Talks

2024

2024

1. Oral presentation at AISTATS, Valencia, Spain

   Equivariant Bootstrapping for Uncertainty Quantification in Inverse Problems., May 2024
2. Imaging inverse problems and generating models workshop, ICMS, Edinburgh, UK

   Equivariant Plug-and-Play algorithms for Image Restoration., Apr 2024
3. SIAM Uncertainty Quantification, Trieste, Italy

   Deep Inverse: a PyTorch Library for Solving Inverse Problems with Deep Learning., Feb 2024
4. Complex Media Lab seminar, ENS, Paris, France

   Jan 2024
5. Institut de Mathematiques de Toulouse, Toulouse, France

   Towards knowledge discovery in imaging using (deep) learning: uncertainty quantification and unsupervised learning., Jan 2024

2023

2023

1. Workshop on co-design in computational imaging, IHP Paris, France

   Tutorial: learning without ground-truth data for computational imaging problems., Nov 2023
2. Invited talk at Computational Imaging Group of Purdue University

   Deep Inverse: a pytorch library for solving inverse problems with deep learning., Nov 2023

   HTML Slides
3. Feature speaker LatinX in AI workshop, ICCV, Paris, France

   Deep Inverse: a pytorch library for solving inverse problems with deep learning., Oct 2023

   HTML Slides
4. 30 years of mathematics for optical imaging, Marseille, France

   Deep Inverse: a pytorch library for solving inverse problems with deep learning., Sep 2023

   HTML Slides
5. ICIAM 2023, Tokyo, Japan

   Aug 2023
6. Maths4DL Conference, London, UK

   Jul 2023

   HTML Slides
7. CREATIS lab, INSA de Lyon, France, UK

   Jun 2023
8. Laboratoire Jean Kuntzmann seminar, Grenoble, France

   May 2023

   HTML
9. Workshop on Mathematical Signal and Image Analysis (MSIA23), TUM, Raitenhaslach, Germany

   Learning to Reconstruct Signals from Binary Measurements, Mar 2023

   HTML Slides
10. Interfacing Bayesian statistics, deep learning, and mathematical analysis for imaging inverse problems Workshop, ICMS, Edinburgh, UK

    Tutorial on Imaging with Equivariant Deep Learning., Jan 2023

    Slides

2022

2022

1. NeurIPS’22, New Orleans, USA

   Unsupervised Learning From Incomplete Measurements for Inverse Problems., Nov 2022

   Poster Slides
2. ICTEAM Seminar, UCLouvain, Louvain-la-Neuve, Belgium

   Learning to solve inverse problems without ground truth., Oct 2022
3. MACS Seminar, Heriot-Watt, Edinburgh, UK

   Learning to solve inverse problems without ground truth., Oct 2022
4. GRETSI’22, Nancy, France

   Unsupervised Learning to Solve Inverse Problems: Application to Single-Pixel Imaging, Sep 2022
5. StatMathAppli Workshop, Frejus, France

   Learning to solve inverse problems without ground truth., Aug 2022

   Slides
6. Swiss Equivariant Learning Workshop, EPFL, Switzerland

   Equivariant Imaging, Jul 2022

   HTML
7. ICASSP’22, Singapore, virtual

   Sketched RT3D: How to reconstruct billions of photons per second, May 2022

   Abs Video Poster

   Single-photon light detection and ranging (lidar) captures depth and intensity information of a 3D scene. Reconstructing a scene from observed photons is a challenging task due to spurious detections associated with background illumination sources. To tackle this problem, there is a plethora of 3D reconstruction algorithms which exploit spatial regularity of natural scenes to provide stable reconstructions. However, most existing algorithms have computational and memory complexity proportional to the number of recorded photons. This complexity hinders their real-time deployment on modern lidar arrays which acquire billions of photons per second. Leveraging a recent lidar sketching framework, we show that it is possible to modify existing reconstruction algorithms such that they only require a small sketch of the photon information. In particular, we propose a sketched version of a recent state-of-the-art algorithm which uses point cloud denoisers to provide spatially regularized reconstructions. A series of experiments performed on real lidar datasets demonstrates a significant reduction of execution time and memory requirements, while achieving the same reconstruction performance than in the full data case.

8. TeSA lab seminar, Toulouse, France

   Equivariant Imaging: learning to solve inverse problems without ground truth., Mar 2022

   Abs HTML

   In recent years, deep neural networks have obtained state-of-the-art performance in multiple imaging inverse problems ranging from medical imaging to computational photography. Networks are generally trained with pairs of signals and associated measurements. However, in various imaging problems, we usually only have access to compressed measurements of the underlying signals, hindering this learning-based approach. Learning from measurement data only is impossible in general, as the compressed observations do not contain information outside the range of the forward sensing operator. In this talk, I will present a new learning framework, called Equivariant Imaging, which overcomes this limitation by exploiting the invariance to transformations (translations, rotations, etc.) present in natural signals. I will also discuss necessary and sufficient conditions for learning without ground truth. Our proposed learning strategy performs as well as fully supervised methods and can handle noisy data. I will show results on various inverse problems, including sparse-view X-ray computed tomography, accelerated magnetic resonance imaging and image inpainting.

9. ENSL Colloquium, ENS de Lyon, France

   Equivariant Imaging: learning to solve inverse problems without ground truth., Jan 2022

   Abs HTML Slides

   In recent years, deep neural networks have obtained state-of-the-art performance in multiple imaging inverse problems ranging from medical imaging to computational photography. Networks are generally trained with pairs of signals and associated measurements. However, in various imaging problems, we usually only have access to compressed measurements of the underlying signals, thus hindering this learning-based approach. Learning from measurement data only is impossible in general, as the compressed observations do not contain information outside the range of the forward sensing operator. In this talk, I will present a new learning framework, called Equivariant Imaging, which overcomes this limitation by exploiting the invariance to transformations (translations, rotations, etc.) present in natural signals. I will also discuss necessary and sufficient conditions for learning without ground truth. Our proposed learning strategy performs as well as fully supervised methods and can handle noisy data. I will show results on various inverse problems, including sparse-view X-ray computed tomography, accelerated magnetic resonance imaging and image inpainting.

2021

2021

1. TOPML Conference, Rice University, virtual

   The role of overparameterization and optimization in CNN denoisers, Apr 2021

   HTML Slides
2. IMB, Université Bordeaux, virtual

   The neural tangent link between CNN denoisers and non-local filters, Jan 2021

   Slides
3. CVPR’21, oral presentation, virtual

   The neural tangent link between CNN denoisers and non-local filters, Jun 2021

   Video Poster

2020

2020

1. PhD defence, ENSEEHIT, France

   Bayesian methods for inverse problems with point clouds: applications to single-photon lidar, Jan 2020

   Slides

2019

2019

1. CAMSAP’19, Guadaloupe, West Indies

   Real-time color 3D reconstruction from single-photon lidar data, Dec 2019

   Slides
2. EUSIPCO’19, La Coruña, Spain

   Fast surface detection in single-photon lidar waveforms, Sep 2019

   Slides
3. 3MT EUSIPCO’19, La Coruña, Spain

   3D reconstruction from single-photon lidar data, Sep 2019

   Video

4. STEM for Britain’19, Westminster Parliament, UK

   3D reconstruction using single-photon lidar data, Mar 2019

   HTML Poster
5. CAMSAP’19, Guadaloupe, West Indies

   Real-time color 3D reconstruction from single-photon lidar data, Dec 2019

   Slides
6. EUSIPCO’18, Rome, Italy

   Bayesian reconstruction of photon-starved images, Sep 2019

   Slides