1 - 6 of 6
-
PublicationHow to Indicate AI at Work on Vehicle Dashboards: Analysis and Empirical Study( 2024)The KARLI project aims to create an adaptive AI system for future vehicles. It’s focusing on motion sickness, level-compliant driver behavior, and AI-HMI (artificial intelligence human-machine Interface). The project explores making AI activities visible through avatars, aiming to enhance user experiences and empower users to understand and influence AI decisions for a positive interaction with technology. AI representations in HMIs range from non-representational to realistic, introducing a classification that includes "HMI-integrated." The analysis explores AI representations in vehicle HMIs, citing Nio's Nomi and Waymo's ride service as examples. AI depictions in films, ranging from abstract (HAL 9000) to realistic (Ava from "Ex Machina"), are examined. The KARLI project aims to differentiate itself by explicitly representing AI activity on screens in non-fictional and automotive contexts. Pros and cons of different levels of abstraction in AI avatars are made. A study predominantly involving females and younger individuals, showing a positive attitude toward AI was conducted. Three design variants of the avatar were tested in a comparative laboratory study. All tested designs received negative Net Promoter Scores, with the abstract figurative design rated the best and the figurative design the creepiest. All designs scored low on "Intention to Use," indicating participants’ reluctance, and "Product Loyalty" echoed this sentiment. A final design was created based on the results of analysis and study.
-
PublicationPattern Recognition. Introduction, Features, Classifiers and Principles(De Gruyter, 2024)The book offers a thorough introduction to Pattern Recognition aimed at master and advanced bachelor students of engineering and the natural sciences. Besides classification - the heart of Pattern Recognition - special emphasis is put on features: their typology, their properties and their systematic construction. Additionally, general principles that govern Pattern Recognition are illustrated and explained in a comprehensible way. Rather than presenting a complete overview over the rapidly evolving field, the book clarifies the concepts so that the reader can easily understand the underlying ideas and the rationale behind the methods. For this purpose, the mathematical treatment of Pattern Recognition is pushed so far that the mechanisms of action become clear and visible, but not farther. Therefore, not all derivations are driven into the last mathematical detail, as a mathematician would expect it. Ideas of proofs are presented instead of complete proofs. From the authors’ point of view, this concept allows to teach the essential ideas of Pattern Recognition with sufficient depth within a relatively lean book.
-
PublicationAI Marketplace: Serving Environment for AI Solutions Using Kubernetes( 2023)Recent advances in the field of artificial intelligence (AI) provide numerous potentials for industrial compa- nies. However, the adoption of AI in practice is still left behind. One of the main reasons is a lack of knowledge about possible AI application areas by industry experts. The AI Marketplace addresses this problem by pro- viding a platform for the cooperation between industry experts and AI developers. An essential function of this platform is a serving environment that allows AI developers to present their solution to industry experts. The solutions are packaged in a uniform way and made accessible to all platform members via the serving environment. In this paper, we present the conceptual design of this environment, its implementation using Amazon Web Services, and illustrate its application on two exemplary use cases.
-
PublicationDeep Learning-Based Action Detection for Continuous Quality Control in Interactive Assistance Systems( 2023)Interactive assistance systems have shown to be useful in various industrial settings, in particular those involving human labor like manual assembly of workpieces. Current systems support workers based on different technologies like projection-based augmented reality, hand or tool tracking or automated inspections using computer vision techniques. While these technologies help to increase product quality significantly, existing solutions are not able to monitor the entire process, which makes it difficult to detect process errors. In this paper, we present a deep-learning based approach for continuous on-the-fly quality control within an interactive assistance system. By using labeled video data of an assembly process, a model can be trained that automatically recognizes and distinguishes single actions and thus control the sequence of subsequent work processes. By integrating the system into the interactive assistance systems, users are made aware on any process errors. Besides presenting the concept and implementation of our deep-learning integration into the assistance system, we describe the created industrial assembly-oriented dataset and present the results from our technical evaluation that shows the potential of applying deep-learning methods into interactive assistance systems.
-
PublicationBest-of-Breed: Service-Oriented Integration of Artificial Intelligence in Interoperable Educational Ecosystems( 2023)
-
PublicationTowards Discriminative and Transferable One-Stage Few-Shot Object Detectors( 2022-10-11T20:58:25Z)Recent object detection models require large amounts of annotated data for training a new classes of objects. Few-shot object detection (FSOD) aims to address this problem by learning novel classes given only a few samples. While competitive results have been achieved using two-stage FSOD detectors, typically one-stage FSODs underperform compared to them. We make the observation that the large gap in performance between two-stage and one-stage FSODs are mainly due to their weak discriminability, which is explained by a small post-fusion receptive field and a small number of foreground samples in the loss function. To address these limitations, we propose the Few-shot RetinaNet (FSRN) that consists of: a multi-way support training strategy to augment the number of foreground samples for dense meta-detectors, an early multi-level feature fusion providing a wide receptive field that covers the whole anchor area and two augmentation techniques on query and source images to enhance transferability. Extensive experiments show that the proposed approach addresses the limitations and boosts both discriminability and transferability. FSRN is almost two times faster than two-stage FSODs while remaining competitive in accuracy, and it outperforms the state-of-the-art of one-stage meta-detectors and also some two-stage FSODs on the MS-COCO and PASCAL VOC benchmarks.
