DEep LEaring for Generating humAn moTION
Description: DELEGATION aims to propose a deep learning-based framework to generate 2D/3D skeleton sequences representing human motions with various expressiveness such as emotions, styles and morphological specificities. We propose to disentangle dynamics features from human motion to discover and learn adequate continuous latent sub-spaces. This allows us to control the generation process by selecting meaningful latent codes corresponding to the desired action as well as its style or emotion. By considering a hierarchy of motion primitives, the framework will be extended to automatically generate complex motion sequences like Activities of Daily Living (ADL). The findings will impact the efficiency of human motion analysis by offering a new way to rapidly increase the amount of data and to better understand the intrinsic features of human motion.
Funder: Agence Nationale de la Recherche
Amount: €181,200
Funding type: research
Funder scheme: AAPG2021
Start date: 2022-02-01
Investigator: Maxime DEVANNE
Institution: Institut de Recherche en Informatique Mathématiques Automatique Signal (IRIMAS) - UR 7499
Funded outputs count: 16
Funder award ID: ANR-21-CE23-0014
OpenAlex ID: G4629210510