Pierre Le Jeune

Pierre Le Jeune

PhD Student in Deep Learning

Sorbonne Paris Nord

Biography

PhD Student at Sorbonne University (USPN) in Paris and research scientist at COSE, a leading french company in the development of embedded systems for intelligence and defense. My work focuses on few-shot object detection applied to aerial images. This project is part of an industrial partnership between COSE and L2TI a laboratory from USPN.

I graduated from both Danish Technical University and Centrale Nantes in 2020 as part of a double-degree exchange program. I hold a MSc in Mathematical Modelling and Computation and a diplôme d’ingénieur with a specialization in computer science and AI. My full resumé can be found here .

Interests
  • Computer Vision
  • Deep Learning
  • Few-shot learning
Education

  • PhD in Artificial Intelligence, 2020

    Sorbonne Paris Nord

  • MSc Mathematical Modelling and Computation, 2018-2020

    Danish Technical University

  • Engineering School, 2016-2020

    Centrale Nantes

Publications

Quickly discover relevant content by filtering publications.
Cross-Scale Query-Support Alignment Approach for Small Object Detection in the Few-Shot Regime
Paper published at ICIP 2023.
Pierre Le Jeune, Anissa Mokraoui
PDF Slides
Cross-Scale Query-Support Alignment Approach for Small Object Detection in the Few-Shot Regime
Object Detection in Aerial Images in Scarce Data Regimes
PhD Thesis.
Pierre Le Jeune
PDF Slides
Object Detection in Aerial Images in Scarce Data Regimes
Extension de l'Intersection over Union pour améliorer la détection d'objets de petite taille en régime d'apprentissage few-shot
Paper published at GRETSI 2023.
Pierre Le Jeune, Anissa Mokraoui
PDF Poster
Extension de l'Intersection over Union pour améliorer la détection d'objets de petite taille en régime d'apprentissage few-shot
Amélioration de la détection d'objets few-shot à travers une analyse de performances sur des images aériennes et naturelles
Paper published at GRETSI 2022.
Pierre Le Jeune, Anissa Mokraoui
PDF Poster
Amélioration de la détection d'objets few-shot à travers une analyse de performances sur des images aériennes et naturelles
Improving Few-Shot Object Detection through a Performance Analysis on Aerial and Natural Images
Paper published at EUSIPCO 2022.
Pierre Le Jeune, Anissa Mokraoui
PDF Talk
Improving Few-Shot Object Detection through a Performance Analysis on Aerial and Natural Images
See all publications

Projects

DTU Master Thesis
3D Ken Burns effect with Generative Adversarial networks.
Pierre Le Jeune
Code Video
DTU Master Thesis
Image Enhancement with GANs
Exposure is a fundamental component of a good picture. However, it can be quite challenging to set the camera parameters to get it right. Overexposed shots can be corrected, but this also demands some expertise.
Pierre Le Jeune
PDF Code Slides
Doom Reinforcement Learning
Training deep reinforcement models to kill demons.
Pierre Le Jeune
PDF Code Video
Doom Reinforcement Learning
What makes a great NBA player
Analysis of the characteristics that define a NBA legend and cool d3.js visualization.
Pierre Le Jeune, Marouann Cheikh
Code
What makes a great NBA player
See all

Recent & Upcoming Talks

Improving Few-Shot Object Detection through a Performance Analysis on Aerial and Natural Images
Presentation for the paper Improving Few-Shot Object Detection through a Performance Analysis on Aerial and Natural Images presented at EUSIPCO 2022 in Belgrade. For more detail see related publication.
Sep 2, 2022 Belgrade, Serbia
Pierre Le Jeune
Slides Article
Improving Few-Shot Object Detection through a Performance Analysis on Aerial and Natural Images
Experience feedback using Representation Learning for Few-Shot Object Detection on Aerial Images
Presentation for the paper Experience feedback using Representation Learning for Few-Shot Object Detection on Aerial Images accepted at ICMLA 2021. For more detail see related publication.
Dec 13, 2021 Pasadena (CA), Virtual
Pierre Le Jeune
Slides Article
Experience feedback using Representation Learning for Few-Shot Object Detection on Aerial Images
Prototypical Faster R-CNN for Few-Shot Object Detection on Aerial Images
A 5-minutes speech at DeepLearn Summer School 2021 in Gran Canaria. I presented some of my work about few-shot object detection and the challenges about this very specific task.
Jul 29, 2021 Las Palmas, Gran Canaria, Spain
Pierre Le Jeune
Slides
Prototypical Faster R-CNN for Few-Shot Object Detection on Aerial Images

Recent Posts

Teaching

 
 
 
 
 
Danish Technical University
Logical Theory for Uncertainty and Learning
Danish Technical University
Sep 2019 – Dec 2019 Copenhagen
 
 
 
 
 
Danish Technical University
Social Data Analysis
Danish Technical University
Feb 2020 – Jun 2020 Copenhagen