1. LLM in the loop: a framework for contextualizing counterfactual segment perturbations in point cloudsVeljka Kočić, Niko Lukač, Dzemail Rozajac, Stefan Schweng, Christoph Gollob, Arne Nothdurft, Karl Stampfer, Javier Del Ser, Andreas Holzinger, 2025, original scientific article Abstract: Point Cloud Data analysis has seen a major leap forward with the introduction of PointNet algorithms, revolutionizing how we process 3D environments. Yet, despite these advancements, key challenges remain, particularly in optimizing segment perturbations to influence model outcomes in a controlled and meaningful way. Traditional methods struggle to generate realistic and contextually appropriate perturbations, limiting their effectiveness in critical applications like autonomous systems and urban planning. This paper takes a bold step by integrating Large Language Models into the counterfactual reasoning process, unlocking a new level of automation and intelligence in segment perturbation. Our approach begins with semantic segmentation, after which LLMs intelligently select optimal replacement segments based on features such as class label, color, area, and height. By leveraging the reasoning capabilities of LLMs, we generate perturbations that are not only computationally efficient but also semantically meaningful. The proposed framework undergoes rigorous evaluation, combining human inspection of LLM-generated suggestions with quantitative analysis of semantic classification model performance across different LLM variants. By bridging the gap between geometric transformations and high-level semantic reasoning, this research redefines how we approach perturbation generation in Point Cloud Data analysis. The results pave the way for more interpretable, adaptable, and intelligent AI-driven solutions, bringing us closer to realworld applications where explainability and robustness are paramount.
Keywords: explainable AI, point cloud data, counterfactual reasoning, LiDAR, 3D point cloud data, interpretability, human-centered AI, large language models, K-nearest neighbors
Published in DKUM: 19.05.2025; Views: 0; Downloads: 2
 Full text (7,24 MB) |
2. LiDAR-Based Maintenance of a Safe Distance between a Human and a Robot ArmDavid Podgorelec, Suzana Uran, Andrej Nerat, Božidar Bratina, Sašo Pečnik, Marjan Dimec, Franc Žaberl, Borut Žalik, Riko Šafarič, 2023, original scientific article Abstract: This paper focuses on a comprehensive study of penal policy in Slovenia in the last 70 years, providing an analysis of statistical data on crime, conviction, and prison populations. After a sharp political and penal repression in the first years after World War II, penal and prison policy began paving the way to a unique "welfare sanction system", grounded in ideas of prisoners' treatment. After democratic reforms in the early 1990s, the criminal legislation became harsher, but Slovenia managed to avoid the general punitive trends characterized by the era of penal state and culture of control. The authoritarian socialist regime at its final stage had supported the humanization of the penal system, and this trend continued in the first years of the democratic reforms in the 1990s, but it lost its momentum after 2000. In the following two decades, Slovenia experienced a continuous harshening of criminal law and sanctions on the one hand and an increasing prison population rate on the other. From 2014 onwards, however, there was a decrease in all segments of penal statistics. The findings of the study emphasize the exceptionalism of Slovenian penal policy, characterized by penal moderation, which is the product of the specific local historical, political, economic, and normative developments.
Keywords: LIDAR, robot, human-robot collaboration, speed and separation monitoring, intelligent control system, geometric data registration, motion prediction
Published in DKUM: 16.02.2024; Views: 417; Downloads: 35
Full text (5,27 MB)
This document has many files! More... |
3. High-resolution spatiotemporal assessment of solar potential from remote sensing data using deep learningMitja Žalik, Domen Mongus, Niko Lukač, 2024, original scientific article Keywords: deep learning, fully convolutional neural network, LiDAR data, digital elevation model, solar energy, solar potential
Published in DKUM: 26.01.2024; Views: 244; Downloads: 116
Full text (6,42 MB)
This document has many files! More... |
