Appearance-based Gaze Estimation Research Articles

Appearance-Based Gaze Estimation as a Benchmark for Eye Image Data Generation Methods

Data augmentation is commonly utilized to increase the size and diversity of training sets for deep learning tasks. In this study, we propose a novel application of an existing image generation approach in the domain of realistic eye images that leverages data collected from 40 subjects. This hybrid method combines the benefits of precise control over the image content provided by 3D rendering, while introducing the previously lacking photorealism and diversity into synthetic images through neural style transfer. We prove its general efficacy as a data augmentation tool for appearance-based gaze estimation when generated data are mixed with a sparse train set of real images. It improved the results for 39 out of 40 subjects, with an 11.22% mean and a 19.75% maximum decrease in gaze estimation error, achieving similar metrics for train and held-out subjects. We release our data repository of eye images with gaze labels used in this work for public access.

Evaluating User Experience and Data Quality in Gamified Data Collection for Appearance-Based Gaze Estimation

Appearance-based gaze estimation, which uses only a regular camera to estimate human gaze, is important in various application fields. While the technique faces data bias issues, data collection protocol is often demanding, and collecting data from a wide range of participants is difficult. It is an important challenge to design opportunities that allow a diverse range of people to participate while ensuring the quality of the training data. To tackle this challenge, we introduce a novel gamified approach for collecting training data. In this game, two players communicate words via eye gaze through a transparent letter board. Images captured during gameplay serve as valuable training data for gaze estimation models. The game is designed as a physical installation that involves communication between players, and it is expected to attract the interest of diverse participants. We assess the game’s significance on data quality and user experience through a comparative user study.

Young-gaze: an appearance-based gaze estimation solution for adolescents

Young-gaze: an appearance-based gaze estimation solution for adolescents

PrivatEyes: Appearance-based Gaze Estimation Using Federated Secure Multi-Party Computation

Latest gaze estimation methods require large-scale training data but their collection and exchange pose significant privacy risks. We propose PrivatEyes - the first privacy-enhancing training approach for appearance-based gaze estimation based on federated learning (FL) and secure multi-party computation (MPC). PrivatEyes enables training gaze estimators on multiple local datasets across different users and server-based secure aggregation of the individual estimators' updates. PrivatEyes guarantees that individual gaze data remains private even if a majority of the aggregating servers is malicious. We also introduce a new data leakage attack DualView that shows that PrivatEyes limits the leakage of private training data more effectively than previous approaches. Evaluations on the MPIIGaze, MPIIFaceGaze, GazeCapture, and NVGaze datasets further show that the improved privacy does not lead to a lower gaze estimation accuracy or substantially higher computational costs - both of which are on par with its non-secure counterparts.

PnP-GA+: Plug-and-Play Domain Adaptation for Gaze Estimation Using Model Variants.

Appearance-based gaze estimation has garnered increasing attention in recent years. However, deep learning-based gaze estimation models still suffer from suboptimal performance when deployed in new domains, e.g., unseen environments or individuals. In our previous work, we took this challenge for the first time by introducing a plug-and-play method (PnP-GA) to adapt the gaze estimation model to new domains. The core concept of PnP-GA is to leverage the diversity brought by a group of model variants to enhance the adaptability to diverse environments. In this article, we propose the PnP-GA+ by extending our approach to explore the impact of assembling model variants using three additional perspectives: color space, data augmentation, and model structure. Moreover, we propose an intra-group attention module that dynamically optimizes pseudo-labeling during adaptation. Experimental results demonstrate that by directly plugging several existing gaze estimation networks into the PnP-GA+ framework, it outperforms state-of-the-art domain adaptation approaches on four standard gaze domain adaptation tasks on public datasets. Our method consistently enhances cross-domain performance, and its versatility is improved through various ways of assembling the model group.

Webcam-based gaze estimation for computer screen interaction.

This paper presents a novel webcam-based approach for gaze estimation on computer screens. Utilizing appearance based gaze estimation models, the system provides a method for mapping the gaze vector from the user's perspective onto the computer screen. Notably, it determines the user's 3D position in front of the screen, using only a 2D webcam without the need for additional markers or equipment. The study presents a comprehensive comparative analysis, assessing the performance of the proposed method against established eye tracking solutions. This includes a direct comparison with the purpose-built Tobii Eye Tracker 5, a high-end hardware solution, and the webcam-based GazeRecorder software. In experiments replicating head movements, especially those imitating yaw rotations, the study brings to light the inherent difficulties associated with tracking such motions using 2D webcams. This research introduces a solution by integrating Structure from Motion (SfM) into the Convolutional Neural Network (CNN) model. The study's accomplishments include showcasing the potential for accurate screen gaze tracking with a simple webcam, presenting a novel approach for physical distance computation, and proposing compensation for head movements, laying the groundwork for advancements in real-world gaze estimation scenarios.

Gaze Estimation by Attention-Induced Hierarchical Variational Auto-Encoder.

Appearance-based gaze estimation has been widely studied recently with promising performance. The majority of appearance-based gaze estimation methods are developed under the deterministic frameworks. However, the deterministic gaze estimation methods suffer from large performance drop upon challenging eye images in low-resolution, darkness, partial occlusions, etc. To alleviate this problem, in this article, we alternatively reformulate the appearance-based gaze estimation problem under a generative framework. Specifically, we propose a variational inference model, that is, variational gaze estimation network (VGE-Net), to generate multiple gaze maps as complimentary candidates simultaneously supervised by the ground-truth gaze map. To achieve robust estimation, we adaptively fuse the gaze directions predicted on these candidate gaze maps by a regression network through a simple attention mechanism. Experiments on three benchmarks, that is, MPIIGaze, EYEDIAP, and Columbia, demonstrate that our VGE-Net outperforms state-of-the-art gaze estimation methods, especially on challenging cases. Comprehensive ablation studies also validate the effectiveness of our contributions. The code will be publicly released.

Appearance debiased gaze estimation via stochastic subject-wise adversarial learning

Recently, appearance-based gaze estimation has been attracting attention in computer vision, and remarkable improvements have been achieved using various deep learning techniques. Despite such progress, most methods aim to infer gaze vectors from images directly, which causes overfitting to person-specific appearance factors. In this paper, we address these challenges and propose a novel framework: Stochastic subject-wise Adversarial gaZE learning (SAZE), which trains a network to generalize the appearance of subjects. We design a Face generalization Network (Fgen-Net) using a face-to-gaze encoder and face identity classifier and a proposed adversarial loss. The proposed loss generalizes face appearance factors so that the identity classifier inferences a uniform probability distribution. In addition, the Fgen-Net is trained by a learning mechanism that optimizes the network by reselecting a subset of subjects at every training step to avoid overfitting. Our experimental results verify the robustness of the method in that it yields state-of-the-art performance, achieving 3.89°and 4.42°on the MPIIFaceGaze and EyeDiap datasets, respectively. Furthermore, we demonstrate the positive generalization effect by conducting further experiments using face images involving different styles generated from the generative model.

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.

Eye-Tracking System with Low-End Hardware: Development and Evaluation

Eye-tracking systems have emerged as valuable tools in various research fields, including psychology, medicine, marketing, car safety, and advertising. However, the high costs of the necessary specialized hardware prevent the widespread adoption of these systems. Appearance-based gaze estimation techniques offer a cost-effective alternative that can rely solely on RGB cameras, albeit with reduced accuracy. Therefore, the aim of our work was to present a real-time eye-tracking system with low-end hardware that leverages appearance-based techniques while overcoming their drawbacks to make reliable gaze data accessible to more users. Our system employs fast and light machine learning algorithms from an external library called MediaPipe to identify 3D facial landmarks. Additionally, it uses a series of widely recognized computer vision techniques, like morphological transformations, to effectively track eye movements. The precision and accuracy of the developed system in recognizing saccades and fixations when the eye movements are mainly horizontal were tested through a quantitative comparison with the EyeLink 1000 Plus, a professional eye tracker. Based on the encouraging registered results, we think that it is possible to adopt the presented system as a tool to quickly retrieve reliable gaze information.

A Complementary Dual-Branch Network for Appearance-Based Gaze Estimation From Low-Resolution Facial Image

Estimating gaze from a low-resolution facial image is a challenging task. Most current networks for gaze estimation focus on using face images of adequate resolution. Their performance degrades when the image resolution decreases due to information loss. This work aims to explore more helpful face and gaze information in a novel way to alleviate the problem of information loss in the low-resolution gaze estimation task. Considering that all faces have a relatively fixed structure, it is feasible to reconstruct the residual information of face and gaze based on the solid constraint of the prior knowledge of face structure through learning an end-to-end mapping from pairs of low-and high-resolution images. This paper proposes a complementary dual-branch network (CDBN) to achieve this task. A fundamental branch is designed to extract features of the major structural information from low-resolution input. A residual branch is employed to reconstruct features containing the residual information as a supplement under the supervision of both the high-resolution image and gaze direction. These two features are then fused and processed for gaze estimation. Experimental results on three widely used datasets, MPIIFaceGaze, EYEDIAP, and RT-GENE, show that the proposed CDBN achieves more accurate gaze estimation from the low-resolution input image compared with the state-of-the-art methods.

Appearance-based gaze estimation with feature fusion of multi-level information elements

Abstract Gaze estimation is a fundamental task in many applications of cognitive sciences, human–computer interaction, and robotics. The purely data-driven appearance-based gaze estimation methods may suffer from a lack of interpretability, which prevents their applicability to pervasive scenarios. In this study, a feature fusion method with multi-level information elements is proposed to improve the comprehensive performance of the appearance-based gaze estimation model. The multi-level feature extraction and expression are carried out from the originally captured images, and a multi-level information element matrix is established. A gaze conduction principle is formulated for reasonably fusing information elements from the established matrix. According to the gaze conduction principle along with the matrix, a multi-level information element fusion (MIEF) model for gaze estimation is proposed. Then, several input modes and network structures of the MIEF model are designed, and a series of grouping experiments are carried out on a small-scale sub-dataset. Furthermore, the optimized input modes and network structures of the MIEF model are selected for training and testing on the whole dataset to verify and compare model performance. Experimental results show that optimizing the feature combination in the input control module and fine-tuning the computational architecture in the feature extraction module can improve the performance of the gaze estimation model, which would enable the reduction of the model by incorporating the critical features and thus improve the performance and accessibility of the method. Compared with the reference baseline, the optimized model based on the proposed feature fusion method of multi-level information elements can achieve efficient training and improve the test accuracy in the verification experiment. The average error is 1.63 cm on phones on the GazeCapture dataset, which achieves comparable accuracy with state-of-the-art methods.

BoT2L-Net: Appearance-Based Gaze Estimation Using Bottleneck Transformer Block and Two Identical Losses in Unconstrained Environments

As a nonverbal cue, gaze plays a critical role in communication, expressing emotions and reflecting mental activity. It has widespread applications in various fields. Recently, the appearance-based gaze estimation method, which utilizes CNN (convolutional neural networks), has rapidly improved the accuracy and robustness of gaze estimation algorithms. Due to their insufficient ability to capture global relationships, the present accuracy of gaze estimation methods in unconstrained environments, has the potential for improvement. To address this challenge, the focus of this paper is to enhance the accuracy of gaze estimation, which is typically measured by mean angular error. In light of Transformer’s breakthrough in image classification and target detection tasks, and the need for an efficient network, the Transformer-enhanced-CNN method is a suitable choice. This paper proposed a novel model for 3D gaze estimation in unconstrained environments, based on the Bottleneck Transformer block and multi-loss methods. Our designed network (BoT2L-Net), incorporates self-attention through the BoT block, utilizing two identical loss functions to predict the two gaze angles. Additionally, the back-propagation network was combined with classification and regression losses, to improve the network’s accuracy and robustness. Our model was evaluated on two commonly used gaze datasets: Gaze360 and MPIIGaze, achieving mean angular errors of 11.53° and 9.59° for front 180° and front-facing gaze angles, respectively, on the Gaze360 testing set, and a mean angular error of 3.97° on the MPIIGaze testing set, outperforming the CNN-based gaze estimation method. The BoT2L-Net model proposed in this paper performs well on two publicly available datasets, demonstrating the effectiveness of our approach.

Appearance-Based Gaze Estimation Method Using Static Transformer Temporal Differential Network

Gaze behavior is important and non-invasive human–computer interaction information that plays an important role in many fields—including skills transfer, psychology, and human–computer interaction. Recently, improving the performance of appearance-based gaze estimation, using deep learning techniques, has attracted increasing attention: however, several key problems in these deep-learning-based gaze estimation methods remain. Firstly, the feature fusion stage is not fully considered: existing methods simply concatenate the different obtained features into one feature, without considering their internal relationship. Secondly, dynamic features can be difficult to learn, because of the unstable extraction process of ambiguously defined dynamic features. In this study, we propose a novel method to consider feature fusion and dynamic feature extraction problems. We propose the static transformer module (STM), which uses a multi-head self-attention mechanism to fuse fine-grained eye features and coarse-grained facial features. Additionally, we propose an innovative recurrent neural network (RNN) cell—that is, the temporal differential module (TDM)—which can be used to extract dynamic features. We integrated the STM and the TDM into the static transformer with a temporal differential network (STTDN). We evaluated the STTDN performance, using two publicly available datasets (MPIIFaceGaze and Eyediap), and demonstrated the effectiveness of the STM and the TDM. Our results show that the proposed STTDN outperformed state-of-the-art methods, including that of Eyediap (by 2.9%).

Towards High Performance Low Complexity Calibration in Appearance Based Gaze Estimation.

Appearance-based gaze estimation from RGB images provides relatively unconstrained gaze tracking from commonly available hardware. The accuracy of subject-independent models is limited partly by small intra-subject and large inter-subject variations in appearance, and partly by a latent subject-dependent bias. To improve estimation accuracy, we have previously proposed a gaze decomposition method that decomposes the gaze angle into the sum of a subject-independent gaze estimate from the image and a subject-dependent bias. Estimating the bias from images outperforms previously proposed calibration algorithms, unless the amount of calibration data is prohibitively large. This paper extends that work with a more complete characterization of the interplay between the complexity of the calibration dataset and estimation accuracy. In particular, we analyze the effect of the number of gaze targets, the number of images used per gaze target and the number of head positions in calibration data using a new NISLGaze dataset, which is well suited for analyzing these effects as it includes more diversity in head positions and orientations for each subject than other datasets. A better understanding of these factors enables low complexity high performance calibration. Our results indicate that using only a single gaze target and single head position is sufficient to achieve high quality calibration. However, it is useful to include variability in head orientation as the subject is gazing at the target. Our proposed estimator based on these studies (GEDDNet) outperforms state-of-the-art methods by more than 6.3%. One of the surprising findings of our work is that the same estimator yields the best performance both with and without calibration. This is convenient, as the estimator works well "straight out of the box," but can be improved if needed by calibration. However, this seems to violate the conventional wisdom that train and test conditions must be matched. To better understand the reasons, we provide a new theoretical analysis that specifies the conditions under which this can be expected. The dataset is available at http://nislgaze.ust.hk. Source code is available at https://github.com/HKUST-NISL/GEDDnet.

Improving Domain Generalization in Appearance-Based Gaze Estimation With Consistency Regularization

Improving Domain Generalization in Appearance-Based Gaze Estimation With Consistency Regularization

Non-Intrusive Real Time Eye Tracking Using Facial Alignment for Assistive Technologies.

Most affordable eye tracking systems use either intrusive setup such as head-mounted cameras or use fixed cameras with infrared corneal reflections via illuminators. In the case of assistive technologies, using intrusive eye tracking systems can be a burden to wear for extended periods of time and infrared based solutions generally do not work in all environments, especially outside or inside if the sunlight reaches the space. Therefore, we propose an eye-tracking solution using state-of-the-art convolutional neural network face alignment algorithms that is both accurate and lightweight for assistive tasks such as selecting an object for use with assistive robotics arms. This solution uses a simple webcam for gaze and face position and pose estimation. We achieve a much faster computation time than the current state-of-the-art while maintaining comparable accuracy. This paves the way for accurate appearance-based gaze estimation even on mobile devices, giving an average error of around 4.5°on the MPIIGaze dataset [1] and state-of-the-art average errors of 3.9°and 3.3°on the UTMultiview [2] and GazeCapture [3], [4] datasets respectively, while achieving a decrease in computation time of up to 91%.

Gaze Estimation via Modulation-Based Adaptive Network With Auxiliary Self-Learning

Given a face image, most of previous works in gaze estimation infer the gaze via a well-trained model with supervised training. However, the distribution of test data may be very different compared to that of training data since samples might be corrupted in real-world scenarios ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g</i> ., taking a photo in strong light). This will lead to a gap between source domain ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e</i> ., training data) and target domain ( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">i.e</i> ., test data). In this paper, we first introduce self-supervised learning into our method for addressing challenging situations in gaze estimation. Moreover, existing appearance-based gaze estimation methods focus on directing towards the development of powerful regressors, which mainly utilize face and eye images simultaneously or face (eye) images only. However, the problem of inter cues between face and eye features has been largely overlooked. To this end, we propose a novel Modulation-based Adaptive Network (MANet) for gaze estimation, which uses high-level knowledge to filter the distractive information and bridges the intrinsic relationship between face and eye features. Further, we combine self-supervised learning and MANet to learn to adapt to challenging cases, such as abnormal lighting conditions and poor-quality images, by minimizing a self-supervised loss and a supervised loss jointly. The experimental results on several datasets demonstrate the effectiveness of our proposed approach with a real-time speed of 900 <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">fps</i> on a PC with an NVIDIA Titan RTX GPU.

Appearance-Based Gaze Estimation for ASD Diagnosis.

Biomarkers, such as magnetic resonance imaging (MRI) and electroencephalogram have been used to help diagnose autism spectrum disorder (ASD). However, the diagnosis needs the assist of specialized medical equipment in the hospital or laboratory. To diagnose ASD in a more effective and convenient way, in this article, we propose an appearance-based gaze estimation algorithm-AttentionGazeNet, to accurately estimate the subject's 3-D gaze from a raw video. The experimental results show its competitive performance on the MPIIGaze dataset and the improvement of 14.7% for static head pose and 46.7% for moving head pose on the EYEDIAP dataset compared with the state-of-the-art gaze estimation algorithms. After projecting the obtained gaze vector onto the screen coordinate, we apply accumulated histogram to taking into account both spatial and temporal information of estimated gaze-point and head-pose sequences. Finally, classification is conducted on our self-collected autistic children video dataset (ACVD), which contains 405 videos from 135 different ASD children, 135 typically developing (TD) children in a primary school, and 135 TD children in a kindergarten. The classification results on ACVD shows the effectiveness and efficiency of our proposed method, with the accuracy 94.8%, the sensitivity 91.1% and the specificity 96.7% for ASD.

A Nonintrusive and Real-Time Classification Method for Driver’s Gaze Region Using an RGB Camera

Driver distraction has become a leading cause of traffic crashes. Visual distraction has the most direct impact on driving safety among various driver distractions. If the driver’s line of sight deviates from the road in front, there will be a high probability of visual distraction. A nonintrusive and real-time classification method for driver’s gaze region is proposed. A Multi-Task Convolutional Neural Network (MTCNN) face detector is used to collect the driver’s face image, and the driver’s gaze direction can be detected with a full-face appearance-based gaze estimation method. The driver’s gaze region is classified by the model trained through the machine learning algorithms such as Support Vector Machines (SVM), Random Forest (RF), and K-Nearest Neighbors (KNN). The simulated experiment and the real vehicle experiment were conducted to test the method. The results show that it has good performance on gaze region classification and strong robustness to complex environments. The models in this paper are all lightweight networks, which can meet the accuracy and speed requirements for the tasks. The method can be a good help for further exploring the visual distraction state level and exert an influence on the research of driving behavior.