@inproceedings{MendelSouzaJrRauberetal., author = {Mendel, Robert and Souza Jr., Luis Antonio de and Rauber, David and Papa, Jo{\~a}o Paulo and Palm, Christoph}, title = {Semi-supervised Segmentation Based on Error-Correcting Supervision}, series = {Computer vision - ECCV 2020: 16th European conference, Glasgow, UK, August 23-28, 2020, Proceedings, Part XXIX}, booktitle = {Computer vision - ECCV 2020: 16th European conference, Glasgow, UK, August 23-28, 2020, Proceedings, Part XXIX}, publisher = {Springer}, address = {Cham}, isbn = {978-3-030-58525-9}, doi = {10.1007/978-3-030-58526-6_9}, pages = {141 -- 157}, abstract = {Pixel-level classification is an essential part of computer vision. For learning from labeled data, many powerful deep learning models have been developed recently. In this work, we augment such supervised segmentation models by allowing them to learn from unlabeled data. Our semi-supervised approach, termed Error-Correcting Supervision, leverages a collaborative strategy. Apart from the supervised training on the labeled data, the segmentation network is judged by an additional network. The secondary correction network learns on the labeled data to optimally spot correct predictions, as well as to amend incorrect ones. As auxiliary regularization term, the corrector directly influences the supervised training of the segmentation network. On unlabeled data, the output of the correction network is essential to create a proxy for the unknown truth. The corrector's output is combined with the segmentation network's prediction to form the new target. We propose a loss function that incorporates both the pseudo-labels as well as the predictive certainty of the correction network. Our approach can easily be added to supervised segmentation models. We show consistent improvements over a supervised baseline on experiments on both the Pascal VOC 2012 and the Cityscapes datasets with varying amounts of labeled data.}, subject = {Semi-Supervised Learning}, language = {en} } @article{GrassmannMengelkampBrandletal., author = {Graßmann, Felix and Mengelkamp, Judith and Brandl, Caroline and Harsch, Sebastian and Zimmermann, Martina E. and Linkohr, Birgit and Peters, Annette and Heid, Iris M. and Palm, Christoph and Weber, Bernhard H. F.}, title = {A Deep Learning Algorithm for Prediction of Age-Related Eye Disease Study Severity Scale for Age-Related Macular Degeneration from Color Fundus Photography}, series = {Ophtalmology}, volume = {125}, journal = {Ophtalmology}, number = {9}, publisher = {Elsevier}, doi = {10.1016/j.ophtha.2018.02.037}, pages = {1410 -- 1420}, abstract = {Purpose Age-related macular degeneration (AMD) is a common threat to vision. While classification of disease stages is critical to understanding disease risk and progression, several systems based on color fundus photographs are known. Most of these require in-depth and time-consuming analysis of fundus images. Herein, we present an automated computer-based classification algorithm. Design Algorithm development for AMD classification based on a large collection of color fundus images. Validation is performed on a cross-sectional, population-based study. Participants. We included 120 656 manually graded color fundus images from 3654 Age-Related Eye Disease Study (AREDS) participants. AREDS participants were >55 years of age, and non-AMD sight-threatening diseases were excluded at recruitment. In addition, performance of our algorithm was evaluated in 5555 fundus images from the population-based Kooperative Gesundheitsforschung in der Region Augsburg (KORA; Cooperative Health Research in the Region of Augsburg) study. Methods. We defined 13 classes (9 AREDS steps, 3 late AMD stages, and 1 for ungradable images) and trained several convolution deep learning architectures. An ensemble of network architectures improved prediction accuracy. An independent dataset was used to evaluate the performance of our algorithm in a population-based study. Main Outcome Measures. κ Statistics and accuracy to evaluate the concordance between predicted and expert human grader classification. Results. A network ensemble of 6 different neural net architectures predicted the 13 classes in the AREDS test set with a quadratic weighted κ of 92\% (95\% confidence interval, 89\%-92\%) and an overall accuracy of 63.3\%. In the independent KORA dataset, images wrongly classified as AMD were mainly the result of a macular reflex observed in young individuals. By restricting the KORA analysis to individuals >55 years of age and prior exclusion of other retinopathies, the weighted and unweighted κ increased to 50\% and 63\%, respectively. Importantly, the algorithm detected 84.2\% of all fundus images with definite signs of early or late AMD. Overall, 94.3\% of healthy fundus images were classified correctly. Conclusions Our deep learning algoritm revealed a weighted κ outperforming human graders in the AREDS study and is suitable to classify AMD fundus images in other datasets using individuals >55 years of age.}, subject = {Senile Makuladegeneration}, language = {en} } @inproceedings{HoengWiedererEderetal., author = {H{\"o}ng, Simon K. and Wiederer, Jonas and Eder, Friedrich and Obergriesser, Mathias and Linner, Thomas}, title = {Towards AI-enhanced facade planning : integrating human expertise with machine learning-driven parametric modeling}, series = {EC³ \& CIB W78 : 2025 European Conference on Computing in Construction \& 42nd CIB W78 IT in Construction Conference}, volume = {6}, booktitle = {EC³ \& CIB W78 : 2025 European Conference on Computing in Construction \& 42nd CIB W78 IT in Construction Conference}, publisher = {European Council for Computing in Construction}, isbn = {978-9-083451-31-2}, issn = {2684-1150}, doi = {10.35490/EC3.2025.320}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-84862}, pages = {192 -- 199}, abstract = {Planning modern facade systems is complex, requiring optimization across multiple domains.This paper proposes an AI-enhanced workflow for facade planning, harnessing computer vision and human input via a Large Language Model.A generative AI system then guides a parametric model to produce 3D facade designs. Automated checks provide feedback to a Reinforcement Learning system, to iteratively determine optimal solutions.These solutions are verified and finalized by human expertise, ensuring improved outcomes with reduce planning time and effort.The approach illustrates how combining advanced AI methods with human expertise can address the multifactorial challenges of facade design within current industry practices.}, language = {en} } @article{HantschWestnerMatschi, author = {Hantsch, Carolin and Westner, Markus and Matschi, Markus}, title = {IT/OT Convergence: Status Quo and Implementation Recommendations}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:898-opus4-77586}, pages = {17}, abstract = {This study examines IT/OT Convergence in 31 predominantly large German companies. It presents an "IT/OT Governance Framework" as a specific model for successful convergence implementation. The in-terviews revealed that companies are at very different stages of maturity in implementing IT/OT conver-gence. The analysis reveals a heterogeneous level of convergence maturity across the interviewed compa-nies, the majority of which are at the beginning or in the implementation phase. Close cooperation between IT and OT is favoured, but without IT completely taking over OT. An IT/OT tandem consisting of a CIO and a central OT responsible person is seen as the ideal way to take responsibility. Six specific recommen-dations were derived from the interviews and included in the IT/OT Governance Framework developed. These include ensuring top management support, using security as a driver, introducing a central OT re-sponsible person, bringing IT and OT teams closer together, continuously demonstrating the added value of convergence, and taking a step-by-step approach to implementation}, language = {en} } @inproceedings{HoOoiWestner, author = {Ho, Joe Ee and Ooi, Boon Yaik and Westner, Markus}, title = {Application Integration Framework for Large Language Models}, series = {2024 5th International Conference on Artificial Intelligence and Data Sciences (AiDAS), Bangkok, Thailand, 03-04 September 2024}, booktitle = {2024 5th International Conference on Artificial Intelligence and Data Sciences (AiDAS), Bangkok, Thailand, 03-04 September 2024}, publisher = {IEEE}, isbn = {979-8-3315-2855-3}, doi = {10.1109/AiDAS63860.2024.10730541}, pages = {1 -- 6}, abstract = {Large Language Models (LLMs) have unlocked new opportunities in processing non-structured information. However, integrating LLM into conventional applications poses challenges due to their non-deterministic nature. This paper introduces a framework designed to effectively integrate LLM into intermediate modules by ensuring more consistent and reliable outputs. The framework includes three key components: the Sieve, which captures and retries processing of incorrect outputs; the Circuit Breaker, which stops processing persistently incorrect outputs; and the Optimizer, which enhances processing efficiency by combining inputs into single prompts. Experimental results employing structured methodology demonstrate the framework's effectiveness, achieving significant improvements a 71.05\% reduction in processing time and an 82.97\% reduction in token usage while maintaining high accuracy. The proposed framework, agnostic to specific LLM implementations, aids the integration of LLMs into diverse applications, enhancing automation and efficiency in fields such as finance, healthcare, and education.}, language = {en} } @inproceedings{PreissWestner, author = {Preiß, Niklas and Westner, Markus}, title = {Towards a Taxonomy for Digital Assistant Technologies: Addressing the Jingle-Jangle Fallacies}, series = {38th Bled eConference: empowering transformation: shaping digital futures for all: conference proceedings, 8.-11.6.2025, Bled}, booktitle = {38th Bled eConference: empowering transformation: shaping digital futures for all: conference proceedings, 8.-11.6.2025, Bled}, publisher = {University of Maribor Press}, isbn = {9789612869984}, doi = {10.18690/um.fov.4.2025.1}, pages = {1 -- 20}, abstract = {This study proposes a unified taxonomy for Digital Assistant Technologies (DATs) to resolve terminological inconsistencies and eliminate »Jingle-Jangle fallacies.« By employing a systematic taxonomy development method on 137 papers, the framework categorizes DATs across four meta-characteristics: AI technology, context, intelligence, and interaction. This taxonomy facilitates the clear differentiation of three primary DAT concepts: assistant, chatbot, and agent. By providing a structured framework, the study enhances conceptual clarity, fosters more focused research, and ensures better alignment of DATs.}, language = {en} } @misc{CapellmannWestner, author = {Capellmann, Dominik and Westner, Markus}, title = {ChatGPT in Software Engineering: Potentials, Challenges and Possible Applications for Unit Testing and Code Debugging}, series = {21st International Conference on Applied Computing 2024}, journal = {21st International Conference on Applied Computing 2024}, language = {en} } @incollection{StadlbauerFrommeldChowetal., author = {Stadlbauer, Jana and Frommeld, Debora and Chow, Rosan and Weber, Karsten}, title = {Partizipative Designpraktiken f{\"u}r verk{\"o}rperte K{\"u}nstliche Intelligenz}, series = {Designforschung und KI}, booktitle = {Designforschung und KI}, editor = {Grabbe, Lars C.}, publisher = {Springer}, address = {Wiesbaden}, isbn = {9783658497972}, issn = {2731-9458}, doi = {10.1007/978-3-658-49798-9_8}, pages = {123 -- 137}, abstract = {Der Beitrag thematisiert partizipative Designpraktiken f{\"u}r verk{\"o}rperte K{\"u}nstliche Intelligenz und untersucht, wie Nutzende st{\"a}rker in den Gestaltungsprozess eingebunden werden k{\"o}nnen. Ausgehend von Design-Thinking-Modellen, die Nutzendengruppen aktiv an der Technikgestaltung beteiligen, geht der Gedanke des Beitrags noch weiter und stellt das Konzept des participatory theatre vor - eine explorative Methode, die im Rahmen der Ausstellung ‚Leben mit Robotern' an der OTH Regensburg eingesetzt wurde. Besuchende konnten sich als Roboter verkleiden und in Rollenspielen Interaktionen mit Technik simulieren. Diese performative Herangehensweise er{\"o}ffnete neue Perspektiven auf das Design von Robotern und auf die Wahrnehmung verk{\"o}rperter KI. Mit einer qualitativen, inhaltsanalytischen Auswertung von knapp 90 Aussagen von Besuchenden wird untersucht, wie Menschen auf das Design von Robotik und KI reagiert haben und inwieweit alltagsrelevante Designprinzipien zur Sprache kommen. Die Ergebnisse zeigen eine mehrheitlich offene bis begeisterte Haltung gegen{\"u}ber Robotern. Kritik und Zur{\"u}ckhaltung wurde seltener ge{\"a}ußert. Es wird diskutiert, inwieweit zentrale Aspekte wie Design einen Einfluss auf die Wahrnehmung von Robotern haben k{\"o}nnen. Der Beitrag betont die gesellschaftliche Relevanz {\"a}sthetischer Gestaltung und pl{\"a}diert f{\"u}r interdisziplin{\"a}re, explorative Forschungsans{\"a}tze, um KI-Systeme nutzungsnah und kulturell sensibel zu entwickeln.}, language = {de} } @inproceedings{MaierHaugHuberetal., author = {Maier, Johannes and Haug, Sonja and Huber, Michaela and Katzky, Uwe and Neumann, Sabine and Perret, J{\´e}r{\^o}me and Prinzen, Martin and Weber, Karsten and Wittenberg, Thomas and W{\"o}hl, Rebecca and Scorna, Ulrike and Palm, Christoph}, title = {Development of a haptic and visual assisted training simulation concept for complex bone drilling in minimally invasive hand surgery}, series = {CARS Conference, 5.10.-7.10.2017}, booktitle = {CARS Conference, 5.10.-7.10.2017}, language = {en} } @misc{OPUS4-349, title = {Bildverarbeitung f{\"u}r die Medizin 2020}, editor = {Tolxdorff, Thomas and Deserno, Thomas M. and Handels, Heinz and Maier, Andreas and Maier-Hein, Klaus H. and Palm, Christoph}, publisher = {Springer Vieweg}, address = {Wiesbaden}, isbn = {978-3-658-29266-9}, doi = {10.1007/978-3-658-29267-6}, abstract = {In den letzten Jahren hat sich der Workshop "Bildverarbeitung f{\"u}r die Medizin" durch erfolgreiche Veranstaltungen etabliert. Ziel ist auch 2020 wieder die Darstellung aktueller Forschungsergebnisse und die Vertiefung der Gespr{\"a}che zwischen Wissenschaftlern, Industrie und Anwendern. Die Beitr{\"a}ge dieses Bandes - einige davon in englischer Sprache - umfassen alle Bereiche der medizinischen Bildverarbeitung, insbesondere Bildgebung und -akquisition, Maschinelles Lernen, Bildsegmentierung und Bildanalyse, Visualisierung und Animation, Zeitreihenanalyse, Computerunterst{\"u}tzte Diagnose, Biomechanische Modellierung, Validierung und Qualit{\"a}tssicherung, Bildverarbeitung in der Telemedizin u.v.m.}, subject = {Bildanalyse}, language = {de} }