A5015879753- Emotion and Mood Recognition
- Face recognition and analysis
- Human Pose and Action Recognition
- Gaze Tracking and Assistive Technology
- Generative Adversarial Networks and Image Synthesis
- EEG and Brain-Computer Interfaces
- Deception detection and forensic psychology
- Adversarial Robustness in Machine Learning
- Anomaly Detection Techniques and Applications
- Video Surveillance and Tracking Methods
- Hate Speech and Cyberbullying Detection
- 3D Shape Modeling and Analysis
- Multimodal Machine Learning Applications
- Hand Gesture Recognition Systems
- ECG Monitoring and Analysis
- Functional Brain Connectivity Studies
- Imbalanced Data Classification Techniques
- Retinal Diseases and Treatments
- Music and Audio Processing
- Psychopathy, Forensic Psychiatry, Sexual Offending
- Model Reduction and Neural Networks
- Cybercrime and Law Enforcement Studies
- Heart Rate Variability and Autonomic Control
- Gait Recognition and Analysis
- Machine Learning in Healthcare
University of Stuttgart
2021-2023
Universitat Pompeu Fabra
2019-2022
University of Augsburg
2020-2021
The human face is one of the most visible features our unique identity as individuals. Interestingly, monozygotic twins share almost identical facial traits and same DNA sequence but could exhibit differences in other biometrical parameters. expansion world wide web possibility to exchange pictures humans across planet has increased number people identified online virtual or doubles that are not family related. Herein, we have characterized detail a set "look-alike" humans, defined by...
In recent years, affective computing and its applications have become a fast-growing research topic. Furthermore, the rise of deep learning has introduced significant improvements in emotion recognition system compared to classical methods. this work, we propose multi-modal model based on techniques using combination peripheral physiological signals facial expressions. Moreover, present an improvement proposed models by introducing latent features extracted from our internal Bio Auto-Encoder...
Human facial tracking is an important task in computer vision, which has recently lost pace compared to other analysis tasks. The majority of current available tracker possess two major limitations: their little use temporal information and the widespread handcrafted features, without taking full advantage large annotated datasets that have become available. In this paper we present a fully end-to-end model based on state art deep architectures can be effectively trained from landmark...
Automatic kinship recognition using Computer Vision, which aims to infer the blood relationship between individuals by only comparing their facial features, has started gain attention recently. The introduction of large datasets, such as Family In Wild (FIW), allowed scale dataset modeling state art deep learning models. Among other tasks, family classification task is lacking any significant progress due its increasing difficulty in relation member size. Furthermore, most current of-the-art...
Facial alignment is an essential task for many higher level facial analysis applications, such as animation, human activity recognition and - computer interaction. Although the recent availability of big datasets powerful deep-learning approaches have enabled major improvements on state art accuracy, performance current can severely deteriorate when dealing with images in highly unconstrained conditions, which limits real-life applicability models. In this paper, we propose a composite...
The development of facial alignment models is growing rapidly thanks to the availability large landmarked datasets and powerful deep learning models. However, important challenges still remain for work on images under extreme conditions, such as severe occlusions or variations in pose illumination. Current attempts overcome this limitation have mainly focused building robust feature extractors with assumption that model will be able discard noise select only meaningful features. an ignores...
Affective Computing has recently attracted the attention of research community, due to its numerous applications in diverse areas.In this context, emergence video-based data allows enrich widely used spatial features with inclusion temporal information.However, such spatio-temporal modelling often results very high-dimensional feature spaces and large volumes data, making training difficult time consuming.This paper addresses these shortcomings by proposing a novel model that efficiently...
The interest in automatic emotion recognition and the larger field of Affective Computing has recently gained momentum. current emergence large, video-based affect datasets offering rich multi-modal inputs facilitates development deep learning-based models for analysis that currently holds state art. However, recent approaches to process these modalities cannot fully exploit them due use oversimplified fusion schemes. Furthermore, efficient temporal information inherent huge data are also...
There is a growing interest in affective computing research nowadays given its crucial role bridging humans with computers. This progress has recently been accelerated due to the emergence of bigger dataset. One recent advance this field use adversarial learning improve model through augmented samples. However, latent features, which feasible learning, not largely explored, yet. technique may also performance models, as analogously demonstrated related fields, such computer vision. To expand...
Lie detection is considered a concern for everyone in their day-to-day life, given its impact on human interactions. Thus, people normally pay attention to both what interlocutors are saying and visual appearance, including the face, find any signs that indicate whether or not person telling truth. While automatic lie may help us understand these lying characteristics, current systems still fairly limited, partly due lack of adequate datasets evaluate performance realistic scenarios. In this...
Gesture is an important mean of non-verbal communication, with visual modality allows human to convey information during interaction, facilitating peoples and human-machine interactions. However, it considered difficult automatically recognise gestures. In this work, we explore three different means hand signs using deep learning: supervised learning based methods, self-supervised methods visualisation techniques applied 3D moving skeleton data. Self-supervised used train fully connected,...
Video captioning aims to describe video contents using natural language format that involves understanding and interpreting scenes, actions events occurs simultaneously on the view. Current approaches have mainly concentrated visual cues, often neglecting rich information available from other important modality of audio information, including their inter-dependencies. In this work, we introduce a novel method trained with multi-modal contrastive loss emphasizes both integration...
Multistep prediction models are essential for the simulation and model-predictive control of dynamical systems. Verifying safety such is a multi-faceted problem requiring both system-theoretic guarantees as well establishing trust with human users. In this work, we propose novel approach, ReLiNet (Recurrent Linear Parameter Varying Network), to ensure multistep Our approach simplifies recurrent neural network switched linear system that constrained guarantee exponential stability, which acts...
Increasing interest in complementary therapies prompts analysis of the objective impact on human physiology. Polarity Therapy (PT) is a branch medicine that relates to energy field therapies. Although previous clinical work has provided evidence patients, present analyzes, for first time, short-term systematic investigation such therapy. Several physiological signals were collected from 25 consecutive chronic anxiety patients seen an outpatient clinic before and after PT, which included...
Current advance of internet allows rapid dissemination information, accelerating the progress on wide spectrum society. This has been done mainly through use website interface with inherent unique human interactions. In this regards usability analysis becomes a central part to improve However, not yet quantitatively evaluated user perception during interaction, especially when dealing range tasks. study, we perform quantitative websites based their usage and relevance. We do by reporting...
In recent years, Affective Computing and its applications have become a fast-growing research topic. Furthermore, the rise of Deep Learning has introduced significant improvements in emotion recognition system compared to classical methods. this work, we propose multi-modal model based on deep learning techniques using combination peripheral physiological signals facial expressions. Moreover, present an improvement proposed models by introducing latent features extracted from our internal...
There is a growing interest in affective computing research nowadays given its crucial role bridging humans with computers. This progress has been recently accelerated due to the emergence of bigger data. One recent advance this field use adversarial learning improve model through augmented samples. However, latent features, which feasible learning, not largely explored, yet. technique may also performance models, as analogously demonstrated related fields, such computer vision. To expand...
World-wide-web, with the website and webpage as main interface, facilitates dissemination of important information. Hence it is crucial to optimize them for better user interaction, which primarily done by analyzing users' behavior, especially eye-gaze locations. However, gathering these data still considered be labor time intensive. In this work, we enable development automatic estimations given a screenshots input. This curation unified dataset that consists screenshots, heatmap website's...
Lie detection is considered a concern for everyone in their day to life given its impact on human interactions. Thus, people normally pay attention both what interlocutors are saying and also visual appearances, including faces, try find any signs that indicate whether the person telling truth or not. While automatic lie may help us understand this lying characteristics, current systems still fairly limited, partly due lack of adequate datasets evaluate performance realistic scenarios. In...