We are building an artificial intelligence research center for automotive applications based in the center of Paris, a project started in 2017 to conduct ambitious research projects, regarding assisted and autonomous driving

Artificial intelligence is already at the core of the transformation of the automotive industry! 10 years ago, there was no real AI in cars. Today, most are packed with software, part of which is AI related. Artificial intelligence is changing mobility everywhere!

Valeo DRIVE4U® Sensors – Enabling autonomous driving in city centers

Multi-sensor perception

Automated driving relies first on a diverse range of sensors, like Valeo’s cameras, LiDARs, radars and ultrasonics. Exploiting at best the outputs of each of these sensors at any instant is fundamental to understand the complex environment of the vehicle. To this end, we explore various deep learning approaches where sensors are considered both in isolation and collectively.

Domain adaptation

Deep learning and reinforcement learning are key technologies for autonomous driving. One of the challenges they face is to adapt to conditions which differ from those met during training. To improve systems’ performance in such situations, we explore so-called domain adaption techniques, as in AdvEnt, our project presented at CVPR 2019.

Uncertainty estimation

When the unexpected happens, when the weather badly degrades, when a sensor gets blocked, the embarked perception system should diagnose the situation and react accordingly, e.g, by calling an alternative system or the human driver. With this in mind, we investigate automatic ways to assess the uncertainty of a system and to predict its performance.

Meet our team

  • Ph.D. student Florent Bartoccioni

    Ph.D. student

    Perception | Scene understanding | Dynamic forecasting

    ENS Rennes | CTU Prague | INRIA

  • Research Scientist Alexandre Boulch

    Research Scientist

    Computer vision | Deep Learning | Geometry processing

    X | MVA | ENPC | ONERA


  • Research Scientist Andrei Bursuc

    Research Scientist

    Machine Learning | Computer Vision | Visual Search | Autonomous systems

    Politehnica | Mines | Inria | Safran


  • Research Scientist Mickaël Chen

    Research Scientist

    Deep Learning | Computer Vision | Deep Generative Models

    Sorbonne Université


  • Principal scientist Matthieu Cord

    Principal scientist

    Deep Learning | Computer Vision | Vision and Language

    Enseirb | CergyU | KULeuven | Ensea | CNRS | SorbonneU | IUF

  • Research Scientist Spyros Gidaris

    Research Scientist

    Deep Learning | Computer Vision

    AUTH | Cortexica | ENPC


  • Research Scientist David Hurych

    Research Scientist

    Machine Learning | Computer Vision | Generative Networks

    CTU-Prague | NII-Tokyo

  • Principal scientist Renaud Marlet

    Principal scientist

    Computer Vision | Photogrammetry | Geometry Processing

    X | Inria | EdinburgU | Simulog | Inria | TrustedLogic | Inria | ENPC


  • Research Scientist Gilles Puy

    Research Scientist

    Deep Learning | Signal & Image Processing

    Supélec | EPFL | INRIA | Technicolor


  • Research Scientist Oriane Siméoni

    Research Scientist

    Computer vision | Deep Learning

    Enseirb | INRIA


  • Research scientist Tuan-Hung Vu

    Research scientist

    Deep Learning | Computer Vision

    Telecom | Inria | NEC


  • Research Scientist Eloi Zablocki

    Research Scientist

    Deep Learning | Computer Vision | Vision and Language

    X | MVA | SorbonneU

  • Ph.D. student Léon Zheng

    Ph.D. student

    Machine Learning | Frugal Learning

    X | MVA | ENS Lyon

Recent Activities x ICCV’21 participates to ICCV, the premier computer vision conference, in October 2021, presenting six papers on the challenge of understanding complex scenes with cameras, radar or lidar. x CVPR’21 participates to CVPR , the premier computer vision conference, in June 2021, presenting three papers, contributing to a tutorial on self-supervised learning, co-organizing the workshop on omnidirectional computer vision and presenting keynotes to SafeAI4AutonomouDriving and Vision4AllSeason workshops.

Woodscape dataset
Four Valeo teams collaborate to release Woodscape, the first public multi-sensor driving dataset with fisheye cameras, named after Robert Wood who invented the fisheye camera in 1906, featuring 9 perception tasks such as 2D and 3D object detection, semantic segmentation and depth estimation.

Carrada dataset
In collaboration with researchers at Télécom Paris, releases the Camera and Automotive Radar with Range-Angle-Doppler Annotations (Carrada) dataset, the first public automotive radar dataset with cars, cyclists and pedestrians precisely annotated in the raw signals.

Can we make driving systems explainable? researchers release a comprehensive survey on the explainability of vision-based driving systems, presenting a wide range of existing techniques for post-hoc or by-design explainability, analysing their current limitation and outlining future research avenues toward a better interpretation of self-driving AI models.

Paying attention to vulnerable road users
Vulnerable road users such as pedestrians must be reliably analysed by ADAS and AD systems; shows that synthetic people in real scenes help to train better detectors (collaboration with CTU Prague) and that a multi-task model can recognize up to 32 attributes, including action and attention (collaboration with EPFL).