Online test-time adaptation addresses the train-test domain gap by adapting the model on unlabeled streaming test inputs before making the final prediction. However, online adaptation for 3D human pose estimation suffers from error accumulation when relying on self-supervision with imperfect predictions, leading to degraded performance over time. To mitigate this fundamental challenge, we propose a novel solution that highlights the use of motion discretization. Specifically, we employ unsupervised clustering in the latent motion representation space to derive a set of anchor motions, whose regularity aids in supervising the human pose estimator and enables efficient self-replay. Additionally, we introduce an effective and efficient soft-reset mechanism by reverting the pose estimator to its exponential moving average during continuous adaptation. We examine long-term online adaptation by continuously adapting to out-of-domain streaming test videos of the same individual, which allows for the capture of consistent personal shape and motion traits throughout the streaming observation. By mitigating error accumulation, our solution enables robust exploitation of these personal traits for enhanced accuracy. Experiments demonstrate that our solution outperforms previous online test-time adaptation methods and validate our design choices.

Despite decades of research on data collection and model architectures, current gaze estimation models encounter significant challenges in generalizing across diverse data domains. Recent advances in self-supervised pre-training have demonstrated remarkable generalization across various vision tasks. However, their effectiveness in gaze estimation remains unexplored. We propose UniGaze, which for the first time leverages large-scale in-the-wild facial datasets for gaze estimation through self-supervised pre-training. Through systematic investigation, we clarify critical factors essential for effective pre-training in gaze estimation. Our experiments reveal that self-supervised approaches designed for semantic tasks fail when applied to gaze estimation, while our carefully designed pre-training pipeline consistently improves cross-domain performance. Through comprehensive experiments of the challenging cross-dataset evaluations and novel protocols, including leave-one-dataset-out and joint-dataset settings, we demonstrate that UniGaze significantly improves generalization across multiple data domains while minimizing reliance on costly labeled data. Source code and model are available at https://github.com/ut-vision/UniGaze.

Generalization to unseen environments is crucial in appearance-based gaze estimation but is primarily hindered by limitations in appearance diversity and gaze range. However, collecting large-scale real data remains challenging due to the inefficiency of physically capturing 3D gaze directions. In this paper, we present a novel fully learning-by-generation pipeline that relies exclusively on synthetic data. While prior synthetic data has shown promise in extending gaze ranges, its limited ability to control appearance still leads to suboptimal performance in unknown environments. Our approach, based on a generative adversarial network (GAN), enables more flexible data synthesis, simultaneously addressing both gaze range and appearance diversity. Furthermore, to effectively leverage synthetic data for improved model generalization, we introduce a two-stage training pipeline that mitigates domain shifts and pseudo-label noise. Extensive experiments and analyses demonstrate that the proposed learning-by-generation pipeline enhances robustness across various scenarios, making it valuable for both data augmentation and pre-training in transfer learning.

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.

Abstract Sound localization is one of the essential tasks in audio-visual learning. Especially, stereo sound localization methods have been proposed to handle multiple sound sources. However, existing stereo-sound localization methods treat sound source localization as a segmentation task and, as a result, require costly annotation of segmentation masks. Another serious problem of the existing stereo-sound localization methods is that they have been trained and evaluated only in a controlled environment, such as a fixed camera and microphone setting with limited variability of scenes. Therefore, their performance on videos recorded in uncontrolled environments, such as in-the-wild videos from the Internet, has not been fully investigated. To address these problems, we propose a weakly supervised method as an extension of a typical stereo-sound localization method by utilizing the spatial relative order of sound sources in recorded videos. The proposed method solves the annotation problem by training the localization model using only sound category labels. Furthermore, our method utilizes the spatial relative order of the sound sources, which is not affected by specific recording settings, and thus can be effectively used for videos recorded in uncontrolled environments. We also collect stereo-recorded videos from YouTube to construct a new dataset to demonstrate the applicability of the proposed method to stereo sounds recorded in various environments. Our method enhances the localization performance by inserting a novel training step that exploits the relative order of sound sources into a typical audio-visual localization method in both existing and newly introduced audio-visual datasets.

We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of daily-life activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards, with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception.

Appearance-based gaze estimation has been actively studied in recent years. However, its generalization performance for unseen head poses is still a significant limitation for existing methods. This work proposes a generalizable multi-view gaze estimation task and a cross-view feature fusion method to address this issue. In addition to paired images, our method takes the relative rotation matrix between two cameras as additional input. The proposed network learns to extract rotatable feature representation by using relative rotation as a constraint and adaptively fuses the rotatable features via stacked fusion modules. This simple yet efficient approach significantly improves generalization performance under unseen head poses without significantly increasing computational cost. The model can be trained with random combinations of cameras without fixing the positioning and can generalize to unseen camera pairs during inference. Through experiments using multiple datasets, we demonstrate the advantage of the proposed method over baseline methods, including state-of-the-art domain generalization approaches. The code will be available at https://github.com/ut-vision/Rot-MVGaze.

Abstract While AI technology is becoming increasingly prevalent in our daily lives, the comprehension of machine learning (ML) among non-experts remains limited. Interactive machine learning (IML) has the potential to serve as a tool for end users, but many existing IML systems are designed for users with a certain level of expertise. Consequently, it remains unclear whether IML experiences can enhance the comprehension of ordinary users. In this study, we conducted a public event using an IML system to assess whether participants could gain technical comprehension through hands-on IML experiences. We implemented an interactive sound classification system featuring visualization of internal feature representation and invited visitors at a science museum to freely interact with it. By analyzing user behavior and questionnaire responses, we discuss the potential and limitations of IML systems as a tool for promoting technical comprehension among non-experts.

We introduce Ego4D, a massive-scale egocentric video dataset and benchmark suite. It offers 3,670 hours of dailylife activity video spanning hundreds of scenarios (household, outdoor, workplace, leisure, etc.) captured by 931 unique camera wearers from 74 worldwide locations and 9 different countries. The approach to collection is designed to uphold rigorous privacy and ethics standards, with consenting participants and robust de-identification procedures where relevant. Ego4D dramatically expands the volume of diverse egocentric video footage publicly available to the research community. Portions of the video are accompanied by audio, 3D meshes of the environment, eye gaze, stereo, and/or synchronized videos from multiple egocentric cameras at the same event. Furthermore, we present a host of new benchmark challenges centered around understanding the first-person visual experience in the past (querying an episodic memory), present (analyzing hand-object manipulation, audio-visual conversation, and social interactions), and future (forecasting activities). By publicly sharing this massive annotated dataset and benchmark suite, we aim to push the frontier of first-person perception. Project page: https://ego4d-data.org/

Unsupervised domain adaptive video action recognition aims to recognize actions of a target domain using a model trained with only out-of-domain (source) annotations. The inherent complexity of videos makes this task challenging but also provides ground for leveraging multi-modal inputs (e.g., RGB, Flow, Audio). Most previous works utilize the multi-modal information by either aligning each modality individually or learning representation via cross-modal self-supervision. Different from previous works, we find that the cross-domain alignment can be more effectively done by using cross-modal interaction first. Cross-modal knowledge interaction allows other modalities to supplement missing transferable information because of the cross-modal complementarity. Also, the most transferable aspects of data can be highlighted using cross-modal consensus. In this work, we present a novel model that jointly considers these two characteristics for domain adaptive action recognition. We achieve this by implementing two modules, where the first module exchanges complementary transferable information across modalities through the semantic space, and the second module finds the most transferable spatial region based on the consensus of all modalities. Extensive experiments validate that our proposed method can significantly outperform state-of-the-art methods on multiple benchmark datasets, including the complex fine-grained dataset EPIC-Kitchens-100.

This article presents a photometric stereo method based on deep learning. One of the major difficulties in photometric stereo is designing an appropriate reflectance model that is both capable of representing real-world reflectances and computationally tractable for deriving surface normal. Unlike previous photometric stereo methods that rely on a simplified parametric image formation model, such as the Lambert's model, the proposed method aims at establishing a flexible mapping between complex reflectance observations and surface normal using a deep neural network. In addition, the proposed method predicts the reflectance, which allows us to understand surface materials and to render the scene under arbitrary lighting conditions. As a result, we propose a deep photometric stereo network (DPSN) that takes reflectance observations under varying light directions and infers the surface normal and reflectance in a per-pixel manner. To make the DPSN applicable to real-world scenes, a dataset of measured BRDFs (MERL BRDF dataset) has been used for training the network. Evaluation using simulation and real-world scenes shows the effectiveness of the proposed approach in estimating both surface normal and reflectances.

Temporal relations among multiple action segments play an important role in action segmentation especially when observations are limited (e.g., actions are occluded by other objects or happen outside a field of view). In this paper, we propose a network module called Graph-based Temporal Reasoning Module (GTRM) that can be built on top of existing action segmentation models to learn the relation of multiple action segments in various time spans. We model the relations by using two Graph Convolution Networks (GCNs) where each node represents an action segment. The two graphs have different edge properties to account for boundary regression and classification tasks, respectively. By applying graph convolution, we can update each node's representation based on its relation with neighboring nodes. The updated representation is then used for improved action segmentation. We evaluate our model on the challenging egocentric datasets namely EGTEA and EPIC-Kitchens, where actions may be partially observed due to the viewpoint restriction. The results show that our proposed GTRM outperforms state-of-the-art action segmentation models by a large margin. We also demonstrate the effectiveness of our model on two third-person video datasets, the 50Salads dataset and the Breakfast dataset.

Interactive machine learning techniques have a great potential to personalize media recognition models for each individual user by letting them browse and annotate a large amount of training data. However, graphical user interfaces (GUIs) for interactive machine learning have been mainly investigated in image and text recognition scenarios, not in other data modalities such as sound. In a scenario where users browse a large amount of audio files to search and annotate target samples corresponding to their own sound recognition classes, it is difficult for them to easily navigate through the overall sample structure due to the non-visual nature of audio data. In this work, we investigate the design issue for interactive sound recognition by comparing different visualization techniques ranging from audio spectrograms to deep learning-based audio-to-image retrieval. Based on an analysis of the user study, we clarify the advantages and disadvantages of audio visualization techniques, and provide design implications for interactive sound recognition GUIs using a massive amount of audio samples.

Abstract We present a method for geometric point light source calibration. Unlike prior works that use Lambertian spheres, mirror spheres, or mirror planes, we use a calibration target consisting of a plane and small shadow casters at unknown positions above the plane. We show that shadow observations from a moving calibration target under a fixed light follow the principles of pinhole camera geometry and epipolar geometry, allowing joint recovery of the light position and 3D shadow caster positions, equivalent to how conventional structure from motion jointly recovers camera parameters and 3D feature positions from observed 2D features. Moreover, we devised a unified light model that works with nearby point lights as well as distant light in one common framework. Our evaluation shows that our method yields light estimates that are stable and more accurate than existing techniques while having a much simpler setup and requiring less manual labor.

Appearance-based gaze estimation methods that only require an off-the-shelf camera have significantly improved but they are still not yet widely used in the human-computer interaction (HCI) community. This is partly because it remains unclear how they perform compared to model-based approaches as well as dominant, special-purpose eye tracking equipment. To address this limitation, we evaluate the performance of state-of-the-art appearance-based gaze estimation for interaction scenarios with and without personal calibration, indoors and outdoors, for different sensing distances, as well as for users with and without glasses. We discuss the obtained findings and their implications for the most important gaze-based applications, namely explicit eye input, attentive user interfaces, gaze-based user modelling, and passive eye monitoring. To democratise the use of appearance-based gaze estimation and interaction in HCI, we finally present OpenGaze (www.opengaze.org), the first software toolkit for appearance-based gaze estimation and interaction.

Learning-based methods are believed to work well for unconstrained gaze estimation, i.e. gaze estimation from a monocular RGB camera without assumptions regarding user, environment, or camera. However, current gaze datasets were collected under laboratory conditions and methods were not evaluated across multiple datasets. Our work makes three contributions towards addressing these limitations. First, we present the MPIIGaze dataset, which contains 213,659 full face images and corresponding ground-truth gaze positions collected from 15 users during everyday laptop use over several months. An experience sampling approach ensured continuous gaze and head poses and realistic variation in eye appearance and illumination. To facilitate cross-dataset evaluations, 37,667 images were manually annotated with eye corners, mouth corners, and pupil centres. Second, we present an extensive evaluation of state-of-the-art gaze estimation methods on three current datasets, including MPIIGaze. We study key challenges including target gaze range, illumination conditions, and facial appearance variation. We show that image resolution and the use of both eyes affect gaze estimation performance, while head pose and pupil centre information are less informative. Finally, we propose GazeNet, the first deep appearance-based gaze estimation method. GazeNet improves on the state of the art by 22 percent (from a mean error of 13.9 degrees to 10.8 degrees) for the most challenging cross-dataset evaluation.

Learning-based gaze estimation has significant potential to enable attentive user interfaces and gaze-based interaction on the billions of camera-equipped handheld devices and ambient displays. While training accurate person- and device-independent gaze estimators remains challenging, person-specific training is feasible but requires tedious data collection for each target device. To address these limitations, we present the first method to train person-specific gaze estimators across multiple devices. At the core of our method is a single convolutional neural network with shared feature extraction layers and device-specific branches that we train from face images and corresponding on-screen gaze locations. Detailed evaluations on a new dataset of interactions with five common devices (mobile phone, tablet, laptop, desktop computer, smart TV) and three common applications (mobile game, text editing, media center) demonstrate the significant potential of cross-device training. We further explore training with gaze locations derived from natural interactions, such as mouse or touch input.

Appearance-based gaze estimation is promising for unconstrained real-world settings, but the significant variability in head pose and user-camera distance poses significant challenges for training generic gaze estimators. Data normalization was proposed to cancel out this geometric variability by mapping input images and gaze labels to a normalized space. Although used successfully in prior works, the role and importance of data normalization remains unclear. To fill this gap, we study data normalization for the first time using principled evaluations on both simulated and real data. We propose a modification to the current data normalization formulation by removing the scaling factor and show that our new formulation performs significantly better (between 9.5% and 32.7%) in the different evaluation settings. Using images synthesized from a 3D face model, we demonstrate the benefit of data normalization for the efficiency of the model training. Experiments on real-world images confirm the advantages of data normalization in terms of gaze estimation performance.

This paper presents a novel interface to support video coding of social attention in the assessment of children with autism spectrum disorder. Video-based evaluations of social attention during therapeutic activities allow observers to find target behaviors while handling the ambiguity of attention. Despite the recent advances in computer vision-based gaze estimation methods, fully automatic recognition of social attention under diverse environments is still challenging. The goal of this work is to investigate an approach that uses automatic video analysis in a supportive manner for guiding human judgment. The proposed interface displays visualization of gaze estimation results on videos and provides GUI support to allow users to facilitate agreement between observers by defining social attention labels on the video timeline. Through user studies and expert reviews, we show how the interface helps observers perform video coding of social attention and how human judgment compensates for technical limitations of the automatic gaze analysis.

Visual attention is highly fragmented during mobile interactions, but the erratic nature of attention shifts currently limits attentive user interfaces to adapting after the fact, i.e. after shifts have already happened. We instead study attention forecasting - the challenging task of predicting users' gaze behaviour (overt visual attention) in the near future. We present a novel long-term dataset of everyday mobile phone interactions, continuously recorded from 20 participants engaged in common activities on a university campus over 4.5 hours each (more than 90 hours in total). We propose a proof-of-concept method that uses device-integrated sensors and body-worn cameras to encode rich information on device usage and users' visual scene. We demonstrate that our method can forecast bidirectional attention shifts and predict whether the primary attentional focus is on the handheld mobile device. We study the impact of different feature sets on performance and discuss the significant potential but also remaining challenges of forecasting user attention during mobile interactions.

Eye contact is an important non-verbal cue in social signal processing and promising as a measure of overt attention in human-object interactions and attentive user interfaces. However, robust detection of eye contact across different users, gaze targets, camera positions, and illumination conditions is notoriously challenging. We present a novel method for eye contact detection that combines a state-of-the-art appearance-based gaze estimator with a novel approach for unsupervised gaze target discovery, i.e. without the need for tedious and time-consuming manual data annotation. We evaluate our method in two real-world scenarios: detecting eye contact at the workplace, including on the main work display, from cameras mounted to target objects, as well as during everyday social interactions with the wearer of a head-mounted egocentric camera. We empirically evaluate the performance of our method in both scenarios and demonstrate its effectiveness for detecting eye contact independent of target object type and size, camera position, and user and recording environment.

We present a computational model to predict users' spatio-temporal visual attention on WIMP-style (windows, icons, menus, pointer) graphical user interfaces. Like existing models of bottom-up visual attention in computer vision, our model does not require any eye tracking equipment. Instead, it predicts attention solely using information available to the interface, specifically users' mouse and keyboard input as well as the UI components they interact with. To study our model in a principled way, we further introduce a method to synthesize user interface layouts that are functionally equivalent to real-world interfaces, such as from Gmail, Facebook, or GitHub. We first quantitatively analyze attention allocation and its correlation with user input and UI components using ground-truth gaze, mouse, and keyboard data of 18 participants performing a text editing task. We then show that our model predicts attention maps more accurately than state-of-the-art methods. Our results underline the significant potential of spatio-temporal attention modeling for user interface evaluation, optimization, or even simulation.

3D gaze information is important for scene-centric attention analysis, but accurate estimation and analysis of 3D gaze in real-world environments remains challenging. We present a novel 3D gaze estimation method for monocular head-mounted eye trackers. In contrast to previous work, our method does not aim to infer 3D eyeball poses, but directly maps 2D pupil positions to 3D gaze directions in scene camera coordinate space. We first provide a detailed discussion of the 3D gaze estimation task and summarize different methods, including our own. We then evaluate the performance of different 3D gaze estimation approaches using both simulated and real data. Through experimental validation, we demonstrate the effectiveness of our method in reducing parallax error, and we identify research challenges for the design of 3D calibration procedures.

We present labelled pupils in the wild (LPW), a novel dataset of 66 high-quality, high-speed eye region videos for the development and evaluation of pupil detection algorithms. The videos in our dataset were recorded from 22 participants in everyday locations at about 95 FPS using a state-of-the-art dark-pupil head-mounted eye tracker. They cover people of different ethnicities and a diverse set of everyday indoor and outdoor illumination environments, as well as natural gaze direction distributions. The dataset also includes participants wearing glasses, contact lenses, and make-up. We benchmark five state-of-the-art pupil detection algorithms on our dataset with respect to robustness and accuracy. We further study the influence of image resolution and vision aids as well as recording location (indoor, outdoor) on pupil detection performance. Our evaluations provide valuable insights into the general pupil detection problem and allow us to identify key challenges for robust pupil detection on head-mounted eye trackers.

Gaze is frequently explored in public display research given its importance for monitoring and analysing audience attention. However, current gaze-enabled public display interfaces require either special-purpose eye tracking equipment or explicit personal calibration for each individual user. We present AggreGaze, a novel method for estimating spatio-temporal audience attention on public displays. Our method requires only a single off-the-shelf camera attached to the display, does not require any personal calibration, and provides visual attention estimates across the full display. We achieve this by 1) compensating for errors of state-of-the-art appearance-based gaze estimation methods through on-site training data collection, and by 2) aggregating uncalibrated and thus inaccurate gaze estimates of multiple users into joint attention estimates. We propose different visual stimuli for this compensation: a standard 9-point calibration, moving targets, text and visual stimuli embedded into the display content, as well as normal video content. Based on a two-week deployment in a public space, we demonstrate the effectiveness of our method for estimating attention maps that closely resemble ground-truth audience gaze distributions.

Appearance-based gaze estimation is believed to work well in real-world settings, but existing datasets have been collected under controlled laboratory conditions and methods have been not evaluated across multiple datasets. In this work we study appearance-based gaze estimation in the wild. We present the MPIIGaze dataset that contains 213,659 images we collected from 15 participants during natural everyday laptop use over more than three months. Our dataset is significantly more variable than existing ones with respect to appearance and illumination. We also present a method for in-the-wild appearance-based gaze estimation using multimodal convolutional neural networks that significantly outperforms state-of-the art methods in the most challenging cross-dataset evaluation. We present an extensive evaluation of several state-of-the-art image-based gaze estimation algorithms on three current datasets, including our own. This evaluation provides clear insights and allows us to identify key research challenges of gaze estimation in the wild.

Images of the eye are key in several computer vision problems, such as shape registration and gaze estimation. Recent large-scale supervised methods for these problems require time-consuming data collection and manual annotation, which can be unreliable. We propose synthesizing perfectly labelled photo-realistic training data in a fraction of the time. We used computer graphics techniques to build a collection of dynamic eye-region models from head scan geometry. These were randomly posed to synthesize close-up eye images for a wide range of head poses, gaze directions, and illumination conditions. We used our model's controllability to verify the importance of realistic illumination and shape variations in eye-region training data. Finally, we demonstrate the benefits of our synthesized training data (SynthesEyes) by out-performing state-of-the-art methods for eye-shape registration as well as cross-dataset appearance-based gaze estimation in the wild.

Head-mounted eye tracking has significant potential for gaze-based applications such as life logging, mental health monitoring, or the quantified self. A neglected challenge for the long-term recordings required by these applications is that drift in the initial person-specific eye tracker calibration, for example caused by physical activity, can severely impact gaze estimation accuracy and thus system performance and user experience. We first analyse calibration drift on a new dataset of natural gaze data recorded using synchronised video-based and Electrooculography-based eye trackers of 20 users performing everyday activities in a mobile setting. Based on this analysis we present a method to automatically self-calibrate head-mounted eye trackers based on a computational model of bottom-up visual saliency. Through evaluations on the dataset we show that our method 1) is effective in reducing calibration drift in calibrated eye trackers and 2) given sufficient data, can achieve gaze estimation accuracy competitive with that of a calibrated eye tracker, without any manual calibration.

This paper presents an unconstrained gaze estimation method using an online learning algorithm. We focus on a desktop scenario, where a user operates a personal computer, and use the mouse-clicked positions to infer, where on the screen the user is looking at. Our method continuously captures the user's head pose and eye images with a monocular camera, and each mouse click triggers learning sample acquisition. In order to handle head pose variations, the samples are adaptively clustered according to the estimated head pose. Then, local reconstruction-based gaze estimation models are incrementally updated in each cluster. We conducted a prototype evaluation in real-world environments, and our method achieved an estimation accuracy of 2.9°.

In this paper, we address the problem of free head motion in appearance-based gaze estimation. This problem remains challenging because head motion changes eye appearance significantly, and thus, training images captured for an original head pose cannot handle test images captured for other head poses. To overcome this difficulty, we propose a novel gaze estimation method that handles free head motion via eye image synthesis based on a single camera. Compared with conventional fixed head pose methods with original training images, our method only captures four additional eye images under four reference head poses, and then, precisely synthesizes new training images for other unseen head poses in estimation. To this end, we propose a single-directional (SD) flow model to efficiently handle eye image variations due to head motion. We show how to estimate SD flows for reference head poses first, and then use them to produce new SD flows for training image synthesis. Finally, with synthetic training images, joint optimization is applied that simultaneously solves an eye image alignment and a gaze estimation. Evaluation of the method was conducted through experiments to assess its performance and demonstrate its effectiveness.

Inferring human gaze from low-resolution eye images is still a challenging task despite its practical importance in many application scenarios. This paper presents a learning-by-synthesis approach to accurate image-based gaze estimation that is person- and head pose-independent. Unlike existing appearance-based methods that assume person-specific training data, we use a large amount of cross-subject training data to train a 3D gaze estimator. We collect the largest and fully calibrated multi-view gaze dataset and perform a 3D reconstruction in order to generate dense training data of eye images. By using the synthesized dataset to learn a random regression forest, we show that our method outperforms existing methods that use low-resolution eye images.

Learning-based approaches using actual human gaze data have been proven to be an efficient way to acquire accurate visual saliency models and attracted much interest in recent years. However, it still remains yet to be answered how different types of stimulus (e.g., fractal images, and natural images with or without human faces) and viewing tasks (e.g., free viewing or a preference rating task) affect learned visual saliency models. In this study, we quantitatively investigate how learned saliency models differ when using datasets collected in different settings (image contextual level and viewing task) and discuss the importance of choosing appropriate experimental settings.

We investigate the appearance-based gaze estimation problem, with respect to its essential difficulty in reducing the number of required training samples, and other practical issues such as slight head motion, image resolution variation, and eye blinking. We cast the problem as mapping high-dimensional eye image features to low-dimensional gaze positions, and propose an adaptive linear regression (ALR) method as the key to our solution. The ALR method adaptively selects an optimal set of sparsest training samples for the gaze estimation via ℓ(1)-optimization. In this sense, the number of required training samples is significantly reduced for high accuracy estimation. In addition, by adopting the basic ALR objective function, we integrate the gaze estimation, subpixel alignment and blink detection into a unified optimization framework. By solving these problems simultaneously, we successfully handle slight head motion, image resolution variation and eye blinking in appearance-based gaze estimation. We evaluated the proposed method by conducting experiments with multiple users and variant conditions to verify its effectiveness.

This paper presents an approach to discover social groups in surveillance videos by incorporating attention-based cues to model group behaviors of pedestrians in videos. Group behaviors are modeled as a set of decision trees with the decisions being basic measurements based on positionbased and attention-based cues. Rather than enforcing explicit models, we apply tree-based learning algorithms to implicitly construct the decision tree models. The experimental results demonstrate that incorporating attention-based cues significantly increased the estimation accuracy compared to the conventional approaches that used position-based cues alone.

We propose a gaze sensing method using visual saliency maps that does not need explicit personal calibration. Our goal is to create a gaze estimator using only the eye images captured from a person watching a video clip. Our method treats the saliency maps of the video frames as the probability distributions of the gaze points. We aggregate the saliency maps based on the similarity in eye images to efficiently identify the gaze points from the saliency maps. We establish a mapping between the eye images to the gaze points by using Gaussian process regression. In addition, we use a feedback loop from the gaze estimator to refine the gaze probability maps to improve the accuracy of the gaze estimation. The experimental results show that the proposed method works well with different people and video clips and achieves a 3.5-degree accuracy, which is sufficient for estimating a user's attention on a display.

We present a method that extracts groups of fixations and image regions for the purpose of gaze analysis and image understanding. Since the attentional relationship between visual entities conveys rich information, automatically determining the relationship provides us a semantic representation of images. We show that, by jointly clustering human gaze and visual entities, it is possible to build meaningful and comprehensive metadata that offer an interpretation about how people see images. To achieve this, we developed a clustering method that uses a joint graph structure between fixation points and over-segmented image regions to ensure a cross-domain smoothness constraint. We show that the proposed clustering method achieves better performance in relating attention to visual entities in comparison with standard clustering techniques.

Characteristics of the human visual field are well known to be different in central (fovea) and peripheral areas. Existing computational models of visual saliency, however, do not take into account this biological evidence. The existing models compute visual saliency uniformly over the retina and, thus, have difficulty in accurately predicting the next gaze (fixation) point. This paper proposes to incorporate human visual field characteristics into visual saliency, and presents a computational model for producing such a saliency map. Our model integrates image features obtained by bottom-up computation in such a way that weights for the integration depend on the distance from the current gaze point where the weights are optimally learned using actual saccade data. The experimental results using a large number of fixation/saccade data with wide viewing angles demonstrate the advantage of our saliency map, showing that it can accurately predict the point where one looks next.

To infer human gaze from eye appearance, various methods have been proposed. However, most of them assume a fixed head pose because allowing free head motion adds 6 degrees of freedom to the problem and requires a prohibitively large number of training samples. In this paper, we aim at solving the appearance-based gaze estimation problem under free head motion without significantly increasing the cost of training. The idea is to decompose the problem into subproblems, including initial estimation under fixed head pose and subsequent compensations for estimation biases caused by head rotation and eye appearance distortion. Then each subproblem is solved by either learning-based method or geometric-based calculation. Specifically, the gaze estimation bias caused by eye appearance distortion is learnt effectively from a 5-seconds video clip. Extensive experiments were conducted to verify the effectiveness of the proposed approach. 1

The problem of estimating human gaze from eye appearance is regarded as mapping high-dimensional features to low-dimensional target space. Conventional methods require densely obtained training samples on the eye appearance manifold, which results in a tedious calibration stage. In this paper, we introduce an adaptive linear regression (ALR) method for accurate mapping via sparsely collected training samples. The key idea is to adaptively find the subset of training samples where the test sample is most linearly representable. We solve the problem via l1-optimization and thoroughly study the key issues to seek for the best solution for regression. The proposed gaze estimation approach based on ALR is naturally sparse and low-dimensional, giving the ability to infer human gaze from variant resolution eye images using much fewer training samples than existing methods. Especially, the optimization procedure in ALR is extended to solve the subpixel alignment problem simultaneously for low resolution test eye images. Performance of the proposed method is evaluated by extensive experiments against various factors such as number of training samples, feature dimensionality and eye image resolution to verify its effectiveness.

We propose a calibration-free gaze sensing method using visual saliency maps. Our goal is to construct a gaze estimator only using eye images captured from a person watching a video clip. The key is treating saliency maps of the video frames as probability distributions of gaze points. To efficiently identify gaze points from saliency maps, we aggregate saliency maps based on the similarity of eye appearances. We establish mapping between eye images to gaze points by Gaussian process regression. The experimental result shows that the proposed method works well with different people and video clips and achieves 6 degrees of accuracy, which is useful for estimating a person's attention on monitors.