Fête de la Science 2025 in Juan-les-Pins

Explore the 3IA Côte d'Azur program for the Fête de la Science 2025 in Juan-les-Pins!

  • Dates: October 11-12, 2025
  • Times:
    • Saturday: 1:00 pm to 7:00 pm
    • Sunday: 10:00 am to 6:00 pm
  • Place: Palais des Congrès Antipolis (60 Chemin des Sables, Antibes)

Saturday program

1:00 pm to 4:00 pm
  • Demo: "Votre cœur virtuel : quand la science crée votre jumeau numérique" ("Your virtual heart: When science builds your digital twin") by Gaëtan Desrues (Researcher at Inria in Epione team, led by Nicholas Ayache - 3IA Scientific Director and Chairholder)
 
3:30 pm to 4:15 pm
  • Conference: "Reconnaissance de comportement humain à partir d'analyse vidéo" ("Human behavior recognition from video analysis") by François Bremond (3IA Chairholder)
4:00 pm to 7:00 pm
  • Demo: "Solving the Traveling Salesman Problem with Neural Networks" by Pierre Pereira (Ph.D. student under supervision of Emanuele Natale - 3IA Chairholder - and Frédéric Giroire - 3IA Fellow)

Abstract: The Traveling Salesman Problem is one of the most famous problem in combinatorial optimization. As such, it is a great benchmark for Deep Learning. This presentation will explain how we typically use neural networks to approximate the TSP, what are the current challenges, and what progress we can hope by tackling such problems.
 
  • Demo: "Understanding What Surgeons Do During Robotic Operations" by Ezem Sura Ekmekci (3IA Ph.D. student, Chair of Nicholas Ayache)

Robotic surgery has transformed modern medicine, allowing surgeons to perform complex procedures with enhanced precision. However, understanding exactly what happens during these operations—identifying each surgical action and when it occurs—remains a significant challenge. This research develops intelligent computer systems that can automatically recognize and detect surgical actions in robotic surgery videos. By teaching computers to "see" and understand what surgeons do in real-time, this work aims to improve surgical training, enhance patient safety, and enable better analysis of surgical techniques. Our approach uses advanced machine learning methods to identify specific actions like picking-up the needle, tying a knot, cutting a suture, creating a foundation for smarter operating rooms where technology can assist surgeons more effectively. This research bridges the gap between artificial intelligence and healthcare, ultimately working toward safer surgeries and better outcomes for patients.

Sunday program

10:00 am to 1:00 pm
  • Demo: "CleverFish : Analyse vidéo basée sur l'IA pour la biologie marine et le suivi des espèces" ("CleverFish : AI-based Video Analysis for Marine Biology and Species Monitoring") by Charles Bouveyron (3IA Chairholder)

The crucial need for reliable, robust, and unbiased biodiversity data to support initiatives such as the 30x30 initiative poses significant scientific and technological challenges. Advances have been made in automating fish biodiversity assessment using computer vision. However, the clear divide between the research domains of ecology and data science hinders the effective use of computer vision tools for ecological tasks.

CleverFish is a new tool designed to bridge the gap between marine biology and artificial intelligence. By addressing the following three challenges, it provides an effective analysis tool:
• Providing an easy-to-use graphical interface;
• Enabling both global and video-specific biodiversity assessment directly within the application;
• Offering the capability for rapid and efficient extraction of temporal and spatial distribution of fish species in a format understandable to ecologists.
 
  • Demo PEACE by Serena Villata (3IA Deputy Scientific Director and Chairholder)

PEACE - Providing Explanations and Analysis for Combating Hate Expressions

Ce démonstrateur décrit une méthode consistant à détecter les discours haineux explicites et implicites dans du texte, mais également à expliquer ces prédictions avec du langage naturel (en utilisant des modèles de langages comme GPT). 

This demonstrator describes a method for detecting both explicit and implicit hate speech in text, while also explaining these predictions using natural language (by leveraging language models such as GPT).
1:00 pm to 3:00 pm
  • Demo: "Représentation apprenable de maillages 3D" ("Learnable 3D mesh representation") by Pierre Alliez (3IA Chairholder)

Dans cette démo, nous présenterons PoNQ, une représentation de maillage innovante basée sur des points 3D enrichis de normales et de métriques d’erreur quadratique (QEM). Ces points sont prédits par un réseau neuronal, puis automatiquement connectés par des triangles via une triangulation de Delaunay, garantissant un maillage sans auto-intersections et toujours fermé. Nous montrons comment PoNQ permet de reconstruire des surfaces à partir de grilles SDF (champs de distance signés, avec des résultats qui surpassent les méthodes récentes, tant sur les critères de qualité de surface que sur les mesures de précision des arêtes.

In this demo, we will present PoNQ, an innovative mesh representation based on 3D points enriched with normals and quadratic error metrics (QEM). These points are predicted by a neural network, then automatically connected by triangles via Delaunay triangulation, ensuring a mesh without self-intersections that is always closed. We show how PoNQ enables surface reconstruction from SDF grids (signed distance fields), with results that outperform recent methods, both in terms of surface quality criteria and edge accuracy measures.
1:00 pm to 4:00 pm
4:00 pm to 6:00 pm

inHEART est un projet de startup du Centre Inria d'Université Côte d'Azur, du CHU de Bordeaux et de l’université de Bordeaux via l’Institut Hospitalo-Universitaire Liryc dont l’objectif est de valoriser plusieurs années de recherche et de développements de haut niveau dans le domaine du guidage des interventions sur les troubles du rythme cardiaque par l’imagerie médicale 3D.

inHEART is a startup project from the Inria Centre at Université Côte d'Azur, Bordeaux University Hospital, and the University of Bordeaux through the Liryc Hospital-University Institute. Its objective is to leverage several years of high-level research and development in the field of image-guided interventions for cardiac arrhythmias using 3D medical imaging.
 
3:00 pm to 6:00 pm
  • Demo: "Modèles statistiques pour l’organisation de données complexes" ("Statistical models for complex data organization") by Vincent Vandewalle (3IA Chairholder) 

Face à la diversité des informations que nous produisons – réponses à un sondage, habitudes d’achat, enregistrements médicaux ou encore courbes de températures – il n’est pas toujours facile d’y voir clair. Les méthodes de clustering offrent un moyen de donner du sens à cette complexité en regroupant les données selon leurs ressemblances. Qu’elles soient qualitatives (préférences, opinions), quantitatives (mesures numériques) ou même fonctionnelles (évolutions dans le temps), ces approches révèlent des groupes cachés au sein des données. Cette démo offrira un aperçu concret de leur fonctionnement et montrera comment elles transforment un flot de données brutes en catégories plus lisibles.

Given the diversity of information we produce – survey responses, purchasing habits, medical records, or temperature curves – it's not always easy to make sense of it all. Clustering methods offer a way to bring meaning to this complexity by grouping data according to their similarities. Whether qualitative (preferences, opinions), quantitative (numerical measurements), or even functional (changes over time), these approaches reveal hidden groups within the data. This demo will provide a concrete overview of how they work and show how they transform a flow of raw data into more readable categories.