Nicolás Violante Grezzi

I'm currently a postdoc at Imagine, École Nationale des Ponts et Chaussées, working with Vincent Lepetit. I completed my Ph.D. at Inria Sophia Antipolis with the GraphDeco group, supervised by George Drettakis.

I earned my master's degree in Mathematics, Vision, and Learning from ENS Paris-Saclay, graduating with highest honors (mention très bien). I also earned a degree in Electrical Engineering from Universidad de la República in Uruguay.

Email  /  CV  /  Scholar  /  LinkedIn  /  Bluesky  /  GitHub

profile photo

News

02/2026

I started a postdoc at Imagine, École Nationale des Ponts et Chaussées, working with Vincent Lepetit.

06/2025

Invited talk at Centre Borelli.

05/2025

Our paper Splat and Replace: 3D Reconstruction with Repetitive Elements was accepted at SIGGRAPH 2025.

09/2024

I started an internship at Adobe, San Francisco, under the supervision of Thibault Groueix.

Research

Articulation in Prime: Primitive-Based Articulated Object Understanding from a Single Casual Video
A. Artykov, T. Ravaud, N. Violante, V. Lepetit
Preprint, 2026
project page

A category-agnostic optimization framework that recovers 3D kinematics of articulated objects from monocular video by fitting geometric primitives organized via joint constraints.

Lighting-Consistent Object Transfer Across Radiance Fields
N. Violante, G. Kopanas, L. Franke, J. Philip, G. Drettakis
Computer Graphics Forum (EGSR), 2026
project page / paper / code / video

A diffusion-based harmonization approach that transfers objects between 3DGS scenes by correcting the illumination of the objects and adding matching shadows and reflections.

Splat and Replace: 3D Reconstruction with Repetitive Elements
N. Violante, A. Meuleman, A. Gauthier, F. Durand, T. Groueix, G. Drettakis
SIGGRAPH, 2025
project page / paper / code / video

We leverage repetitive elements in 3D scenes — segmenting, registering, and sharing information across repeated instances in a 3DGS reconstruction — to improve novel view synthesis quality under poor coverage and occlusions.

Physically-based Lighting of 3D Generative Models of Cars
N. Violante, A. Gauthier, S. Diolatzis, T. Leimkühler, G. Drettakis
Computer Graphics Forum (Eurographics), 2024
project page / paper / code / video

A 3D GAN framework for multi-view coherent 360° car synthesis with disentangled reflectance, enabling interactive relighting with arbitrary environment maps.


Design inspired by Jon Barron's website.