I'm Mickaël Chen,

Currently a Deep Learning Research Scientist at Valeo.ia in Paris, working with Deep Generative and Spatio-Temporal Models.

You can contact me in French or English.


STEEX: Steering Counterfactual Explanations with Semantics

We propose a method to generate counterfactual images that scales up to complex scenes, with application for autonomous driving tasks. To do so, we leverage recent segmentation-to-image networks.

Paul Jacob, Éloi Zablocki, Hédi Ben-Younes, Mickaël Chen, Patrick Pérez, Matthieu Cord

[Preprint] [GitHub]

Will be presented at ECCV2022, Tel-Aviv

A Neural Tangent Kernel Perspective of GANs

GAN analyses by modeling the discriminator in its infinite-width limit. By taking into account the impact of the neural architecture and gradient descent training, this modelization explains, among other results, why GANs can be trained even though an optimal arbitrarily powerful discriminator would provide no gradients.

Jean-Yves Franceschi*, Emmanuel de Bézenac*, Ibrahim Ayed*, Mickaël Chen, Sylvain Lamprier, and Patrick Gallinari

[Preprint] [GitHub]

Will be presented at ICML 2022, Baltimore.

Raising context awareness in motion forecasting

We find that state-of-the-art forecasting methods tend to overly rely on the agent's dynamics and fail to exploit the context cues provided at its input. We fix this by introducing a forecasting model equipped with a training procedure designed to promote the use of semantic contextual information.

Hédi Ben-Younes*, Éloi Zablocki*, Mickaël Chen, Patrick Pérez, Matthieu Cord


Will be presented at the CVPR WAD 2022, New-Orleans.

Stochastic Latent Residual Video Prediction

A fully latent temporal model for stochastic video prediction that achieves state-of-the-art performances.

Jean-Yves Franceschi*, Edouard Delasalles*, Mickaël Chen, Sylvain Lamprier, and Patrick Gallinari

[Preprint] [Project page] [GitHub]

Presented at ICML 2020.

ReDO: Unsupervised Object Segmentation by Redrawing

We discover meaningful segmentation masks by redrawing regions of the images independently.

Mickaël Chen, Thierry Artières and Ludovic Denoyer.

[Paper] [Preprint] [GitHub] [Poster] [Slides]

Presented at NeurIPS 2019, Vancouver.

Multi-view Data Generation Without View Supervision

We propose a generative model for multi-view data by decomposing the latent space between content and view.

Mickaël Chen, Ludovic Denoyer and Thierry Artières.

[Paper] [GitHub] [Poster] [Slides]

Presented at ICRL 2018, Vancouver.

Adversarial learning for modeling human motion

We generate Motion Capture Sequences by separating Emotion and Action.

Qi Wang, Thierry Artières, Mickaël Chen, Ludovic Denoyer

[Journal][Preprint (short)]

Presented at ESANN 2018. Journal extension published in The Visual Computer

Multi-view Generative Adversarial Networks

We encode uncertainty in gaussian embeddings using bidirectional GANs.

Mickaël Chen and Ludovic Denoyer.

[Paper] [Preprint] [Poster] [Slides]

Presented at ECML-PKDD 2017, Skopje.


During my Ph.D. I also worked as teaching assistant for the department of Computer Sciences of Sorbonne Université on the following courses:

  • 2016 - 2019: Object Oriented Programming with Java (Licence 2) [ressources]

  • 2018 - 2019: Statistics for computer science (Licence 3)

  • 2017 - 2018: Data science project (Master 1)

  • 2017 - 2018: Introduction to Programming with C (Licence 1)

  • 2016 - 2017: Shell and Bash scripting for Linux (Licence 2)

Ants simulation (OOP/Java exam)
Boids simulation. (OOP/Java exam)

Find me on Social Media