A survey revisiting multi-view shape-from-shading, providing intuitive evidence for uniqueness of solutions based on epipolar geometry, and presenting SfS-NeuS, an effective neural implicit surface method that outperforms state-of-the-art 3D reconstruction approaches on real-world examples.

An extended framework that incorporates multi-view normal and reflectance maps into radiance-based surface reconstruction, achieving state-of-the-art performance on MVPS benchmark datasets with improved fine-grained detail reconstruction.

A comprehensive comparison of photogrammetry and photometric 3D reconstruction techniques for archaeological features, proposing a promising new method that combines both families in a multi-view, multi-lighting context.

A versatile paradigm for integrating multi-view reflectance and normal maps acquired through photometric stereo.

An extended multi-view stereo technique to reconstruct objects within a transparent, refractive medium by modeling the refractive interface's geometry and solving it using a discrete method validated on synthetic and real data.

A pipeline to recover precisely the geometry of a convex polyhedral object from multiple views under circular motion.

Supervised by Lilian Calvet and Jean-Denis Durou at IRIT.

Meshroom plugin implementing the open-source Open-RNb method for 3D surface reconstruction. Runs a two-phase albedo scaling pipeline and outputs reconstructed OBJ meshes in world coordinates from multi-view photometric stereo inputs.

Meshroom plugin integrating SDM-UniPS, a universal photometric stereo method that estimates surface normal maps, albedo, roughness, and metallic parameter maps from multi-light image sets.

Meshroom plugin for LINO-UniPS, a photometric stereo method leveraging diffusion model priors to estimate surface normals from multi-lighting images. Supports arbitrary numbers of input images with GPU acceleration.

Meshroom plugin for Uni-MS-PS, a universal photometric stereo method using a multi-scale encoder-decoder transformer architecture for high-resolution surface normal estimation under unknown lighting.

A hands-on tutorial with 7 progressively complex examples — from a minimal node to SfMData manipulation with pyalicevision and ML inference (DeepLabV3 semantic segmentation). Serves as the learning resource for Meshroom plugin development.

Computation for this work was performed using HPC resources from the Juliet cluster operated by MesoNet.