Flanders AI Research Day

Program

Breakout sessions (I) (12:00-13:00)

LLMs in Flanders: Training, Generation and NLP Applications for Text and Speech

Description: Foundation models are powering a wide range of applications and research methods. Within the Flemish AI program, a lot of effort is dedicated to improving and understanding these models better. This session briefly introduce the foundations of LLMs and cover some of our research, for instance the creation of Dutch LLMs, state-of-the-art Dutch speech recognition,  controllable text generation, novel human-inspired training methods ,and LLMs for legal QA.

Moderator: Pieter Delobelle (KULeuven)

Speakers: Prof. Thomas Demeester (UGent), Ehsan Lotfi and Nicolae Banari (UAntwerpen), Ying Jiao (KU Leuven), Prof. Hugo Van Hamme (KU Leuven)

Advances in Explainable AI (deep dive I)

Note: (The explainable AI deep dive session consists of 2 parts, it is possible to follow both, or a single one)

Description: AI is advancing at a never seen before pace. It is not unthinkable that important decisions in a wide array of domains, including health, tax, insurance, will be made by machines operating autonomously in the near future. However, to get there we will not only need new AI methods, but we will also need to develop some core capabilities in order to employ these methods in responsible ways. In this deep dive session 3 researchers will present research that is being conducted in Flanders in order to address important topics such as: making sure that we can understand how AI methods operate and how we can make sure that they do not discriminate.

Speakers: Prof. David Martens (UAntwerpen), Prof. José Oramas (UAntwerpen), Prof. Nikos Deligiannis (VUB)

Advances in AI-based Knowledge Discovery

Description: Knowledge discovery from data is the endeavor to unearth useful and actionable insights from complex data. As we record ever more data, it is becoming an important tool for scientists and engineers in almost every discipline. However, data comes in many varieties, users can often not explain well what they are looking for, and knowledge discovery is not a clearly defined problem. This makes it one of the most challenging settings to develop machine learning solutions for. In this session, we showcase some of the advances made within the Flanders AI Research Program to facilitate visualization and understanding of data.

Speakers: Arne Gevaert (UGent); Ceder Dens (UAntwerpen); Edith Heiter & Fuyin Lai (UGent); Shakir Showkat Sofi (KU Leuven); and Sonny Achten (KU Leuven)

Moderator: Prof. Jefrey Lijffijt (UGent) 

Reinforcement Learning (RL) for Industrial Applications

Description: We would like to present the design, training and deployment of RL and show it applied in 2 industrial applications related to Smart Machines and Autonomous Systems use cases. 

Moderator: Bruno Depraetere (Flanders Make)

Speakers: Bruno Depraetere (Flanders Make), Denis Steckelmacher (VUB), Gaoyuan Liu (VUB), Tom Lefebvre (UGent)

Situated AI (GC2) advances and highlights

Description: The goal of Situated AI (GC2) is to build agents that are designed to behave successfully in their environment. Such agents: Exist in a dynamic (rapidly changing) environment; Can sense or perceive; Can manipulate or change through their actions; Should coordinate their actions to achieve common goals; Should be human-centered and trustworthy (safe, accountable, and reliable). In this deep dive session we will provide some research highlights of the results that have been achieved in the Situated AI grand challenge so far.

Uncovering Hidden Patterns: Advanced Anomaly Detection with Generative AI

Description: In this talk, we explore the power of Generative AI in detecting anomalies for industrial applications. Generative AI offers a dynamic approach by learning the underlying structure of data and uncovering hidden patterns that may signal anomalies.

First talk: Knowledge guided deep learning

Speaker: Prof. Peter Karsmakers (KULeuven)

Second talk: An Agentic Retrieval Augmented Generation approach for failure resolution in manufacturing processes (Mathias Verbeke)

Speaker: Prof. Mathias Verbeke (KULeuven)

Moderator: Dries Verhees (Flanders Make)

Breakout sessions (II) (14:30-15:30)

AI on the Edge

First talk: Generative AI Multi-Agent Systems at the Edge: Strategies for Inference, Workflow Optimization, and Hardware Implications

Description: This lecture will explore the challenges and opportunities presented by generative AI in multi-agent systems operating at the edge. We will dive into strategies for optimizing inference and fine-tuning agentic workflows directly on edge devices, addressing the constraints and advantages of decentralized processing. The session will also cover business cases for deploying generative multi-agent systems at the edge, from real-time decision-making to enhanced autonomy in IoT and edge computing environments. Additionally, we will examine the hardware implications for supporting these advanced workflows, focusing on how edge devices must adapt to handle the computational and memory demands of generative models. Finally, the lecture will touch on retrieval-augmented generation (RAG) workflows and the critical role of self-reflection and memory in building multi-agent systems that can learn, adapt, and improve autonomously, offering insights into the next generation of AI-powered edge solutions.

Speaker: Tanguy Coenen, imec

Second talk: FAIM Box v2 - a toolbox for AI researchers and system developers 

Description: Imagine LEGO Mindstorms for edge-AI-development: a multipurpose platform to test your ideas and to get your edge AI projects moving. At imec we are building this: the FAIM Box v2 is a hardware and software kit with a performant FPGA at the core. It allows you to connect sensors, AI-chips, cameras, an Open WiFi or a radar module, with your PC. It can also become your stand-alone engine for your next robot or intelligent assistant. We provide a compact module with high-speed interfaces and a software platform on which you can build and test your creations. If you want to know more, please join us for this session. 

Speaker: Maarten van der Burgt, imec

Neurosymbolic learning and reasoning for trustworthy AI

Description: The abilities to learn and to reason are central to (artificial) intelligence.The focus in AI today is very much on learning, but one should not learn what one already knows. The challenge therefore is to use the avaible knowledge to guide and constrain the learning, and to reason with the resulting models in a trustworthy manner. This requires the integration of different paradigms for AI, which is the focus of neurosymbolic AI, often viewed as the next wave in AI.  The talk will contrast traditional AI techniques (such as search and rule-based AI) with those of deep learning and large language models. I will argue that we need to combine the best of both worlds to arrive at trustworthy AI, especially  AI that  exploits knowledge,  provides genuine explanations and guarantees with regard to safety and robustness. I will also argue that Neurosymbolic AI = Logic + Probability + Neural Networks. This will allow me to specify a high-level recipe  for developing neurosymbolic AI approaches: start from a logic, add a probabilistic interpretation, and then turn neural networks into `neural predicates'. Probability is interpreted broadly here, and is necessary to provide a quantitative and differentiable component to the logic. At the semantic and the computation level, one can then combine logical circuits (ako proof structures) labeled with probability, and neural networks in computation graphs.

Speaker: Prof. Luc De Raedt (KULeuven)

AI for Images & Video

Moderator: Prof. Tinne Tuytelaars (KULeuven)

First talk: Animate Your Motion: Turning Still Images into Dynamic Videos

Description: In recent years, diffusion models have made remarkable strides in text-to-video generation, sparking a quest for enhanced control over video outputs to more accurately reflect user intentions. Traditional efforts predominantly focus on employing either semantic cues, like images or depth maps, or motion-based conditions, like moving sketches or object bounding boxes. Semantic inputs offer a rich scene context but lack detailed motion specificity; conversely, motion inputs provide precise trajectory information but miss the broader semantic narrative. For the first time, we integrate both semantic and motion cues within a diffusion model for video generation, as demonstrated in Fig 1. To this end, we introduce the Scene and Motion Conditional Diffusion (SMCD), a novel methodology for managing multimodal inputs. It incorporates a recognized motion conditioning module and investigates various approaches to integrate scene conditions, promoting synergy between different modalities. For model training, we separate the conditions for the two modalities, introducing a two-stage training pipeline. Experimental results demonstrate that our design significantly enhances video quality, motion precision, and semantic coherence.

Speaker: Bo Wan (KULeuven)

Second talk: Easily Accessible 3D Content - free viewpoint video met neural rendering 

Description: Our current technology allows us to easily view 2D images and videos on various devices for both work and entertainment. However, compared to 2D content, 3D content offers richer information and a more immersive experience. This naturally raises the question: can 3D content also be seamlessly integrated into our daily lives? Imagine a future where we can browse 3D immersive scenes as effortlessly as we scroll through TikTok or YouTube. One major challenge, however, is the significant storage requirements of 3D content. In this presentation, I will briefly discuss our recent work on compact 3D representations for novel view synthesis. Specifically, we propose temporal tri-plane radiance fields for efficient free-viewpoint video, enabling online streaming and photorealistic rendering of dynamic scenes. For static scenes, we introduce a multi-level implicit representation for 3D Gaussian splatting, which achieves high-quality rendering with a compression rate hundreds of times greater than the traditional methods.

Speaker: Minye Wu (KULeuven)

Third talk: Recognizing Tables from Industrial Documents: Challenges and Solutions based on Visual modeling

Description: This presentation will explore advanced techniques for improving table recognition in industrial documents, as part of the RAVLIDP Enrichment project. Industrial documents, such as invoices, manuals, and technical specification documents, often contain complex tables that are challenging to accurately extract using conventional methods. The presentation will outline the challenges and the latest advancements in table recognition. By visually comparing methods in a series of example documents, the challenges inherent in table recognition will be clearly illustrated. Once diverse layouts and complex structures are highlighted, the properties of tables will be analyzed to better understand their defining characteristics.

Speaker: Van Beersel Arthur (VUB)

Advances in Explainable AI (deep dive 2)

Note: The explainable AI deep dive session consists of 2 parts, it is possible to follow both, or a single one.

Description: AI is advancing at a never seen before pace. It is not unthinkable that important decisions in a wide array of domains, including health, tax, insurance, will be made by machines operating autonomously in the near future. However, to get there we will not only need new AI methods, but we will also need to develop some core capabilities in order to employ these methods in responsible ways. In this deep dive session 3 researchers will present research that is being conducted in Flanders in order to address important topics such as: making sure that we can understand how AI methods operate and how we can make sure that they do not discriminate."

Speakers: Prof. Katrien Verbert (KU Leuven), Prof. Toon Calders (UAntwerpen), Prof. Vincent Ginis (VUB), Thomas Doom (Uantwerp)

Introduction to Bayesian Optimization: Techniques, Applications, and Strategies

Description: For the sub-class of low-dimensional black-box optimization problems that are expensive to evaluate, Bayesian optimization (BO) has become a very popular alternative. BO has applications ranging from hyperparameter tuning of deep learning models and design optimization in engineering to stochastic optimization in operational research. Bayesian optimization builds a probabilistic surrogate model, usually a Gaussian process, based on all previous evaluations. Gaussian processes are not only able to predict the quality of new solutions, but also approximate the uncertainty around the prediction. This information is then used to decide what solution to evaluate next, explicitly trading off exploration (high uncertainty) and exploitation (high quality). This trade-off is modeled by the acquisition function which quantifies how ‘interesting’ a solution is to evaluate. The workshop will introduce the general BO framework for black-box optimization with and without noise; the most commonly used acquisition functions will be explained. Furthermore, we will cover multiobjective Bayesian optimization, with a particular focus on noise handling strategies, and give examples of practical applications such as simulation-optimization and hyperparameter optimization.

Moderators: Prof. Ivo Couckuyt, Prof. Tom Dhaene (imec, UGent)

Infosession - Flemish funded projects: bridging research & industry

Description: This session explores how various VLAIO instruments facilitate collaboration between research institutions and companies, i.e. O&O, Baekeland & innovation mandate, ICON, etc. Discover how these instruments support joint efforts in advancing innovation, bridging the gap between academic research and industrial application, and foster effective partnerships to drive progress in various sectors.

Privacy Aware Digital Assistant for Operators

Description: We would like to showcase how LLMs GPT based models can be tuned and customized to assist Operators in an Industrial environment while respecting their privacy

Moderators: Ewoud Verhelst & Bart Van Doninck (Flanders Make)

Exploring Efficiency in Multimodal Training (PEFT) and Learning by Teaching (LbT)

First talk: Introducing Routing Functions to Vision-Language Parameter-Efficient Fine-Tuning with Low-Rank Bottlenecks (multimodal, vision and language)

Description: Mainstream parameter-efficient fine-tuning (PEFT) methods, such as LoRA or Adapter, project a model’s hidden states to a lower dimension, allowing pre-trained models to adapt to new data through this low-rank bottleneck. However, PEFT tasks involving multiple modalities, like vision-language (VL) tasks, require not only adaptation to new data but also learning the relationship between different modalities. Targeting at VL PEFT tasks, we propose a family of operations, called routing functions, to enhance VL alignment in the low-rank bottlenecks. These feature routing functions adopt linear operations and do not introduce new trainable parameters. In-depth analyses are conducted to study their behavior. In various VL PEFT settings, the routing functions significantly improve performance of the original PEFT methods, achieving over 20% improvement on VQAv2 (RoBERTa-large+ViT-L/16) and 30% on COCO Captioning (GPT2-medium+ViT-L/16). Also when fine-tuning a pre-trained multimodal model such as CLIP-BART, we observe smaller but consistent improvements across a range of VL PEFT tasks. Our code is available at https://github.com/tingyu215/Routing_VLPEFT.

Speaker: Tingyu Qu (KULeuven)

Second talk: Can LLMs Learn by Teaching? A Preliminary Study

Description: Teaching to improve student models (e.g., knowledge distillation) is an extensively studied methodology in LLMs. However, for humans, teaching not only improves students but also improves teachers. We ask: Can LLMs also learn by teaching (LbT)? If yes, we can potentially unlock the possibility of continuously advancing the models without solely relying on human-produced data or stronger models. In this paper, we provide a preliminary exploration of this ambitious agenda. We show that LbT ideas can be incorporated into existing LLM training/prompting pipelines and provide noticeable improvements. Specifically, we design three methods, each mimicking one of the three levels of LbT in humans: observing students’ feedback, learning from the feedback, and learning iteratively, with the goals of improving answer accuracy without training and improving models’ inherent capability with fine-tuning. The findings are encouraging. For example, similar to LbT in human, we see that: (1) LbT can induce weak-to-strong generalization: strong models can improve themselves by teaching other weak models; (2) Diversity in students might help: teaching multiple students could be better than teaching one student or the teacher itself. We hope that this early promise can inspire future research on LbT and more broadly adopting the advanced techniques in education to improve LLMs.

Speaker: Zifu Wang (KULeuven)

Third talk: On Compatible Loss Functions for Semantic Segmentation

Speaker: Zifu Wang (KULeuven)

Fourth talk: Combining multiple learning tasks with multi-view kernel principal component analysis

Speaker: Sonny Achten (KULeuven)

Fifth talk: Highly preliminary title: Visualization and interpretation of decision making pathways for optimizing neural network architectures

Speaker: Christopher Patzanovsky (Uhasselt)

Moderator: Prof. Matthew Blaschko (KULeuven)

Breakout sessions (III) (16:00-17:00)

Overview of GC1 AI-driven Data Science

Description: In this seminar, we will elaborate on the research activities of the Grand Challenge on AI-Driven Data Science. We will present some key results from our work packages: AI-driven Data Engineering, Symbolic and Subsymbolic Learning and Reasoning, AI-assisted Data Exploration, Automation in Machine Learning, Understanding Images and Audio, Responsible AI, and Decision Support Systems.

AI for the living environment

First talk: Enhancing environmental modelling using AI & remote sensing

Description: Machine learning techniques have already been applied in the domains of environmental modelling and remote sensing for many years. Recently new AI techniques are being proposed and explored and challenge the current state of the art.This lecture will give an overview of the current practices and applications including: Time series modelling for water and air quality forecasting; Using an AI foundation model for airborne change detection; Effortless training and rapid classification of remote sensing data; Integration of AI based environmental models into digital twin environments. 

Speakers: Stijn Vranckx, Tanja Van Achteren (VITO)

Second talk: Bayesian optimal design for species distribution modelling

Description: In various spatio-temporal applications, the acquisition of data can be labour intensive or costly such that in practice one is interested in an efficient design for data collection. Although a rich literature is available on spatio-temporal survey designs, less is known on how to develop optimum designs for integrating field surveys with modern acquisition techniques such as citizen science. In this work we present a Bayesian design for species distribution modelling. The approach seeks to improve designs for occupancy modelling via the integration of presence-only and presence-absence data within the design optimization. For this purpose, the prior knowledge of the presence-only data is captured through a Poisson process model and used to inform about the sampling location for a site-occupancy model. Results show that in this way surveys for presence-absence data can be made more efficient while improving inference. Future research aims to extend these techniques to other spatio-temporal applications.

Speaker: Max Savery (UGent)

Third talk: Object-centric representation learning with applications in remote sensing

Description: Visual representation learning aims to learn powerful and general visual foundation models capable of extracting informative representations from any input image. Two dominant ways to represent an image are global (one vector per image) and local/patch-wise (one vector per image patch) representations. Humans, on the other hand, represent complex scenes in terms of objects rather than patches. To be more in line with humans, a field of object-centric learning has emerged in the recent years. While still in its infancy, it gives a promise of foundation models capable of discovering objects in a completely unsupervised manner and outputting a single informative representation per object. This presentation will give an overview of the current state of the field as well as introduce our recent contributions on the topic. Finally, we will touch upon remote sensing as an interesting application domain for object-centric models. We will introduce the unique challenges of remote sensing and present a possible path towards tackling these challenges within the context of the FAIR2 project.

Speaker: Nikola Đukić (KULeuven)

AI for Healthier Aging: Concrete innovations in MS Diagnosis and Elder Care (30 min, from 16:00 to 16:30)

Description: Robust Interpretation of Machine-learning Extracted features from OCT data. Using artificial intelligence (AI) to improve the diagnosis and prognosis of Multiple Sclerosis (MS). By analyzing Optical Coherence Tomography (OCT) scans of the optic nerve, AI models can supplement neurologists' assessments and identify key features related to MS progression. This research aims to enhance the accuracy and efficiency of MS diagnosis and treatment.

Hydration & Happiness (H2O): A rObot Water Butler to Improve Elderly Health.
This project aims to develop a socially acceptable robot assistant to help elderly residents in nursing homes stay hydrated. The robot will autonomously deliver water to residents in both the communal dining area and their private rooms.

Neurosymbolic learning and reasoning - Deep dive

First talk: Interpretable Neurosymbolic Concept Reasoning

Description: Deep learning methods are highly accurate, yet their opaque decision process prevents them from earning full human trust. Concept-based models aim to address this issue by learning tasks based on a set of human-understandable concepts. However, state-of-the-art concept-based models rely on high-dimensional concept embedding representations which lack a clear semantic meaning, thus questioning the interpretability of their decision process. To overcome this limitation, we propose the Deep Concept Reasoner (DCR), the first interpretable concept-based model that builds upon concept embeddings. In DCR, neural networks do not make task predictions directly, but they build syntactic rule structures using concept embeddings. DCR then executes these rules on meaningful concept truth degrees to provide a final interpretable and semantically-consistent prediction in a differentiable manner.

Speaker: Prof. Giuseppe Marra (KULeuven)

Second talk: Knowledge guided deep learning

Description: Deep learning has achieved remarkable success across various applications in recent years. However, many of these applications tend to overlook the integration of available background knowledge, requiring the learning algorithm to relearn this information from representative data. To address such inefficiency, there has been a growing interest in approaches that combine domain knowledge with data when learning deep neural networks. This approach falls within the broader field of neuro-symbolic artificial intelligence. In this talk, I will begin by providing a brief overview of existing methods in this area of research. I will then focus specifically on an in-house framework called Constraint Guided Gradient Descent (CGGD), which leverages hard constraints to enhance the learning process of neural networks. The talk will conclude by discussing potential future directions for this line of research.

Speaker: Prof. Peter Karsmakers (KULeuven)

Third talk:Decision-Focused Learning: Foundations, State of the Art, Benchmark and Future Opportunities

Description: Increasingly, combinatorial optimisation problems follow a predict + optimize paradigm, where part of the parameters (costs, volumes, capacities) are predicted from data, and those predictions are fed into a downstream combinatorial optimisation problem. How to best train these predictive models? Decision-focused learning (DFL) is an emerging paradigm in machine learning which trains a model to optimize decisions, integrating prediction and optimization in an end-to-end system. This paradigm holds the promise to revolutionize decision-making in many real-world applications which operate under uncertainty, where the estimation of unknown parameters within these decision models often becomes a substantial roadblock to high-quality solutions. This talk presents a comprehensive review of DFL. It provides an in-depth analysis of the various techniques devised to integrate machine learning and optimization models, introduces a taxonomy of DFL methods distinguished by their unique characteristics, and conducts an extensive empirical evaluation of these methods proposing suitable benchmark dataset and tasks for DFL. Finally, we'll provides valuable insights into current and potential future avenues in DFL research.

Speaker: Prof. Tias Guns (KULeuven)

Fourth talk: Knowledge Graphs for Machine Learning

Speaker: Ioannis Dasoulas (KULeuven, FlandersMake)

Aleatoric and epistemic uncertainty in machine learning: a tutorial introduction

Description: Without any doubt, the notion of uncertainty is of major importance in machine learning and constitutes a key element of modern machine learning methodology. In recent years, it has gained in importance due to the increasing relevance of machine learning for practical applications, many of which are coming with safety requirements. In the statistics literature, two inherently different sources of uncertainty are commonly distinguished, referred to as aleatoric and epistemic. While the former refers to variability due to inherently random effects, the latter is uncertainty caused by a lack of knowledge and hence relates to the epistemic state of an agent. Thus, epistemic uncertainty can in principle be reduced on the basis of additional information, while aleatoric uncertainty is non-reducible. The distinction between different types of uncertainty and their quantification has also been adopted in the recent ML literature, and various methods for quantifying aleatoric and epistemic uncertainty have been proposed. This tutorial aims to give an overview of these methods, with specific focus on novel approaches for distinguishing and representing aleatoric and epistemic uncertainty. By the end of the tutorial, attendees will have a comprehensive understanding of the fundamental concepts and recent advances in this field.

Speaker: Prof. Willem Waegeman (UGent)

AI for E2E optimization in manufacturing systems

Description: We would like to present how AI models can be trained and deployed in a digital backbone of a manufacturing system / process. We will touch different topics such as parsing techniques to automatically scan datasheets of consumables, inline monitoring of assets under variable contexts using innovative AI techniques, Edge AI to meet real time requirements, etc.

Speakers: Prof. Bert Pluymers (KULeuven), Jeroen Jordens (Flanders Make), Willy Carlos (VUB), Sebastian Rojas Gonzalez (UHasselt) and Prof. Inneke Van Nieuwenhuyse (UHasselt)

Moderator: Jeroen Jordens (Flanders Make)

Sustainable AI: How to Avoid the Next AI Winter

Description: Is AI still going to be here in 20 years?" While the current hype cycle would suggest so, history warns us of the potential for another AI winter. The explosive growth of AI comes with a significant downside—its skyrocketing energy consumption. AI models are doubling in size by a factor of 100 every two years, while Moore’s Law can only support a modest factor of 2 in hardware advancements. This gap threatens the long-term sustainability of AI, posing a critical challenge for the future of innovation. At the same time, companies like NVIDIA are thriving thanks to specialized AI chips, but can this success last without addressing the growing energy demands? In Flanders, we are uniquely positioned to tackle this challenge, with our strong expertise in both hardware and software giving us a competitive edge. This keynote will explore how we can make AI sustainable by combining cutting-edge solutions in both areas. We will delve into the potential of neuromorphic computing at the edge and how breakthroughs in IMEC technology could make large language models up to 100 times more energy efficient. By driving innovation in these key areas, we can ensure that AI avoids the pitfalls of past winters and continues to thrive for decades to come.

Speaker: Prof. Steven Latré (imec)

AI for a Smarter Grid: Innovations in Data Analysis and Visualization (30 min, from 16:30 to 17:00)

Description: Develop advanced GIS visualization and analysis tools for LV grid digital twins to transform vast amounts of data into actionable information. Enhance the predictive clustering method for sampling electricity consumption and production timeseries data to improve grid congestion forecasting.

UC Demos: Discover the power of AI across different sectors

At the Flanders AI Research Day, you'll have the opportunity to explore groundbreaking AI solutions that are transforming various sectors. Discover how AI is revolutionizing industry, the environment, society, and healthcare through our Use Case (UC) demos.

Industry

• Smart Machines - Monitoring & Control

• Monitoring & Control of Production & Machine

• Production Optimization

• Straight-Through-Digitalization

• Autonomous Systems

• Human-Machine Flexible Production

• Robotics for Agrifood

• Prognostics and Health Management for Assets

Planet & Energy

• Living Environment

• Smart Grid

• Renewable Energy Production

Society

• Public Employment

• Digital Humanities

• Education & Training

• Collaborative Learning

• Conversational XAI

• Intelligent Intersections

Health

• Monitoring @ Home

• Real-World Evidence

• Medical Imaging

• Single Cell

• Digital Twin Cardio

• Intensive Care Units

• Personalized Dermatology

• Epidemic Decision Making

• Fitness on the Edge

• Sports

Come and discover how AI is transforming the world around us at the Flanders AI Research Day 2024

Location

Wintercircus Gent
Lammerstraat 13, 
9000 Gent