Interpretability and Explainability of Deep Models

The concept of explainability has several facets and the need for explainability is strong in safety-critical applications such as autonomous driving where deep learning models are now widely used. As the underlying mechanisms of these models remain opaque, explainability and trustworthiness have become major concerns. Among other things, we investigate methods providing explanations to a black-box visual-based systems in a post-hoc fashion, as well as approaches that aim at building more interpretable self-driving systems by design.

Publications

STEEX: Steering Counterfactual Explanations with Semantics

Paul Jacob, Éloi Zablocki, Hédi Ben-Younes, Mickaël Chen, Patrick Pérez, Matthieu Cord
European Conference on Computer Vision, 2022


Explainability of deep vision-based autonomous driving systems: Review and challenges

Éloi Zablocki*, Hédi Ben-Younes*, Patrick Pérez, Matthieu Cord
International Journal of Computer Vision, 2022


Driving behavior explanation with multi-level fusion

Hédi Ben-Younes*, Éloi Zablocki*, Patrick Pérez, Matthieu Cord
Pattern Recognition (PR), 2022