Repository der Technischen Hochschule Ingolstadt
Not a member yet
4170 research outputs found
Sort by
"Do you trust me?": Increasing User-Trust by Integrating Virtual Agents in Explainable AI Interaction Design
No full textBenchmarking Perturbation-Based Saliency Maps for Explaining Atari Agents
Get PDFOne of the most prominent methods for explaining the behavior of Deep Reinforcement Learning (DRL) agents is the generation of saliency maps that show how much each pixel attributed to the agents' decision. However, there is no work that computationally evaluates and compares the fidelity of different perturbation-based saliency map approaches specifically for DRL agents. It is particularly challenging to computationally evaluate saliency maps for DRL agents since their decisions are part of an overarching policy, which includes long-term decision making. For instance, the output neurons of value-based DRL algorithms encode both the value of the current state as well as the expected future reward after doing each action in this state. This ambiguity should be considered when evaluating saliency maps for such agents. In this paper, we compare five popular perturbation-based approaches to create saliency maps for DRL agents trained on four different Atari 2,600 games. The approaches are compared using two computational metrics: dependence on the learned parameters of the underlying deep Q-network of the agents (sanity checks) and fidelity to the agents' reasoning (input degradation). During the sanity checks, we found that a popular noise-based saliency map approach for DRL agents shows little dependence on the parameters of the output layer. We demonstrate that this can be fixed by tweaking the algorithm such that it focuses on specific actions instead of the general entropy within the output values. For fidelity, we identify two main factors that influence which saliency map approach should be chosen in which situation. Particular to value-based DRL agents, we show that analyzing the agents' choice of action requires different saliency map approaches than analyzing the agents' state value estimation
Relevance-Based Data Masking: A Model-Agnostic Transfer Learning Approach for Facial Expression Recognition
Get PDFDeep learning approaches are now a popular choice in the field of automatic emotion recognition (AER) across various modalities. Due to the high costs of manually labeling human emotions however, the amount of available training data is relatively scarce in comparison to other tasks. To facilitate the learning process and reduce the necessary amount of training-data, modern approaches therefore often rely on leveraging knowledge from models that have already been trained on related tasks where data is available abundantly. In this work we introduce a novel approach to transfer learning, which addresses two shortcomings of traditional methods: The (partial) inheritance of the original models structure and the restriction to other neural network models as an input source. To this end we identify the parts in the input that have been relevant for the decision of the model we want to transfer knowledge from, and directly encode those relevant regions in the data on which we train our new model. To validate our approach we performed experiments on well-established datasets for the task of automatic facial expression recognition. The results of those experiments are suggesting that our approach helps to accelerate the learning process
Industrial Image Grouping Through Pre-Trained CNN Encoder-Based Feature Extraction and Sub-Clustering
Get PDFA common challenge faced by many industries today is the classification of unlabeled image data from production processes into meaningful groups or patterns for better documentation and analysis. This paper presents a sequential approach for leveraging industrial image data to identify patterns in products or processes for plant floor operators. The dataset used is sourced from steel production, and the model architecture integrates feature reduction through convolutional neural networks (CNNs) like VGG, EfficientNet, and ResNet, followed by clustering algorithms to assign appropriate labels to the observed data. The model’s selection criteria combine clustering metrics, including entropy minimization and silhouette score maximization. Once primary clusters are identified, sub-clustering is performed using near-labels, which are pre-assigned to images with initial distinctions. A novel metric, C-Score, is introduced to assess cluster convergence and grouping accuracy. Experimental re sults demonstrate that this method can address challenges in detecting variations across images, improving pattern recognition and classification
Environment Setup and Model Benchmark of the MuFoRa Dataset
Get PDFAdverse meteorological conditions, particularly fog and rain, present significant challenges to computer vision algorithms and autonomous systems. This work presents MuFoRa a novel, controllable, and measured multimodal dataset recorded at CARISSMA’s indoor test facility, specifically designed to assess perceptual difficulties in foggy and rainy environments. The dataset bridges research gap in the public benchmarking datasets, where quantifiable weather parameters are lacking. The proposed dataset comprises synchronized data from two sensor modalities: RGB stereo cameras and LiDAR sensors, captured under varying intensities of fog and rain. The dataset incorporates synchronized meteorological annotations, such as visibility through fog and precipitation levels of rain, and the study contributes a detailed explanation of the diverse weather effects observed during data collection in the methods section. The dataset’s utility is demonstrated through a baseline evaluation example, asse ssing the performance degradation of state-of-the-art YOLO11 and DETR 2D object detection algorithms under controlled and quantifiable adverse weather conditions. The public release of the dataset (https://doi.org/10.5281/zenodo.14175611) facilitates various benchmarking and quantitative assessments of advanced multimodal computer vision and deep learning models under the challenging conditions of fog and rain
Shifting expectations – how the responsibility for the habitability of the planet is shifting
No full textThe decision-point-dilemma: Yet another problem of responsibility in human-AI interaction
No full textAI as decision support supposedly helps human agents make ‘better’ decisions more efficiently. However, research shows that it can, sometimes greatly, influence the decisions of its human users. While there has been a fair amount of research on intended AI influence, there seem to be great gaps within both theoretical and practical studies concerning unintended AI influence. In this paper I aim to address some of these gaps, and hope to shed some light on the ethical and moral concerns that arise with unintended AI influence. I argue that unintended AI influence has important implications for the way we perceive and evaluate human-AI interaction. To make this point approachable from both the theoretical and practical side, and to avoid anthropocentrically-laden ambiguities, I introduce the notion of decision points. Based on this, the main argument of this paper will be presented in two consecutive steps: i) unintended AI influence doesn’t allow for an appropriate determination of decision points - this will be introduced as decision-point-dilemma, and ii) this has important implications for the ascription of responsibility
Comparison of Localization Algorithms between Reduced-Scale and Real-Sized Vehicles Using Visual and Inertial Sensors
No full textSeebeck Coefficient Modification via Extreme High-Speed Laser Material Deposition for Tool Materials
No full textDuring blanking and cold forming of metals, thermoelectricity almost always occurs. While thermoelectric voltages can be used for temperature measurement via a tool-workpiece thermocouple, currents significantly influence adhesion formation. In both cases, the thermoelectric behavior of tool and workpiece materials, characterized by the Seebeck coefficient, plays a decisive role. While a large difference in coefficients increases the accuracy of temperature measurement, similar coefficients improve wear behavior. Currently, there is no method to adjust the Seebeck coefficient of materials without experimental procedures, which restricts the selection of tool materials based on their Seebeck coefficient. This study presents a novel approach for adjusting the Seebeck coefficient of tool steels using tailored coatings applied by extreme high-speed laser material deposition (EHLA). Therefore, an analysis was conducted to investigate the effects of chemical composition, substrate material and its heat treatment on the thermoelectric and mechanical behavior of the coating. The results demonstrate that targeted modification via tailored EHLA coatings is possible