Silvan Mertes

Andreas Triantafyllopoulos, Björn W. Schuller, Gökçe İymen et al.

Speech is the fundamental mode of human communication, and its synthesis has long been a core priority in human-computer interaction research. In recent years, machines have managed to master the art of generating speech that is understandable by humans. But the linguistic content of an utterance encompasses only a part of its meaning. Affect, or expressivity, has the capacity to turn speech into a medium capable of conveying intimate thoughts, feelings, and emotions -- aspects that are essential for engaging and naturalistic interpersonal communication. While the goal of imparting expressivity to synthesised utterances has so far remained elusive, following recent advances in text-to-speech synthesis, a paradigm shift is well under way in the fields of affective speech synthesis and conversion as well. Deep learning, as the technology which underlies most of the recent advances in artificial intelligence, is spearheading these efforts. In the present overview, we outline ongoing trends and summarise state-of-the-art approaches in an attempt to provide a comprehensive overview of this exciting field.

Luuk H. Boulogne, Julian Lorenz, Daniel Kienzle et al.

Challenges drive the state-of-the-art of automated medical image analysis. The quantity of public training data that they provide can limit the performance of their solutions. Public access to the training methodology for these solutions remains absent. This study implements the Type Three (T3) challenge format, which allows for training solutions on private data and guarantees reusable training methodologies. With T3, challenge organizers train a codebase provided by the participants on sequestered training data. T3 was implemented in the STOIC2021 challenge, with the goal of predicting from a computed tomography (CT) scan whether subjects had a severe COVID-19 infection, defined as intubation or death within one month. STOIC2021 consisted of a Qualification phase, where participants developed challenge solutions using 2000 publicly available CT scans, and a Final phase, where participants submitted their training methodologies with which solutions were trained on CT scans of 9724 subjects. The organizers successfully trained six of the eight Final phase submissions. The submitted codebases for training and running inference were released publicly. The winning solution obtained an area under the receiver operating characteristic curve for discerning between severe and non-severe COVID-19 of 0.815. The Final phase solutions of all finalists improved upon their Qualification phase solutions.HSUXJM-TNZF9CHSUXJM-TNZF9C

Alexander Heimerl, Silvan Mertes, Tanja Schneeberger et al.

Job interviews are usually high-stakes social situations where professional and behavioral skills are required for a satisfactory outcome. Professional job interview trainers give educative feedback about the shown behavior according to common standards. This feedback can be helpful concerning the improvement of behavioral skills needed for job interviews. A technological approach for generating such feedback might be a playful and low-key starting point for job interview training. Therefore, we extended an interactive virtual job interview training system with a Generative Adversarial Network (GAN)-based approach that first detects behavioral weaknesses and subsequently generates personalized feedback. To evaluate the usefulness of the generated feedback, we conducted a mixed-methods pilot study using mock-ups from the job interview training system. The overall study results indicate that the GAN-based generated behavioral feedback is helpful. Moreover, participants assessed that the feedback would improve their job interview performance.

Lennart Eing, Cristina Luna-Jiménez, Silvan Mertes et al.

This paper introduces a novel application of Video Joint-Embedding Predictive Architectures (V-JEPAs) for Facial Expression Recognition (FER). Departing from conventional pre-training methods for video understanding that rely on pixel-level reconstructions, V-JEPAs learn by predicting embeddings of masked regions from the embeddings of unmasked regions. This enables the trained encoder to not capture irrelevant information about a given video like the color of a region of pixels in the background. Using a pre-trained V-JEPA video encoder, we train shallow classifiers using the RAVDESS and CREMA-D datasets, achieving state-of-the-art performance on RAVDESS and outperforming all other vision-based methods on CREMA-D (+1.48 WAR). Furthermore, cross-dataset evaluations reveal strong generalization capabilities, demonstrating the potential of purely embedding-based pre-training approaches to advance FER. We release our code at https://github.com/lennarteingunia/vjepa-for-fer.

Alexander Heimerl, Pooja Prajod, Silvan Mertes et al.

We present a multi-modal stress dataset that uses digital job interviews to induce stress. The dataset provides multi-modal data of 40 participants including audio, video (motion capturing, facial recognition, eye tracking) as well as physiological information (photoplethysmography, electrodermal activity). In addition to that, the dataset contains time-continuous annotations for stress and occurred emotions (e.g. shame, anger, anxiety, surprise). In order to establish a baseline, five different machine learning classifiers (Support Vector Machine, K-Nearest Neighbors, Random Forest, Long-Short-Term Memory Network) have been trained and evaluated on the proposed dataset for a binary stress classification task. The best-performing classifier achieved an accuracy of 88.3% and an F1-score of 87.5%.

Tobias Huber, Maximilian Demmler, Silvan Mertes et al.

Counterfactual explanations are a common tool to explain artificial intelligence models. For Reinforcement Learning (RL) agents, they answer "Why not?" or "What if?" questions by illustrating what minimal change to a state is needed such that an agent chooses a different action. Generating counterfactual explanations for RL agents with visual input is especially challenging because of their large state spaces and because their decisions are part of an overarching policy, which includes long-term decision-making. However, research focusing on counterfactual explanations, specifically for RL agents with visual input, is scarce and does not go beyond identifying defective agents. It is unclear whether counterfactual explanations are still helpful for more complex tasks like analyzing the learned strategies of different agents or choosing a fitting agent for a specific task. We propose a novel but simple method to generate counterfactual explanations for RL agents by formulating the problem as a domain transfer problem which allows the use of adversarial learning techniques like StarGAN. Our method is fully model-agnostic and we demonstrate that it outperforms the only previous method in several computational metrics. Furthermore, we show in a user study that our method performs best when analyzing which strategies different agents pursue.

Silvan Mertes, Christina Karle, Tobias Huber et al.

Explanation mechanisms from the field of Counterfactual Thinking are a widely-used paradigm for Explainable Artificial Intelligence (XAI), as they follow a natural way of reasoning that humans are familiar with. However, all common approaches from this field are based on communicating information about features or characteristics that are especially important for an AI's decision. We argue that in order to fully understand a decision, not only knowledge about relevant features is needed, but that the awareness of irrelevant information also highly contributes to the creation of a user's mental model of an AI system. Therefore, we introduce a new way of explaining AI systems. Our approach, which we call Alterfactual Explanations, is based on showing an alternative reality where irrelevant features of an AI's input are altered. By doing so, the user directly sees which characteristics of the input data can change arbitrarily without influencing the AI's decision. We evaluate our approach in an extensive user study, revealing that it is able to significantly contribute to the participants' understanding of an AI. We show that alterfactual explanations are suited to convey an understanding of different aspects of the AI's reasoning than established counterfactual explanation methods.

Silvan Mertes, Andreas Margraf, Steffen Geinitz et al.

Visual inspection software has become a key factor in the manufacturing industry for quality control and process monitoring. Semantic segmentation models have gained importance since they allow for more precise examination. These models, however, require large image datasets in order to achieve a fair accuracy level. In some cases, training data is sparse or lacks of sufficient annotation, a fact that especially applies to highly specialized production environments. Data augmentation represents a common strategy to extend the dataset. Still, it only varies the image within a narrow range. In this article, a novel strategy is proposed to augment small image datasets. The approach is applied to surface monitoring of carbon fibers, a specific industry use case. We apply two different methods to create binary labels: a problem-tailored trigonometric function and a WGAN model. Afterwards, the labels are translated into color images using pix2pix and used to train a U-Net. The results suggest that the trigonometric function is superior to the WGAN model. However, a precise examination of the resulting images indicate that WGAN and image-to-image translation achieve good segmentation results and only deviate to a small degree from traditional data augmentation. In summary, this study examines an industry application of data synthesization using generative adversarial networks and explores its potential for monitoring systems of production environments. \keywords{Image-to-Image Translation, Carbon Fiber, Data Augmentation, Computer Vision, Industrial Monitoring, Adversarial Learning.

Tobias Hallmen, Silvan Mertes, Dominik Schiller et al.

Affective speech analysis is an ongoing topic of research. A relatively new problem in this field is the analysis of vocal bursts, which are nonverbal vocalisations such as laughs or sighs. Current state-of-the-art approaches to address affective vocal burst analysis are mostly based on wav2vec2 or HuBERT features. In this paper, we investigate the use of the wav2vec successor data2vec in combination with a multitask learning pipeline to tackle different analysis problems at once. To assess the performance of our efficient multitask learning architecture, we participate in the 2022 ACII Affective Vocal Burst Challenge, showing that our approach substantially outperforms the baseline established there in three different subtasks.

Silvan Mertes, Dominik Schiller, Florian Lingenfelser et al.

Generative adversarial networks offer the possibility to generate deceptively real images that are almost indistinguishable from actual photographs. Such systems however rely on the presence of large datasets to realistically replicate the corresponding domain. This is especially a problem if not only random new images are to be generated, but specific (continuous) features are to be co-modeled. A particularly important use case in \emph{Human-Computer Interaction} (HCI) research is the generation of emotional images of human faces, which can be used for various use cases, such as the automatic generation of avatars. The problem hereby lies in the availability of training data. Most suitable datasets for this task rely on categorical emotion models and therefore feature only discrete annotation labels. This greatly hinders the learning and modeling of smooth transitions between displayed affective states. To overcome this challenge, we explore the potential of label interpolation to enhance networks trained on categorical datasets with the ability to generate images conditioned on continuous features.

Silvan Mertes, Dominik Schiller, Michael Dietz et al.

In the field of affective computing, where research continually advances at a rapid pace, the demand for user-friendly tools has become increasingly apparent. In this paper, we present the AffectToolbox, a novel software system that aims to support researchers in developing affect-sensitive studies and prototypes. The proposed system addresses the challenges posed by existing frameworks, which often require profound programming knowledge and cater primarily to power-users or skilled developers. Aiming to facilitate ease of use, the AffectToolbox requires no programming knowledge and offers its functionality to reliably analyze the affective state of users through an accessible graphical user interface. The architecture encompasses a variety of models for emotion recognition on multiple affective channels and modalities, as well as an elaborate fusion system to merge multi-modal assessments into a unified result. The entire system is open-sourced and will be publicly available to ensure easy integration into more complex applications through a well-structured, Python-based code base - therefore marking a substantial contribution toward advancing affective computing research and fostering a more collaborative and inclusive environment within this interdisciplinary field.

Silvan Mertes, Tobias Huber, Christina Karle et al.

In this paper, we demonstrate the feasibility of alterfactual explanations for black box image classifiers. Traditional explanation mechanisms from the field of Counterfactual Thinking are a widely-used paradigm for Explainable Artificial Intelligence (XAI), as they follow a natural way of reasoning that humans are familiar with. However, most common approaches from this field are based on communicating information about features or characteristics that are especially important for an AI's decision. However, to fully understand a decision, not only knowledge about relevant features is needed, but the awareness of irrelevant information also highly contributes to the creation of a user's mental model of an AI system. To this end, a novel approach for explaining AI systems called alterfactual explanations was recently proposed on a conceptual level. It is based on showing an alternative reality where irrelevant features of an AI's input are altered. By doing so, the user directly sees which input data characteristics can change arbitrarily without influencing the AI's decision. In this paper, we show for the first time that it is possible to apply this idea to black box models based on neural networks. To this end, we present a GAN-based approach to generate these alterfactual explanations for binary image classifiers. Further, we present a user study that gives interesting insights on how alterfactual explanations can complement counterfactual explanations.

Dominik Müller, Niklas Schröter, Silvan Mertes et al.

COVID-19 presence classification and severity prediction via (3D) thorax computed tomography scans have become important tasks in recent times. Especially for capacity planning of intensive care units, predicting the future severity of a COVID-19 patient is crucial. The presented approach follows state-of-theart techniques to aid medical professionals in these situations. It comprises an ensemble learning strategy via 5-fold cross-validation that includes transfer learning and combines pre-trained 3D-versions of ResNet34 and DenseNet121 for COVID19 classification and severity prediction respectively. Further, domain-specific preprocessing was applied to optimize model performance. In addition, medical information like the infection-lung-ratio, patient age, and sex were included. The presented model achieves an AUC of 79.0% to predict COVID-19 severity, and 83.7% AUC to classify the presence of an infection, which is comparable with other currently popular methods. This approach is implemented using the AUCMEDI framework and relies on well-known network architectures to ensure robustness and reproducibility.

Fabio Hellmann, Silvan Mertes, Mohamed Benouis et al.

In recent years, the increasing availability of personal data has raised concerns regarding privacy and security. One of the critical processes to address these concerns is data anonymization, which aims to protect individual privacy and prevent the release of sensitive information. This research focuses on the importance of face anonymization. Therefore, we introduce GANonymization, a novel face anonymization framework with facial expression-preserving abilities. Our approach is based on a high-level representation of a face, which is synthesized into an anonymized version based on a generative adversarial network (GAN). The effectiveness of the approach was assessed by evaluating its performance in removing identifiable facial attributes to increase the anonymity of the given individual face. Additionally, the performance of preserving facial expressions was evaluated on several affect recognition datasets and outperformed the state-of-the-art methods in most categories. Finally, our approach was analyzed for its ability to remove various facial traits, such as jewelry, hair color, and multiple others. Here, it demonstrated reliable performance in removing these attributes. Our results suggest that GANonymization is a promising approach for anonymizing faces while preserving facial expressions.

Tobias Huber, Silvan Mertes, Stanislava Rangelova et al.

Virtual Reality (VR) games that feature physical activities have been shown to increase players' motivation to do physical exercise. However, for such exercises to have a positive healthcare effect, they have to be repeated several times a week. To maintain player motivation over longer periods of time, games often employ Dynamic Difficulty Adjustment (DDA) to adapt the game's challenge according to the player's capabilities. For exercise games, this is mostly done by tuning specific in-game parameters like the speed of objects. In this work, we propose to use experience-driven Procedural Content Generation for DDA in VR exercise games by procedurally generating levels that match the player's current capabilities. Not only finetuning specific parameters but creating completely new levels has the potential to decrease repetition over longer time periods and allows for the simultaneous adaptation of the cognitive and physical challenge of the exergame. As a proof-of-concept, we implement an initial prototype in which the player must traverse a maze that includes several exercise rooms, whereby the generation of the maze is realized by a neural network. Passing those exercise rooms requires the player to perform physical activities. To match the player's capabilities, we use Deep Reinforcement Learning to adjust the structure of the maze and to decide which exercise rooms to include in the maze. We evaluate our prototype in an exploratory user study utilizing both biodata and subjective questionnaires.

Pol van Rijn, Silvan Mertes, Dominik Schiller et al.

Recent TTS systems are able to generate prosodically varied and realistic speech. However, it is unclear how this prosodic variation contributes to the perception of speakers' emotional states. Here we use the recent psychological paradigm 'Gibbs Sampling with People' to search the prosodic latent space in a trained GST Tacotron model to explore prototypes of emotional prosody. Participants are recruited online and collectively manipulate the latent space of the generative speech model in a sequentially adaptive way so that the stimulus presented to one group of participants is determined by the response of the previous groups. We demonstrate that (1) particular regions of the model's latent space are reliably associated with particular emotions, (2) the resulting emotional prototypes are well-recognized by a separate group of human raters, and (3) these emotional prototypes can be effectively transferred to new sentences. Collectively, these experiments demonstrate a novel approach to the understanding of emotional speech by providing a tool to explore the relation between the latent space of generative models and human semantics.

Silvan Mertes, Tobias Huber, Katharina Weitz et al.

With the ongoing rise of machine learning, the need for methods for explaining decisions made by artificial intelligence systems is becoming a more and more important topic. Especially for image classification tasks, many state-of-the-art tools to explain such classifiers rely on visual highlighting of important areas of the input data. Contrary, counterfactual explanation systems try to enable a counterfactual reasoning by modifying the input image in a way such that the classifier would have made a different prediction. By doing so, the users of counterfactual explanation systems are equipped with a completely different kind of explanatory information. However, methods for generating realistic counterfactual explanations for image classifiers are still rare. Especially in medical contexts, where relevant information often consists of textural and structural information, high-quality counterfactual images have the potential to give meaningful insights into decision processes. In this work, we present GANterfactual, an approach to generate such counterfactual image explanations based on adversarial image-to-image translation techniques. Additionally, we conduct a user study to evaluate our approach in an exemplary medical use case. Our results show that, in the chosen medical use-case, counterfactual explanations lead to significantly better results regarding mental models, explanation satisfaction, trust, emotions, and self-efficacy than two state-of-the-art systems that work with saliency maps, namely LIME and LRP.

Dominik Schiller, Silvan Mertes, Elisabeth André

Traditional approaches to automatic emotion recognition are relying on the application of handcrafted features. More recently however the advent of deep learning enabled algorithms to learn meaningful representations of input data automatically. In this paper, we investigate the applicability of transferring knowledge learned from large text and audio corpora to the task of automatic emotion recognition. To evaluate the practicability of our approach, we are taking part in this year's Interspeech ComParE Elderly Emotion Sub-Challenge, where the goal is to classify spoken narratives of elderly people with respect to the emotion of the speaker. Our results show that the learned feature representations can be effectively applied for classifying emotions from spoken language. We found the performance of the features extracted from the audio signal to be not as consistent as those that have been extracted from the transcripts. While the acoustic features achieved best in class results on the development set, when compared to the baseline systems, their performance dropped considerably on the test set of the challenge. The features extracted from the text form, however, are showing promising results on both sets and are outperforming the official baseline by 5.7 percentage points unweighted average recall.