Percepts et Concepts

Deep Learning Breakthrough in Computer Vision

Practical information
14 December 2021

ENS, salle des Résistants, 45 rue d'Ulm, 75005 Paris


Séance invitée, dans le cadre du séminaire de master "Percepts & Concepts" organisé par Paul Egré, avec le soutien du projet ANR AMBISENSE.

Abstract : The term AI in computer vision generally refers to deep learning, which is one of several supervised learning models. Deep learning represents a huge breakthrough in computer vision. While many tasks are concerned, we'll focus first on image classification. CNNs are the basic building blocks of deep learning applied to images. We will explain what CNNs are, how they work, how they're applied to image classification, and why this framework is different from approaches used in the past. We will then present other uses of deep learning in computer vision, as well as the limitations and challenges of these approaches.

NB: the session will take place online in case of any problem related to the evolution of Covid (keep posted for updates).