Dr. Tobias Baur

Research Associate
Human - Centered Multimedia
Phone: +49 821 – 598 2305
Email: baur@hcm-lab.de
Room: 2044 (N)
Visiting hours: By agreement
Address: Universitätsstraße 6a, 86159 Augsburg

Research Topics

  • Affective Computing
  • Artificial Emotional Intelligence
  • eXplainable AI
  • Semi-Supervised Active Machine Learning
  • HCI

Current Lectures

(applied filters: | Semester: current | Institution: Multimodale Mensch-Technik-Interaktion | Lehrende: Tobias Baur | Typen: Vorlesung, Vorlesung + Übung, Praktikum)
name semester type
Praktikum Interactive Machine Learning winter semester 2019/20 Praktikum

Awards

  • 8th International Conference on Affective Computing & Intelligent Interaction (ACII 2019)  -  Best Paper Award
  • ACM Multimedia 2013 Open Source Challenge - Honorable Mention
  • UMAP 2013 - Most Participative Demo  Award
  • IDGEI 2013, FDG - Best Paper Award

Talks / Keynotes

Keynote/Tutorial

"NOVA - A tool for eXplainable Cooperative Machine Learning"

ECML/PKDD Summer School Würzburg ( EPSS19)

Demonstration

"I see what you did there: Understanding when to trust a ML model with NOVA"

8th International Conference on Affective Computing & Intelligent Interaction ( ACII 2019)

Tutorial

"Social Signal Interpretation for Virtual Agents: Basic Concepts and Implementation in the SSI Framework"

Fifteenth International Conference on Intelligent Virtual Agents ( IVA 2015) Delft, Netherlands, 2015

 

Networks / Memberships

  • Association for the advancement of Affecitve Computing   ( AAAC)
  • IHK AI Network für die bayerisch-schwäbischen KI-Verantwortlichen ( AI-Network Augsburg)

Editorials

  • Guest Editor: Multimodal Technologies and Interaction: Special Issue "Modeling Interaction with Virtual Characters"   
  • PC: MRC 2020: Eleventh International Workshop Modelling and Reasoning in Context, Santiago de Compostela, Spain, June 8-12, 2020

Software Projects

zwei virtuelle Charaktere im Dialog, über dem Bayes'schen Netz, welches ihr Verhalten steuert
© University of Augsburg

NOVA

NOVA is a tool for annotating and analyzing behaviours in social interactions. A main feature of NOVA is that it allows to employ a collaborative annotation database where annotation work can be split between multiple sides, but also between a human annotator and a machine by supporting human annotators with machine learning techniques already during the annotation task - A process we call Collaborative Machine Learning.

NOVA allows framewise labeling for a precise coding experience, and value-continuous annotations for labeling e.g emotions or social attitudes. The interface is customizable and allows loading and labeling data of multiple persons. The Cooperative Machine Learning capabilities allow to train and evaluate machine learning models, such as Support Vector machines or Artificial neural networks directly from the interface with both, a session completion step, where a model is trained on the first minutes of an annotated sessions to predict the remaining session, and a session transfer step where a model is trained on multiple sessions to predict completly unknown data. With the help of human input the models can then be refined. NOVA further integrates eXplainable AI algorithms to give inisghts into a Machine Learning model’s inner workings.

 

 

Project Github Page

 

Article on eXplainable Cooperative Machine Learning with NOVA at KI-Künstliche Intelligenz

 

 

 

© University of Augsburg

The Social Signal Interpretation Framework

 

The Social Signal Interpretation (SSI) framework offers tools to record, analyse and recognize human behavior in real-time, such as gestures, mimics, head nods, and emotional speech. Following a patch-based design pipelines are set up from autonomic components and allow the parallel and synchronized processing of sensor data from multiple input devices. In particularly SSI supports the machine learning pipeline in its full length and offers a graphical interface that assists a user to collect own training corpora and obtain personalized models. In addition to a large set of built-in components SSI also encourages developers to extend available tools with new functions. For inexperienced users an easy-to-use XML editor is available to draft and run pipelines without special programming skills. SSI is written in C++ and optimized to run on computer systems with multiple CPUs.

 

Project Github Page

Current funded projects

DEEP: Mehrschichtige Verarbeitung von Emotionen für Soziale Agenten Kombination einer Interpretation von Sozialen Signalen und einem Computermodell für Emotionen von Dialogpartnern

Finished funded projects

ARIA Valuspa: Artificial Retrieval of Information Assistants – Virtual Agents with Linguistic Understanding, Social skills, and Personalised Aspects
EMPAT: Empathische Trainingsbegleiter für den Bewerbungsprozess
TARDIS: Training young Adult's Regulation of emotions and Development of social Interaction Skills
ILHAIRE: Incorporating Laughter into Human Avatar Interactions: Research and Experiments
FMLA: Forum Maschinelles Lernen Augsburg

Search