http://hdl.handle.net/2268.2/6048 Tue, 15 Sep 2026 02:50:28 GMT 2026-09-15T02:50:28Z http://hdl.handle.net/2268.2/23373 Title: Automated Generation of Impactful Text Formatting Using Large Language Models Abstract: This master's thesis aims to develop an automated system for generating structured and visually impactful content for Zeiko, a Belgian start-up that offers a software for interactive document creation. The project addresses the need for a solution that helps Zeiko’s users transform long and unstructured textual content into concise, structured, and visually engaging formats compatible with Zeiko's software. The proposed solution uses large language models to analyze and summarize input text while preserving coherence and key elements. A central component of the system is a fine-tuned Mistral 7B model, trained specifically for structured JSON generation. This model extracts key information such as titles, summaries, and highlights from unstructured text, and returns a structured JSON designed for infographic templates. The fine-tuning process was managed using custom datasets and its effectiveness was validated through evaluations. The generated outputs are mapped into predefined infographic templates known as z-blocks, a format specific to the company. Additionally, a complementary management platform was developed to allow Zeiko’s design team to deploy new templates with associated metadata. The entire solution is deployed on Google Cloud using Docker and Kubernetes to ensure scalability and reliability. Evaluations demonstrate that the system significantly improves the speed and consistency of content creation. It also confirms that fine-tuning models like Mistral 7B is an effective method for extracting and structuring information from raw text, providing a practical tool to simplify document production. Sun, 29 Jun 2025 22:00:00 GMT http://hdl.handle.net/2268.2/23373 2025-06-29T22:00:00Z http://hdl.handle.net/2268.2/20386 Title: Development of the software architecture of a mobile robot Abstract: This master's thesis focuses on building a software infrastructure for a unique form of humanoid robot, developed within the RoboCup competition framework. The primary aim was to integrate a real-time operating system, such as Preempt-RT, Xenomai, or RTAI, to minimize latency and optimize system performance. An appropriate Linux distribution was also chosen as the software development base. Following a thorough evaluation based on various criteria, including latency, Preempt-RT was selected as the real-time operating system, accompanied by the Fedora distribution, chosen for its minimal size. The second part of this thesis concentrates on establishing efficient communication between electronic and software components. The chosen communication mode is USB (Universal Serial Bus), and results demonstrate satisfactory data transfer between these components. The third part involves implementing an event logging system, based on four selectable log levels by the user, enabling quick issue tracking and resolution during robot program execution. Recorded logs are detailed, including time down to the nanosecond, message level, content, as well as the relevant line and file. Users can select the base level, filtering logged messages into a file located in a specific directory. Furthermore, robust interfaces for task creation and inter-task communication have been designed to meet future needs. The primary goal was to optimize the efficiency and responsiveness of information exchange, thus contributing to the overall robot performance. The task-related interface facilitates the creation of tasks, assigning priorities and periods as needed, and, if required, assigning them to a designated processor core. The second interface enables tasks to communicate with each other: if a task requires another task's content, it can request and receive it once available. During this content wait, the task is paused to conserve processor resources. To conclude, this thesis has established a robust software infrastructure for a humanoid robot intended for the RoboCup competition. Through the selection of Preempt-RT and Fedora, system performance has been optimized. The use of USB for communication and the implementation of a logging system have also been validated. The interfaces created for task management and communication meet future needs for efficiency and responsiveness.; Cette thèse de master se concentre sur la construction d'une infrastructure logicielle pour une forme original de robot humanoïde, développée dans le cadre de la compétition RoboCup. L'objectif principal était d'intégrer un système d'exploitation temps réel, tel que Preempt-RT, Xenomai ou RTAI, afin de minimiser la latence et d'optimiser les performances du système. Une distribution Linux appropriée a également été choisie comme base de développement logiciel. Suite à une évaluation approfondie basée sur divers critères, dont la latence, Preempt-RT a été sélectionné comme système d'exploitation temps réel, accompagné de la distribution Fedora, choisie pour sa taille minimale. La deuxième partie de cette thèse se concentre sur l'établissement d'une communication efficace entre les composants électroniques et logiciels. Le mode de communication choisi est l'USB (Universal Serial Bus), et les résultats montrent un transfert de données satisfaisant entre ces composants. La troisième partie implique la mise en place d'un système de journalisation d'événe-ments, basé sur quatre niveaux de journal sélectionnables par l'utilisateur, permettant un suivi et une résolution rapides des problèmes lors de l'exécution du programme du robot. Les journaux enregistrés sont détaillés, incluant l'heure jusqu'à la nanoseconde, le niveau du message, son contenu, ainsi que la ligne et le fichier pertinents. Les utilisateurs peuvent sélectionner le niveau de base, filtrant ainsi les messages enregistrés dans un fichier situé dans un répertoire spécifique. De plus, des interfaces robustes ont été conçues pour la création et la communication des tâches, afin de répondre aux besoins futurs. Elles visent principalement à optimiser l'efficacité et la réactivité des échanges d'informations, contribuant ainsi à la performance globale du robot. L'interface de gestion des tâches facilite leur création, permettant d'attribuer des priorités et des périodes si nécessaire, ainsi que de les affecter à des cœurs de processeur spécifiques. La seconde interface permet aux tâches d'interagir entre elles : lorsqu'une tâche requiert le contenu d'une autre, elle peut en faire la demande et le recevoir dès qu'il est disponible. Durant cette attente, la tâche est mise en pause pour économiser les ressources du processeur. En conclusion, cette thèse a développé une infrastructure logicielle solide pour un robot humanoïde compétitif. La sélection de Preempt-RT et de Fedora a amélioré les performances du système, tandis que l'utilisation de l'USB et la mise en place d'un système de journalisation ont été confirmées. Les interfaces conçues pour la gestion des tâches et la communication anticipent efficacement les futurs besoins en termes d'efficacité et de réactivité. Sun, 23 Jun 2024 22:00:00 GMT http://hdl.handle.net/2268.2/20386 2024-06-23T22:00:00Z http://hdl.handle.net/2268.2/14575 Title: Master thesis : Person re-identification in a camera network Abstract: In the modern world, data collection is increasingly used to gain insights into our lifestyles. Companies are tracking us in every way imaginable to fully understand our needs. Smart cameras are one method that can be used to capture new information about customer behavior. In this thesis, we provide an implemented approach to detect, track, and re-identify people using live streams from non-overlapping surveillance cameras. To do so, we combine the well-known computer vision tasks of detection and tracking. In addition, we include a re-identification principle that allows a person to be re-identified as they move from one camera to another. In other words, we allow the assignment of a unique identifier to each tracked person and the retrieval of a person’s identifier when they have not been observed by a camera for a certain period of time. In our case, person re-identification relies heavily on a fairly recent deep learning model, the OSNet-AIN model. It provides essential information about the appearance of different people. These data obtained allow us to differentiate one person from another through the different cameras. However, other information such as the movement of people is also taken into account to improve this re-identification. Evaluated on videos provided by ARHS Spkiseed, the implemented method realizes only one error during re-identification. We also obtain frame rates between two and six taking into account that no GPU was used. The algorithm can therefore work in real time. Sun, 26 Jun 2022 22:00:00 GMT http://hdl.handle.net/2268.2/14575 2022-06-26T22:00:00Z http://hdl.handle.net/2268.2/7812 Title: Individuals Profiling: Deciphering the role of epigenetic methylation and gene expression in multiple cancer types Abstract: With the evolution of technology and the increasing availability of multi-omics cancer datasets, new opportunities are rising for data integration in order to get a better understanding of cancer by developing methods that allow the integration and extraction of data from those large databases. In this thesis, we analyzed gene expression and DNA methylation data across Pancreatic ductal adenocarcinoma (PDAC) and Lung Adenocarcinoma(LUAD) in order to understand the role of these two processes in the cancerous behavior of tumor tissues. Sun, 08 Sep 2019 22:00:00 GMT http://hdl.handle.net/2268.2/7812 2019-09-08T22:00:00Z