👋 Hey, I'm Théo Jaunet

I am interested in Deep Neural Networks Interpretability (XAI) with the help of Data Visualization to tackle their exploitation of biases.
I'm just defended my Ph.D. at LIRIS lab, INSA-Lyon (France)! You can have more information about it here!

📢 I am currently looking for a post-doc in visualization for XAI! 📢

Publications

Sim2RealViz

Sim2RealViz, is a visual analytics tool to assist experts in understanding and reducing this gap for robot ego-pose estimation tasks, i.e. the estimation of a...

XAI4Debugging @ NeurIPS 2021 — Théo Jaunet, Guillaume Bono, Romain Vuillemot, Christian Wolf

VisQA

We introduce VisQA, a visual analytics tool that explores the question of reasoning vs. bias exploitation in Visual Question Answering systems.

IEEE Vis 2021 — Théo Jaunet*, Corentin Kervadec*, Grigory Antipov, Moez Baccouche, Romain Vuillemot, Christian Wolf * Both authors contributed equally

Reasoning Patterns

We introduce an in-depth analysis of reasoning patterns at works in Transformer-based VQA models, and propose to transfer these patterns from an oracle to a...

CVPR 2021 — Corentin Kervadec*, Théo Jaunet*, Grigory Antipov, Moez Baccouche, Romain Vuillemot, Christian Wolf * Both authors contributed equally

Théo Guesser

We propose new techniques for the visualization of the sim2real gap, which provide insights into the difference in performance obtained when neural networks are applied...

VISXAI @ IEEE Vis 2020 — Théo Jaunet, Romain Vuillemot, Christian Wolf

DRLViz- Understanding Decisions and Memory

A visual analytics interface to interpret the internal memory of an agent (e.g. a robot) trained using deep reinforcement learning.

EuroVis 2020 — Théo Jaunet, Romain Vuillemot, Christian Wolf

Memory Reduction

We built Doom player AI using Deep Reinforcement learning. While playing, it builds and updates an inner representation (memory) of the game, its environment. Reducing...

VISXAI @ IEEE Vis 2019 — Théo Jaunet, Romain Vuillemot, Christian Wolf — Best Submission Prize

Towards Fuzzy Geo-Set Visual Analysis

Reasoning over spatial regions is frequent in everyday life. For instance, when moving to a new city, one may want to pick a locations based...

SetVa @ IEEE Vis 2019 — Romain Vuillemot, Liqun Liu, Théo Jaunet

Teaching

2021-2022

  • ECL MOS 5.5 - Interactive Data Visualization (Master level)

    Visualization of Deep Neural Networks Lecture (1h30) at Ecole Centrale Lyon (ECL)

  • IA6 - HCI and AI (Master level)

    Interpretability of Deep Learning Lecture (2 hours) at Lyon 1 University

2020-2021

  • IA5 - Bio-Inspired Intelligence (Master level)

    Group Projects of Deep Q Networks (6 hours) at Lyon 1 University

  • ECL MOS 5.5 - Interactive Data Visualization (Master level)

    Visualization of Deep Neural Networks Lecture (1h30) & Application Project (2 hours)
    Group Projects of Visualizations (14 hours) Ecole Centrale Lyon (ECL)

  • IA6 - HCI and AI (Master level)

    Interpretability of Deep Learning Lecture (2 hours) & Application project (2 hours) at Lyon 1 University

2019-2020

  • IA5 - Bio-Inspired Intelligence (Master level)

    Group Projects of Deep Q Networks (6 hours) at Lyon 1 University

  • ECL MOS 5.5 - Interactive Data Visualization (Master level)

    Group Projects of Visualizations (14 hours) at Ecole Centrale Lyon (ECL)

  • INF TC - Python (Undergraduate level)

    Introduction to Python classes (8 hours) at Ecole Centrale Lyon (ECL)

2018-2019

  • DS2 - Data Visualization (Master level)

    Group Projects of Visualizations (16 hours) at Lyon 1 University