The proposed EU regulation for Artificial Intelligence (AI), the AI Act, has sparked some debate about role of explainable (XAI) in high-risk systems. Some argue that black-box models will have to be replaced with transparent ones, others using XAI techniques might help achieving compliance. This work aims bring clarity as regards context Act and focuses particular on requirements transparency human oversight. After outlining key points describing current limitations techniques, this paper...

Engagement is crucial to designing intelligent systems that can adapt the characteristics of their users. This paper focuses on automatic analysis and classification engagement based humans' robot's personality profiles in a triadic human-human-robot interaction setting. More explicitly, we present study involves two participants interacting with humanoid robot, investigate how participants' personalities be used together predict state each participant. The fully system first trained Big...

Although deep face recognition has achieved impressive results in recent years, controversy arisen regarding racial and gender bias of the models, questioning their deployment into sensitive scenarios. This work quantifies for first time demographic imbalance popular public datasets terms identity, ethnicity. We also publicly release DemogPairs, a new validation set with 10.8K facial images 58.3M identity verification pairs, distributed demographically-balanced folds Asian, Black White...

This article discusses transparency obligations introduced in the Artificial Intelligence Act, recently proposed European regulatory framework for artificial intelligence (AI). An analysis of extent to which current approaches AI documentation satisfy requirements is presented and their suitability as a basis future technical standards assessed.

The recent introduction in the market of low cost yet high fidelity head-mounted displays with stereoscopic 3D perspective such as Oculus Rift opens door to novel virtual reality experiences for gaming. However, QoE metrics and methodologies this kind platforms are still unexplored. In paper, a comparative study on how these impact gaming regard conventional 2D PC screens is presented. Through an experiment where we invited 22 users play forklift driving serious game both environments,...

Super-resolution (SR) has achieved great success due to the development of deep convolutional neural networks (CNNs). However, as depth and width increase, CNN-based SR methods have been faced with challenge computational complexity in practice. More-over, most train a dedicated model for each target resolution, losing generality increasing memory requirements. To address these limitations we introduce OverNet, but lightweight network solve SISR at arbitrary scale factors single model. We...

Abstract This work focuses on facial processing , which refers to artificial intelligence (AI) systems that take images or videos as input data and perform some AI-driven obtain higher-level information (e.g. a person’s identity, emotions, demographic attributes) newly generated imagery with modified attributes). Facial tasks, such face detection, identification, expression recognition attribute manipulation, are generally studied separate research fields without considering particular...

Abstract Despite recent efforts by the Artificial Intelligence (AI) community to move towards standardised procedures for documenting models, methods, systems or datasets, there is currently no methodology focused on use cases aligned with risk-based approach of European AI Act (AI Act). In this paper, we propose a new framework documentation that call case cards , based modelling included in Unified Markup Language (UML) standard. Unlike other methodologies, focus intended purpose and...

The identification of non-verbal emotional signals, and especially facial expressions, is essential for successful social communication among humans. Previous research has reported an age-related decline in emotion identification, argued socio-emotional or aging-brain model explanations. However, more perceptual differences the gaze strategies that accompany processing with advancing age have been under-explored yet. In this study, 22 young (22.2 years) older (70.4 adults were instructed to...

Abstract Nowadays, Artificial Intelligence (AI) is present in all sectors of the economy. Consequently, both data-the raw material used to build AI systems- and have an unprecedented impact on society there a need ensure that they work for its benefit. For this reason, European Union has put data trustworthy at center recent legislative initiatives. An important element these regulations transparency, understood as provision information relevant stakeholders support their understanding...

With the upcoming enforcement of EU AI Act, documentation high-risk systems and their risk management information will become a legal requirement playing pivotal role in demonstration compliance. Despite its importance, there is lack standards guidelines to assist with drawing up aligned Act. This paper aims address this gap by providing an in-depth analysis Act's provisions regarding technical documentation, wherein we particularly focus on management. On basis analysis, propose Cards as...

Interpersonal human-human interaction is a dynamical exchange and coordination of social signals, feelings emotions usually performed through across multiple modalities such as facial expressions, gestures, language. Developing machines able to engage humans in rich natural interpersonal interactions requires capturing dynamics. This paper addresses primitive emotional contagion during dyadic which roles are prefixed. Primitive was defined the tendency people have automatically mimic...

The interpretation of user facial expressions is a very useful method for emotional sensing and it constitutes an indispensable part affective Human Computer Interface designs. Facial are often classified into one several basic emotion categories. This categorical approach seems poor to treat faces with blended emotions, as well measure the intensity given emotion. paper presents effective system classification, where evaluated psychological 2-dimensional continuous approach. At its output,...

Visual attention constitutes a very important feature of the human visual system (HVS). Every day when watching videos, images or browsing Internet, people are confronted with more information than they able to process, and analyze only part in front them. In parallel, crowdsourcing has become particularly hot topic, enabling scale subjective experiments large crowd diversity terms nationalities, social background, age, etc. This paper describes novel framework aim bridge these two fields,...

Evaluating (audio)visual quality and Quality of Experience (QoE) from the user's perspective, has become a key element in optimizing users' experiences their quality. Traditionally, focus lies on how multi-level features are perceived by human user. The interest however gradually expanded towards cognitive, affective behavioral processes that may impact on, be an of, or influenced QoE, which have been underinvestigated so far. In addition, there is major discrepancy between new, broadly...

Most current multi-object trackers focus on short-term tracking, and are based deep complex systems that often cannot operate in real-time, making them impractical for video-surveillance. In this paper we present a long-term, multi-face tracking architecture conceived working crowded contexts where faces the only visible part of person. Our system benefits from advances fields face detection recognition to achieve long-term is particularly unconstrained motion occlusions people. It follows...

The success of affective interfaces lies in the fusion emotional information coming from different modalities. This paper proposes a scalable methodology for fusing multiple affect sensing modules, allowing subsequent addition new modules without having to retrain existing ones. It relies on 2-dimensional model and is able output continuous path characterizing user's progress over time.