Guillaume Bourmaud

I am an Associate Professor at Bordeaux INP, with teaching activities at the engineering school ENSEIRB-MATMECA and research activities at the IMS Laboratory, within the Signal and Image group.

My research lies at the intersection of 3D computer vision and machine learning, with a current focus on learning 4D geometry from visual experience. I am particularly interested in how the three-dimensional structure of dynamic scenes, together with their evolution over time, can be recovered directly from raw 2D images, without relying on labeled data. More broadly, my work investigates how geometric representations of the visual world can be learned directly from visual data.

Over the years, I have worked on 3D foundation models and pre-training, self-supervised learning, deep generative modeling, learning-based feature and image matching, deep learning architectures, robust optimization, Gaussian processes, visual SLAM, 3D reconstruction and localization, as well as filtering and optimization on manifolds.

A selection of publications is available here, while the complete list can be found here.