Rémy Siegfried

About myself

I am from Saint-Maurice in Valais. I studied at EPFL where I obtained a Bachelor in Microengineering (2014) and then a Master (MSc) in Robotics and Autonomous Systems (2016). Then, I worked for 5 years at the Idiap Research institute, where I got my PhD (2021) under the supervision of Jean-Marc Odobez in the Electrical Engineering Doctoral Program (EDEE) of EPFL.

Current work and research interests

I work as a postdoctoral researcher at Idiap in the Perception and Activity Understanding group.

I am interested in machine learning and data analysis in general and more specifically in human perception and its application to human-computer/robot interactions and activity understanding.

Past projects

  • 2021-2022 I worked on the P3 project (Innosuisse, SNSF) project with the HES-SO and a company that wanted to improve their feasibility study for new ordered parts and the optimization of process parameters. The outcome is a web interface that provide a tool to browse past produced parts using visual comparison and a model that predicts important processing variables depending on the process parameters.
  • 2021-2022 In the frame of the NATAI project ("Agora" project, SNSF), I collaborated with the Musée de la Main (CHUV, Lausanne) and provide them a live gaze tracking demo for their exhibition on artificial intelligence. The result is an autonomous demonstrator that displays a scene, estimate what the user is looking at, and adapts the audio accordingly.
  • 2018-2021 The main funding for my thesis comes from the MuMMER project ("Horizon2020", EU), where I worked on modeling and infering attention in human-robot interactions. Exploiting color and depth images as well as audio data, my goal was to estimate the individual attention of a group of people that is interacting with a robot in order to better understand conversations dynamic. I explored different topics and tasks, like unsupervised gaze estimation calibration, eye movements recognition, and attention estimation in arbitray settings.
  • 2017 During my PhD, I was involved in the UBImpressed project ("Sinergia", SNSF), whose goal was to study the building of the first impression and its application in hospitality employees training. I mainly worked on gaze estimation and calibration.
  • 2016 I did my Master Project at MOBOTS group (EPFL) under the supervision of Francesco Mondada. I worked in the field of learning analytics with mobile robots. I worked on methods that use the logs taken during a robot programming lecture to provide useful information to teachers and students in order to increase the learning outcome of lectures. I was then hired for 6 more monthes to continue my master project and develop a tool that provides online hints to students learning robotic programming based on the results of my master project.
  • 2015 I worked during 7 monthes for senseFly (in Cheseaux-sur-Lausanne) on the motor control of their quadrotor and the development of a new camera interface (hardware) for a fixed-wing drone.
  • 2014-2015 During my studies, I performed two semester projects: one on the implementation of safety behaviour on quadrotor formation (at DISAL (EPFL)) and a second on the design of legs for a quadruped robot (at BioRob (EPFL))

Material

  • Code and data: Unsupervised gaze estimation calibration in conversation and manipulation settings (GitHub)
    In our 2021 journal paper, we introduced a method to allow the unsupervised calibration of a gaze estimator using contextual prior based on top-down attention (i.e. related to the current task).
    The data used in our experiments is available here: Idiap page Zenodo
  • ManiGaze dataset (Idiap page
    The ManiGaze dataset was created to evaluate gaze estimation from remote RGB and RGB-D (standard vision and depth) sensors in Human-Robot Interaction (HRI) settings, and more specifically during object manipulation tasks. The recording methodology was designed to let the user behave freely and encourage a natural interaction with the robot, as well as to automatically collect gaze targets, since a-posteriori annotation is almost impossible for gaze.
  • VFOA module (GitHub
    A python package for the basic visual focus of attention estimation of people in a 3D scene (geometrical and statistical models).

Publications

Events and media

Contact

Personal Page

remy.siegfried@idiap.ch



Tel: +41 27 721 7707
Office: 308-6
Contact