Elliot Vincent

Summary

I am a second-year PhD student (expected graduation on 09/2024) in the Imagine research group at Ecole des Ponts ParisTech and in the Willow team at Inria Paris. My research topics are machine learning and computer vision for remote sensing. My work currently focuses on satellite image time series and their analysis with few or no annotations for tasks such as land use classification or change detection.

Supervisors

I work under the supervision of Mathieu Aubry (LIGM, Ecole des Ponts, Univ Gustave Eiffel, CNRS, France) and Jean Ponce (Inria and DIENS (ENS-PSL, CNRS, Inria) and Center for Data Science, New York University).

Publications

Learnable Earth Parser: Discovering 3D Prototypes in Aerial Scans
Romain Loiseau, Elliot Vincent, Mathieu Aubry, Loïc Landrieu
arXiv 2023
We propose an unsupervised method for parsing large 3D scans of real-world scenes into interpretable parts. Our goal is to provide a practical tool for analyzing 3D scenes with unique characteristics in the context of aerial surveying and mapping, without relying on user annotations. Our method offers significant advantage over existing approaches.
Pixel-wise Agricultural Image Time Series Classification: Comparisons and a Deformable Prototype-based Approach
Elliot Vincent, Jean Ponce, Mathieu Aubry
arXiv 2023
We benchmark multivariate time series classification methods on four satellite image time series dataset and propose a prototype-based approach, adding invariance to spectral deformations and temporal shifts and improving over existing baselines.
A Model You Can Hear: Audio Classification with Playable Prototypes
Romain Loiseau, Baptiste Bouvier, Yann Teytaut, Elliot Vincent, Mathieu Aubry, Loïc Landrieu
CVPR Sight and Sound Workshop 2022 - ISMIR 2022
We propose an audio identification model based on learnable spectral prototypes. Our model can be trained with or without supervision and reaches state-of-the-art results for speaker and instrument identification, while remaining easily interpretable.
Unsupervised Layered Image Decomposition into Object Prototypes
Tom Monnier, Elliot Vincent, Jean Ponce, Mathieu Aubry
ICCV 2021
An unsupervised learning framework to decompose images into object layers modeled as transformations of learnable sprites.

Resume

[Mar-Aug 2020] Research internship at Imagine and Willow on unsupervised 2D scene decomposition
[2019-2020]       Mathematics, Vision and Learning master of the ENS and engineering program of ENPC
[Mar-Aug 2019] Internship at IBM Research Zürich on automated methods for diagram understanding
[2016-2019]       Master in Applied Mathematics for Computer Science at École polytechnique

About me

In addition to my interest for computer vision and machine learning, I also graduated with a Master's degree in public policies for environmental protection. I love animals of all sorts but have a weird and particular obsession for otters. On my free time, you will see me playing football with friends or enjoying a nice swim at the pool.

Affiliations

- LIGM (UMR 8049), Ecole des Ponts, Univ. Gustave Eiffel, CNRS
- Inria and DIENS (ENS-PSL, CNRS, Inria)