Bruno Lecouat

Machine Learning Researcher, Apple
Zürich, Switzerland

I am a machine learning researcher at Apple in Zürich, working with Stephan Richter and Vladlen Koltun.

Before that I completed a PhD at Inria and École Normale Supérieure, advised by Jean Ponce and Julien Mairal, on machine learning for computational imaging. I defended in November 2023; the thesis, Image Reconstruction from Multiple Shots with Trainable Algorithms, is available on HAL.

From 2022 to 2024 I was co-founder and CTO of Enhance Lab, building image quality enhancement software for smartphone cameras out of that research. Earlier, I studied signal processing and communication theory at Télécom Paris and electrical engineering at the National University of Singapore, and spent a year at A*STAR in Singapore, working on semi-supervised learning and anomaly detection with Chuan-Sheng Foo and Vijay Chandrasekhar.

Burst super-resolution

Merging 30 noisy raw frames from a handheld camera into a single image, upsampled by a factor of 4 along each dimension. On the left, the 20 megapixel JPEG produced by the camera's own ISP. On the right, our result, developed with a handcrafted ISP, at 325 megapixels. That is far too large for a browser to load at once, so it is served as a tiled deep-zoom image. Both views are synchronized: drag to pan, scroll to zoom.

Camera JPEG

Ours, from 30 raw frames

Handheld Panasonic Lumix GX9. More scenes, side-by-side comparisons and synthetic results are on the project page.

Note: the two renderings are not meant to match. Tone, color and sharpening come from two different pipelines, so they differ by construction. What the comparison shows is the quality of the restored signal, that is noise and fine detail, rather than color rendition.

Publications

Fine Dense Alignment of Image Bursts through Camera Pose and Depth Estimation

Bruno Lecouat*, Yann Dubois de Mont-Marin*, Théo Bodrito*, Julien Mairal, Jean Ponce

arXiv, 2023 arXiv

High Dynamic Range and Super-Resolution from Raw Image Bursts

Bruno Lecouat, Thomas Eboli, Jean Ponce, Julien Mairal

SIGGRAPH 2022 arXiv

Lucas-Kanade Reloaded: End-to-End Super-Resolution from Raw Image Bursts

Bruno Lecouat, Jean Ponce, Julien Mairal

ICCV 2021 arXiv project page

A Flexible Framework for Designing Trainable Priors with Adaptive Smoothing and Game Encoding

Bruno Lecouat, Jean Ponce, Julien Mairal

NeurIPS 2020 arXiv code

Fully Trainable and Interpretable Non-Local Sparse Models for Image Restoration

Bruno Lecouat, Jean Ponce, Julien Mairal

ECCV 2020 arXiv code

Optimistic Mirror Descent in Saddle-Point Problems: Going the Extra (-gradient) Mile

Panayotis Mertikopoulos, Bruno Lecouat, Houssam Zenati, Chuan-Sheng Foo, Vijay Chandrasekhar, Georgios Piliouras

ICLR 2020 arXiv code

Semi-Supervised Deep Learning for Abnormality Classification in Retinal Images

Bruno Lecouat, Ken Chang, Chuan-Sheng Foo, Balagopal Unnikrishnan, James M. Brown, Houssam Zenati, Andrew Beers, Vijay Chandrasekhar, Jayashree Kalpathy-Cramer, Pavitra Krishnaswamy

NeurIPS 2019, Machine Learning for Health workshop arXiv

Adversarially Learned Anomaly Detection

Houssam Zenati, Manon Romain, Chuan-Sheng Foo, Bruno Lecouat, Vijay Chandrasekhar

ICDM 2019 arXiv code

Semi-Supervised Learning with GANs: Revisiting Manifold Regularization

Bruno Lecouat*, Chuan-Sheng Foo*, Houssam Zenati, Vijay Chandrasekhar

ICLR 2019, workshop track arXiv code

Efficient GAN-Based Anomaly Detection

Houssam Zenati*, Chuan-Sheng Foo*, Bruno Lecouat, Gaurav Manek, Vijay Chandrasekhar

Preprint, 2018 arXiv code