaze data time. The accuracy and robustness of both the automated gaze mapping and the screen matching indicate that aDAM can be applied to a wide range of products. aDAM allows, for the first time, automated AOI analysis of tangible screen-based UIs with AOIs that dynamically change over time. The algorithm requires some additional initial input for the setup and training, but analyzed gaze data duration and effort is only determined by computation time and does not require any additional manual work thereafter. The efficiency of the approach has the potential for a broader adoption of mobile eye tracking in usability testing for the development of new products and may contribute to a more data-driven usability engineering process in the future.

A Dual-Cameras-Based Driver Gaze Mapping System With an Application on Non-Driving Activities Monitoring

Characterisation of the driver’s non-driving activities (NDAs) is of great importance to the design of the take-over control strategy in Level 3 automation. Gaze estimation is a typical approach to monitor the driver’s behaviour since the eye gaze is normally engaged with the human activities. However, current eye gaze tracking techniques are either costly or intrusive which limits their applicability in vehicles. This paper proposes a low-cost and non-intrusive dual-cameras based gaze mapping system that visualises the driver’s gaze using a heat map. The challenges introduced by complex head movement during NDAs and camera distortion are addressed by proposing a nonlinear polynomial model to establish the relationship between the face features and eye gaze on the simulated driver’s view. The Root Mean Square Error of this system in the in-vehicle experiment for the X and Y direction is 7.80±5.99 pixel and 4.64±3.47 pixel respectively with the image resolution of $1440 \times 1080$ pixels. This system is successfully demonstrated to evaluate three NDAs with visual attention. This technique, acting as a generic tool to monitor driver’s visual attention, will have wide applications on NDA characterisation for intelligent design of take over strategy and driving environment awareness for current and future automated vehicles.

Semantic 3D gaze mapping toward comprehending attention of multiple people

Semantic 3D gaze mapping toward comprehending attention of multiple people

Automating Areas of Interest Analysis in Mobile Eye Tracking Experiments based on Machine Learning.

For an in-depth, AOI-based analysis of mobile eye tracking data, a preceding gaze assign-ment step is inevitable. Current solutions such as manual gaze mapping or marker-based approaches are tedious and not suitable for applications manipulating tangible objects. This makes mobile eye tracking studies with several hours of recording difficult to analyse quan-titatively. We introduce a new machine learning-based algorithm, the computational Gaze-Object Mapping (cGOM), that automatically maps gaze data onto respective AOIs. cGOM extends state-of-the-art object detection and segmentation by mask R-CNN with a gaze mapping feature. The new algorithm’s performance is validated against a manual fixation-by-fixation mapping, which is considered as ground truth, in terms of true positive rate (TPR), true negative rate (TNR) and efficiency. Using only 72 training images with 264 labelled object representations, cGOM is able to reach a TPR of approx. 80% and a TNR of 85% compared to the manual mapping. The break-even point is reached at 2 hours of eye tracking recording for the total procedure, respectively 1 hour considering human working time only. Together with a real-time capability of the mapping process after completed train-ing, even hours of eye tracking recording can be evaluated efficiently. (Code and video examples have been made available at: https://gitlab.ethz.ch/pdz/cgom.git)

Point of Gaze Estimation Using Corneal Surface Reflection and Omnidirectional Camera Image

We present a human point of gaze estimation system using corneal surface reflection and omni-directional image taken by a fish eye. Only capturing an eye image, our system entables to find where a user is looking in 360P surrounding scene image. We first generates multiple perspective scene images from an equi-rectangular image and perform registration between corneal reflection and perspective images. We then compute the point of gaze using a 3D eye model and project the point to an omni-directional image. We evaluated the robustness of registration and accuracy of PoG estimations using two indoor and five outdoor scenes, and found that gaze mapping error was 5.526[deg] on average. This result shows the potential to the marketing and outdoor training system.

ArUco/Gaze Tracking in Real Environments

The emergence of affordable mobile eye-trackers has allowed to study gaze behavior in real-world environments. However, the gaze mapping from recorded video to a static reference image is a complex and open problem. Finding a reference image within the video frames, i.e., image matching, can give satisfying results, but occluded or overlapped objects are almost impossible to locate using this technique. We suggest using ArUco fiducial markers (and their associated software library available in OpenCV) to map gaze to dynamic Areas Of Interest (AOIs) within a reference image. Although such markers have been used previously, technical details of marker detection and mapping have been sparse. The current approach consists of three steps: (1) define an AOI using markers, then (2) resolve any conflict among overlapping AOIs, and (3) map the gaze point to the reference image. A dynamic AOI can be defined using one or more corner markers. When camera rotations are limited and the object is relatively orthogonal to the camera, it is possible to define an AOI using only one corner marker. When the camera rotates, its pose estimation is required to project corner points to the camera image plane. An AOI can also be defined with four corner markers, which has the advantage of robustness with respect to camera rotations, and no a priori required knowledge of the physical dimensions of the object. The two approaches can be combined, e.g., when using four corner markers and one of the markers is most (due to occlusion or view angle), the basis vectors can be used to interpolate the position of the lost marker. When two or more AOIs overlap and all the markers are tracked, gaze should be marked on the AOI closer to the camera. The distance to an object can be defined knowing the length of the object, the number of pixels spanned on the image and the pre-computed camera focal parameter. Once the AOIs are defined and marker overlaps are resolved, the gaze point can be mapped to the coordinates of the reference image using homography.

Gaze behavior during curb approach

Mobile device use is becoming increasingly prevalent during everyday activities. The simultaneous use of such technology while performing various activities may increase risk of human error and accidents, such as falls. In addition to effects on cognitive and attentional resources, mobile device use may influence visual behavior, thus affecting the user’s ability to perceive obstacles in the environment and respond appropriately. The purpose of this study was to examine the influence of texting on gaze behavior during the approach and negotiation of a simulated curb. Wireless eye-tracking was used to record gaze during locomotion, with and without concurrent texting. Semantic Gaze Mapping (SMI, Inc.) was used to co-register participants’ gaze to different areas of interest (AOIs). Texting resulted in significant reductions in gaze dwell time on the curb and the number of times the curb was fixated. These findings suggest that mobile device use may have a negative effect on navigation during locomotion by reducing visual attention to task-relevant areas along the future path of travel.

Idiosyncratic Feature-Based Gaze Mapping

It is argued that polynomial expressions that are normally used for remote, video-based, low cost eye tracking systems, are not always ideal to accommodate individual differences in eye cleft, position of the eye in the socket, corneal bulge, astigmatism, etc. A procedure to identify a set of polynomial expressions that will provide the best possible accuracy for a specific individual is proposed. It is also proposed that regression coefficients are recalculated in real-time, based on a subset of calibration points in the region of the current gaze and that a real-time correction is applied, based on the offsets from calibration targets that are close to the estimated point of regard.It was found that if no correction is applied, the choice of polynomial is critically important to get an accuracy that is just acceptable. Previously identified polynomial sets were confirmed to provide good results in the absence of any correction procedure. By applying real-time correction, the accuracy of any given polynomial improves while the choice of polynomial becomes less critical. Identification of the best polynomial set per participant and correction technique in combination with the aforementioned correction techniques, lead to an average error of 0.32° (sd = 0.10°) over 134 participant recordings.The proposed improvements could lead to low-cost systems that are accurate and fast enough to do reading research or other studies where high accuracy is expected at framerates in excess of 200 Hz.

Combining dynamic head pose–gaze mapping with the robot conversational state for attention recognition in human–robot interactions

The ability to recognize the visual focus of attention (VFOA, i.e. what or whom a person is looking at) of people is important for robots or conversational agents interacting with multiple people, since it plays a key role in turn-taking, engagement or intention monitoring. As eye gaze estimation is often impossible to achieve, most systems currently rely on head pose as an approximation, creating ambiguities since the same head pose can be used to look at different VFOA targets. To address this challenge, we propose a dynamic Bayesian model for the VFOA recognition from head pose, where we make two main contributions. First, taking inspiration from behavioral models describing the relationships between the body, head and gaze orientations involved in gaze shifts, we propose novel gaze models that dynamically and more accurately predict the expected head orientation used for looking in a given gaze target direction. This is a neglected aspect of previous works but essential for recognition. Secondly, we propose to exploit the robot conversational state (when he speaks, objects to which he refers) as context to set appropriate priors on candidate VFOA targets and reduce the inherent VFOA ambiguities. Experiments on a public dataset where the humanoid robot NAO plays the role of an art guide and quiz master demonstrate the benefit of the two contributions.

Mapping Functions in Gaze Tracking

Eye gaze is one of the powerful modes for interaction with the computer. Gaze estimation systems compute the direction of eye gaze. The gaze region on the screen determines the focus and attention of an individual. The gaze tracking systems consists of two major functionalities, feature extraction and determination of gaze mapping functions. The features are the parameters that characterize the pupil and the iris. The gaze mapping functions determine the coordinates on the screen based on interpolation or approximation. In this paper, an in depth analysis of mapping functions is dealt. An attempt has been made to understand the literature of gaze tracking systems.