Guénolé Fiche

I am a research scientist at NAVER LABS Europe, working on human-centric computer vision.

I received my PhD from CentraleSupélec in 2024, under the supervision of Renaud Séguier and Simon Leglaive. I investigated how latent representations learned by generative models can be leveraged by human pose and shape estimation models. During my PhD, I had the opportunity to visit IRI with Francesc Moreno-Noguer and Antonio Agudo, and to be advised by Xavier Alameda-Pineda from INRIA Grenoble.

Prior to that, I graduated in Mathematical and software engineering from INSA Rouen, with a major in "Mathematics for data science".

Email  /  CV  /  Twitter  /  Github

This work proposes a new approach to human mesh recovery (HMR) based on masked generative modeling. By tokenizing the human pose and shape, we formulate the HMR task as generating a sequence of discrete tokens conditioned on an input image.

This work introduces a novel paradigm to address the human pose and shape estimation problem, involving a low-dimensional discrete latent representation of the human mesh and framing human pose and shape estimation as a classification task.

We introduce a motion prior to capture the short-term dependencies of human motion and an unsupervised learned denoising method unifying regression- and optimization-based approaches in a single framework for real-time 3D human pose estimation.

We introduce SwimXYZ, a synthetic dataset of swimming motions and videos. SwimXYZ contains 3.4 million frames annotated with ground truth 2D and 3D joints, as well as 240 sequences of swimming motions in the SMPL parameters format.

Website source code borrowed from Jon Barron's public academic website.