Maxime Devanne
Generative approaches to overcome the lack of data in human movement analysis.

Recent advances in generative artificial intelligence are opening new perspectives for the modeling, synthesis, and analysis of human movement. This presentation will address the adaptation of generative models to the representation of human movement in the form of multivariate time series. Applications in rehabilitation movement assessment and gait analysis will be presented, illustrating how data augmentation, transfer learning, and generative modeling approaches can help overcome the lack of available data. Particular attention will also be given to the evaluation of these methods and the synthetic data they produce.

Maxime Devanne is a Lecturer in Computer Science at the University of Haute-Alsace and a member of the MSD research team within the IRIMAS laboratory.
His main research interests focus on time series analysis and human movement using deep learning methods. He is the scientific lead of the ANR JCJC DELEGATION project (2022–2026), dedicated to the study of deep generative models for human movement. He also coordinated the REVEIL (2022–2023) and DeepRehab (2020–2021) projects, which focused on the analysis and modeling of rehabilitation movements.