Clément Godard

I am a Senior Research Engineer at Google where I work on real-time 3D reconstruction and neural rendering. Before that I spent some time at Niantic and Skydio where I worked on depth estimation using deep neural networks. I did my PhD at UCL with Gabe Brostow where I worked on monuclar depth estimation.

I got a MSc in Computer Vision and Graphics from UCL and an MEng from Supélec. I also spent a summer at Google Seattle working on Jump and another summer at Facebook Seattle working on burst denoising using deep learning.

Email  /  CV  /  Google Scholar  /  GitHub  /  flickr  /  stuff


I am quite interested in the intersection of computer vision and graphics.
In particular I enjoy 3D reconstruction using neural networks, especially in the case of neural rendering.

SimpleRecon: 3D Reconstruction Without 3D Convolutions
Mohamed Sayed, John Gibson, Jamie Watson, Victor Prisacariu, Michael Firman, and Clément Godard
ECCV, 2022
project page / paper / video / code

We built a fast and accurate multiview depth estimator for monocular videos.

Camera Pose Estimation and Localization with Active Audio Sensing
Karren Yang, Michael Firman, Eric Brachmann, and Clément Godard
ECCV, 2022

We use audio sensing to improve the performance of visual localization methods on three tasks: relative pose estimation, place recognition, and absolute pose regression.

Digging Into Self-Supervised Monocular Depth Estimation
Clément Godard, Oisin Mac Aodha, Michael Firman and Gabriel J. Brostow
ICCV, 2019
paper / video / code

We dug deeper into self-supervision for monocular depth estimation and improved on monodepth.

Deep Burst Denoising
Clément Godard, Kevin Matzen and Matt Uyttendaele
ECCV, 2018
project page / paper

We trained a recurrent neural network to denoise bursts of images.

Unsupervised Monocular Depth Estimation with Left-Right Consistency
Clément Godard, Oisin Mac Aodha and Gabriel J. Brostow
CVPR, 2017 oral
project page / video / code

We trained a fully convolutional neural net to infer single frame depth only using stereo pairs as training data and beat supervised methods on KITTI.

Multi-view Reconstruction of Highly Specular Surfaces in Uncontrolled Environments
Clément Godard*, Peter Hedman*, Wenbin Li and Gabriel J. Brostow
3DV, 2015 oral
project page / video / code

We reconstruct mirror-like objects from two dozen images in uncontrolled environments, while handling inter-reflections.

this page template was shamelessly stolen from Jon Barron