The “Conformal Prediction” team of the DEEL project has been awarded the Best paper award at WAISE 2022 for their work in artificial intelligence on object detection with probabilistic guarantees. The authors of the paper are : Florence de Grancey Thales AVS France SAS, Jean-Luc Adam Renault, Lucian Alecu Continental Automotive France, Sébastien Gerchinovitz IRT Saint Exupéry – Institut de Mathématiques de Toulouse, Franck Mamalet IRT Saint Exupéry and David Vigouroux IRT Saint Exupéry.
The WAISE conference explores new ideas on safety engineering for AI-based systems, ethical design, regulation and standards for AI-based systems.
It was held in Munich on September 6, 2022 at SAFECOMP, the 41st International Conference on Safety, Reliability and Security of Computer Systems. It is an opportunity for AI researchers to exchange about their work and discoveries.
Object Detection With Probabilistic Guarantees: a Conformal Prediction Approach
Florence de Grancey – Thales AVS France SAS, Jean-Luc Adam – Renault, Lucian Alecu – Continental Automotive France, Sébastien Gerchinovitz – IRT Saint Exupéry – Institut de Mathématiques de Toulouse, Franck Mamalet – IRT Saint Exupéry et David Vigouroux – IRT Saint Exupéry
Providing reliable uncertainty quantification for complex visual tasks such as object detection is of utmost importance for safety-critical applications such as autonomous driving, tumor detection, etc. Conformal prediction methods offer simple yet practical means to build uncertainty estimations that come with probabilistic guarantees. In this paper we apply such methods to the task of object localization and illustrate our analysis on a pedestrian detection use-case. We highlight both theoretical and practical implications of our analysis.
The DEEL (DEpendable Explainable Learning) project involves academic and industrial partners in the development of reliable, robust, explainable and certifiable artificial intelligence technology bricks applied to critical systems.
The project covers 4 themes: reliability, explainability, certifiability and confidentiality by design.
The DEEL project is led by ANITI (Artificial and Natural Intelligence Toulouse Institute – Institut d’Intelligence Artificielle et Naturelle de Toulouse) and the IRT Saint Exupéry ; in Québec by IVADO (l’Institut de Valorisation des Données – Artificial Intelligence Research and Transfer Institute), the CRIAQ (Consortium de Recherche et d’Innovation en Aérospatiale au Québec – Consortium for Research and Innovation in Aerospace in Québec) and the IID, (Institut Intelligence et Données -Institute Intelligence and Data) at Université Laval.
In France, with the collaboration of major industrial partners in the aeronautics, space and automotive sectors such as Airbus, Thales, Continental, Renault, Scalian, SNCF, Apsys, DGA or Cnes – as well as academic partners and institutions that are also part of the DEEL project such as the Paul Sabatier Toulouse III University – UPS, the Toulouse Midi-Pyrénées Federal University, ISAE-SUPAERO, DGA and the CNES.
In Quebec, academic and industrial partners are also involved in the project. On the industry side: Thales Canada, CAE, Bombardier Aerospace and Bell Helicopter Textron Canada, and on the academic side: McGill University, Université Laval, Polytechnique Montréal, Université du Québec à Montréal and Université de Montréal.
The project is also supported by Aerospace Valley and the Toulouse Métropole Attractiveness Agency, and the French Ministry of Europe and Foreign Affairs.
This distinction at the WAISE 2022 conference affirms collaboration between academic and industrial partners as a key driver of innovation and technological progress and promises exciting prospects for the future of artificial intelligence.