Abstract Recent years have seen growing interest in contactless authentication, with several soft biometrics solutions being explored and producing promising results. This paper introduces a novel soft biometric approach for Verification that exploits users’ eye features and gaze behavior measured during the free observation of animated stimuli. The main novelty of our research lies in the specific stimuli employed: small squares moving according to various symmetrical patterns, which, to our knowledge, have never been considered before in similar investigations. In a user study involving 36 participants, we acquired gaze data through a remote eye tracker without performing any preliminary calibration. Both the closed- and open-set cases were considered. Experiments performed with various machine learning classifiers showed good verification performance, with a best accuracy of 88.84% and a best Equal Error Rate (EER) of 5.0% for the closed-set case, and a best accuracy of 87.59% and a best EER of 6.09% for the open-set case.

PatchCore is a widely used algorithm for industrial anomaly detection, due to its high performance compared to alternative approaches. This paper focuses on optimizing its architecture for the challenging case of one-shot anomaly detection. More precisely, we modified the feature extraction step by substituting the original ResNet50 with the Anti-aliased ResNet50, and the dimensionality reduction phase by replacing the original Average Pooling with the Gaussian Random Projection. Comparative experiments conducted on the MVTec Anomaly Detection (MVTec-AD) dataset showed that the proposed modifications can significantly increase performance with respect to the original PatchCore on several common categories of objects.

Finding defects during production is crucial in industrial manufacturing. While various detection methods exist, this paper proposes a short review on techniques based on visual saliency. Saliency approaches try to mimic how humans see and are characterized by advantages such as accuracy, speed, and ability to handle complex situations. With solutions based on deep learning becoming increasingly dominant, this review aims to provide a brief analysis of the current state of research in this field, possibly stimulating continued exploration and development of new techniques.

Gaze-based writing is one of the most widespread eye tracking applications for human–computer interaction. While eye tracking communication has traditionally been employed as an assistive technology, declining prices of eye trackers now make it a feasible alternative to keyboards or touchscreens in many contexts (for example, the interaction with public info points). In this paper we propose Leyenes, a text entry method based on smooth pursuit, a natural eye movement that occurs when the gaze follows a moving target. Our approach requires no explicit calibration by the user, allowing for more spontaneous interaction and enabling eye input even when calibration is difficult to achieve or maintain. To the best of our knowledge, Leyenes is the first text entry technique based on smooth pursuit that considers both (approximate) gaze speed and position and employs a linear interface instead of the more common circular layouts. The results of the user study we conducted show that the proposed solution is slow but robust, with a very low error rate, which makes it particularly suitable for extemporaneous writing of short text.

In the automotive field, the development of new vehicles is nowadays a challenge for manufacturers, which try to find a trade-off between "beauty of design" and driver safety. Modern cars often have touch displays and hardly any physical buttons. Touch interaction, however, may distract the driver from the road, which is a major safety issue. For the evaluation of the ergonomics of in-vehicle infotainment interfaces, this paper proposes a test protocol and an evaluation method based on eye tracking technology. To the best of our knowledge, there are currently no standard protocols or methodologies for objective assessment of infotainment systems in vehicles. Unlike what often happens, we were able to test our proposals not simply through simulators but implementing actual experiments with real cars, which we believe is an important aspect that characterizes our work.

Gaze-based interaction has until now been almost an exclusive prerogative of the assistive field, as it is considered not sufficiently performing compared to traditional communication methods based on keyboards, pointing devices, and touch screens. However, situations such as the one we are experiencing now due to the COVID-19 pandemic highlight the importance of touchless communication, to minimize the spread of the disease. In this paper, as an example of the potential pervasive use of eye tracking technology in public contexts, we propose and study five interfaces for a gaze-controlled scale, to be used in supermarkets to weigh fruits and vegetables. Given the great heterogeneity of potential users, the interaction must be as simple and intuitive as possible and occur without the need for calibration. The experiments carried out confirm that this goal is achievable and show strengths and weaknesses of the five interfaces.

Gaze-based text entry is undoubtedly one of the most useful applications of eye-tracking technology for human-machine interaction, both in the assistive context (users with severe motor disabilities can exploit such writing modalities to communicate with the world) and as a way to allow touchless text input in everyday life. Different eye-driven text entry methods have been developed to date, and almost all of them require preliminary calibration procedures to work correctly. When a short text, such as a password or a PIN, needs to be entered without using hands or voice, calibration may be perceived as an unnecessary nuisance (and may not be properly maintained in public places due to “ambient noise,” caused, for example, by nearby people). Inadequate calibration may also be a problem in case of assistive uses. In this article we presentSPEye, a calibration-free eye-controlled writing technique based on smooth pursuit. Although its writing speed is significantly lower than that of ordinary calibrated methods, the absence of an initial calibration makes it suitable for short text entry. The technique has been tested through several experiments, obtaining good performances in terms of key strokes per character and total error rate metrics, and receiving positive feedback from the participants in the tests.

Robots, very widespread in industrial settings, are now used for many different purposes. Even if several tasks can be fully automated, the presence of a human operator is often needed. However, direct robot control is generally complex, and it is thus important to explore alternative and comfortable ways of communication. In this paper, we propose a hybrid approach to human-robot interaction that exploits both manual and gaze-based input. While simple, our solution goes in the direction of enhancing the control capabilities of robots by human operators.

As security becomes crucial in an increasing number of industrial contexts, the need arises for new ways to check or authenticate the identity of people. In this paper, we present a method that exploits gaze data to implement a soft biometric technique. Specifically, the user's gaze behavior is inspected during the unconstrained observation of different kinds of static images. The obtained results, achieved using a machine learning approach, are generally satisfying, although more experiments will be necessary to fully confirm the viability of the proposed method.

We present an eye tracking system developed to study the driver's gaze behavior while driving a car. Two main challenges distinguish this project from similar works with analogous purposes: (1) the use of a cheap portable remote eye tracker, normally employed for indoor applications in controlled light conditions; and (2) the use of an ordinary webcam to shoot the road in front of the car. The purpose is to generate a video where the driver's fixations are graphically and dynamically superimposed on the road shot by webcam (taking into account the different viewpoints of driver and camera). The carried-out experiments confirmed that the implemented system, while perfectible, fulfills the requirements initially set.

In this paper we describe a study on image grid display with automatic vertical scrolling. While scroll operations are normally carried out manually by the user, in the context of RSVP (Rapid Serial Visual Presentation) techniques this work considers a presentation mode in which the image grid is automatically scrolled. Through experiments carried out with 50 testers, we have investigated user performance while looking for specific target subjects within large collections of images. Different numbers of columns and scrolling speeds have been considered. The search task implied both clicking on the identified target pictures and simply vocally stating their visual recognition. To this purpose, and to identify possible specific gaze behaviours, eye tracking technology has been exploited. The obtained results show that number of columns and scroll speed do affect search performance. Moreover, the user’s gaze tends to focus on different screen areas depending on the values of these two parameters. Although it is not possible to definitely find an optimal columns–speed combination that is valid in all cases, the particular context of use can suggest feasible solutions according to one’s needs. To the best of our knowledge, image grid display with automatic scrolling has never been studied to date.

Hand detection and gesture recognition are two of the most studied topics in human–computer interaction (HCI). The increasing availability of sensors able to provide real‐time depth measurements, such as time‐of‐flight cameras or the more recent Kinect, has helped researchers to find more and more efficient solutions for these issues. With the main aim to implement effective gesture‐based interaction systems, this study presents an approach to hand detection and tracking that exploits two different video streams: the depth one and the colour one. Both hand and gesture recognition are based only on geometrical and colour constraints, and no learning phase is needed. The use of a Kalman filter to track hands guarantees system robustness also in presence of many persons in the scene. The entire procedure is designed to maintain a low computational cost and is optimised to efficiently execute HCI tasks. As use cases two common applications are described: a virtual keyboard and a three‐dimensional object manipulation virtual environment. These applications have been tested with a representative sample of non‐trained users to assess the usability and flexibility of the system.

Besides proposing a short survey of eye tracking studies directly or indirectly connected with e-learning technology, in this paper we describe our ongoing research primarily directed to the development of an e-learning platform where eye data are exploited to obtain information about the student's “emotional state”. The results of the preliminary tests presented here will be the starting point for more sophisticated experiments we are going to carry out.

Eye gaze interaction for disabled people is often dealt with by designing ad-hoc interfaces, in which the big size of their elements compensates for both the inaccuracy of eye trackers and the instability of the human eye. Unless solutions for reliable eye cursor control are employed, gaze pointing in ordinary graphical operating environments is a very difficult task. In this paper we present an eye-driven cursor for MS Windows which behaves differently according to the context. When the user's gaze is perceived within the desktop or a folder, the cursor can be discretely shifted from one icon to another. Within an application window or where there are no icons, on the contrary, the cursor can be continuously and precisely moved. Shifts in the four directions (up, down, left, right) occur through dedicated buttons. To increase user awareness of the currently pointed spot on the screen while continuously moving the cursor, a replica of the spot is provided within the active direction button, resulting in improved pointing performance.

Rapid Serial Visual Presentation (RSVP) is now a well-established category of image display methods. In this paper we compare four RSVP techniques when applied to very large collections of images (thousands), in order to extract the highest quantity of items that match a textual description. We report on experiments with more than 30 testers, in which we exploit an eye tracking system to perform the selection of images, thus obtaining quantitative and qualitative data about the efficacy of each presentation mode with respect to this task. Our study aims at confirming the feasibility and convenience of an eye tracking approach for effective image selection in RSVP techniques, compared to the mouse-click traditional selection method, in view of a future where eye trackers might become nearly as common as LCD displays are now. We propose an interpretation of the experimental data and provide short considerations on technical issues.

To date, several eye input methods have been developed, which, however, are usually designed for specific purposes (e.g. typing) and require dedicated graphical interfaces. In this paper we present Eye-S, a system that allows general input to be provided to the computer through a pure eye-based approach. Thanks to the "eye graffiti" communication style adopted, the technique can be used both for writing and for generating other kinds of commands. In Eye-S, letters and general eye gestures are created through sequences of fixations on nine areas of the screen, which we call hotspots. Being usually not visible, such sensitive regions do not interfere with other applications, that can therefore exploit all the available display space.

Userspsila eyes can be a meaningful source of information for e-learning systems. What we look at, and the way we do that, can in fact be exploited to improve the learning process, disclosing information which would otherwise remain concealed. In this paper we describe an e-learning environment where eye tracking is used to observe user behavior, in order to adapt content presentation in real-time. To achieve such purpose, we consider both the way learning activities are carried out and those eye signals that can be related to the userpsilas ldquoemotional statesrdquo.

In this paper we describe some alternative methods intended for rapid and effective browsing of large collections of images. Specifically, we address the user who, not having a clear idea about what to search, needs to explore the entire image database to identify what he or she likes. The purpose of our approaches is to find techniques characterized by good trade-offs between browsing time and quality of the exploration.

We present here a working prototypal Web site where a new approach for on-line interaction has been implemented and tested. Our intention is to improve the usability of Web sites (particularly e-commerce ones) by letting users directly interact with pictorial representations of objects (e.g. items on sale) in a way that resembles their behavior in real-world stores. Our discussion focuses both on the manifold usability benefits of the approach and on practical implementation issues.

The last few decades have witnessed the birth and growth of a new sensibility to transportation efficiency. In particular the need for efficient and improved people and goods mobility has pushed researchers to address the problem of intelligent transportation systems. This paper surveys the most advanced approaches to (partial) customization of the road following task, using on-board systems based on artificial vision. The functionalities of lane detection, obstacle detection and pedestrian detection are described and classified, and their possible application in future road vehicles is discussed.

In various circumstances, it is possible to arrive at the need to specify sequences of operations that a “machine” has to perform to achieve a purpose. This paper will present VISPS, a visual system originally designed to specify mission plans for the SARA autonomous submarine robot. Although this is the particular setting for which the system had been initially devised, thanks to its flexibility it can be easily configured to adjust to different contexts and situations, even if it is always based on the same simple basic visual mechanism.