Hello!

I am Benoît, a 4th year PhD student at EPFL’s CvLab, supervised by Prof. Pascal Fua.

My work is focused on finding good representations for 3D surface reconstruction and manipulation with neural networks.
I was a research intern at Microsoft Research in 2021 and Meta Reality Labs in 2022.


Publications

(full list on scholar)

DrapeNet: Garment Generation and Self-Supervised Draping

Ren Li*, Luca De Luigi*, Benoit Guillard, Mathieu Salzmann, Pascal Fua ; at CVPR 2023 ( * indicates equal contributions)

[Project Page] [Paper] [Code]

We use MeshUDF as a learned parameterization of garments, and learn to drape them without supervision - in the manner of SNUG, but with a single network for a whole category.

DIG: Draping Implicit Garment over the Human Body

Ren Li, Benoit Guillard, Edoardo Remelli, Pascal Fua ; at ACCV 2022 (Oral)

[Project Page] [Paper] [Code]

Reconstructing and draping garments over a parametric body model, with neural network trained to avoid collisions.

Learning to Simulate Realistic LiDARs

Benoit Guillard, Sai Vemprala, Jayesh K. Gupta, Ondrej Miksik, Vibhav Vineet, Pascal Fua, Ashish Kapoor ; at IROS 2022

[Blog Post] [Paper] [Video]

A neural network to simulate how objects interact with LiDARs given their appearance, deployed to a standard driving simulator.

MeshUDF: Fast and Differentiable Meshing of Unsigned Distance Field Networks

Benoit Guillard, Federico Stella, Pascal Fua ; at ECCV 2022

[Project Page] [Paper] [Code]

An extension of Marching Cubes to mesh non-watertight surfaces, applied the output of Unsigned Distance Field networks. We also derive gradients for the reconstructed vertex positions wrt. the UDF field.

Sketch2Mesh: Reconstructing and Editing 3D Shapes from Sketches

Benoit Guillard*, Edoardo Remelli*, Pierre Yvernay, Pascal Fua ; at ICCV 2021 ( * indicates equal contributions)

[Paper] [Code]

A deep learning based pipeline to reconstruct and locally edit 3D shapes from sketches.

UCLID-Net: Single View Reconstruction in Object Space

Benoit Guillard, Edoardo Remelli, Pascal Fua ; at NeurIPS 2020

[Paper] [Code]

A network architecture mixing 3D convolutions and local surface patches for reconstructing object from a single image.



Past, unpublished work

AnalogNet: Convolutional Neural Network Inference on Analog Focal Plane Sensor Processors

Matthew Z Wong*, Benoit Guillard*, Riku Murai, Sajad Saeedi, Paul Kelly ( * indicates equal contributions)

[arXiv] [Master thesis]

A low-level and super lightweight CNN implementation running directly on the analog light sensor of a special camera, SCAMP5.



Academic services

I was a reviewer for CVPR (2022 best reviewer award), ICCV, ECCV, ACCV, NeurIPS (2022 top reviewer), ICLR, ICML, SIGGRAPH, SIGGRAPH Asia, C&G, WACV, Eurographics.