Megane Millan

Experience

• 

  University of Birmingham

  Jun 2016 - Aug 2016

  Technical internship

  - team work on an autonomous robot (named Bob) belonging to the STRANDS project- aim : allow the robot to take the lift and integrate it in the main behavior of the robot- programmed in python, C++, openCV and ROS - pattern recognition part ( floor number recognition in the elevator)

• 

  Angus.ai

  Feb 2017 - Aug 2017

  Stagiaire en vision par ordinateur - Intern in computer vision

  - Human tracking using a 2D camera- Development of a gaze following algorithm using a 2D camera

• 

  ISIR - Institut des Systèmes Intelligents et de Robotique

  Oct 2017 - Sept 2022

  Learning a new sport or manual work is complex. Indeed, many gestures haveto be assimilated in order to reach a good level of skill. However, learning thesegestures cannot be done alone. Indeed, it is necessary to see the gesture executionwith an expert eye in order to indicate corrections for improvement. However,experts, whether in sports or in manual works, are not always available to analyzeand evaluate a novice’s gesture.In order to help experts in this task of analysis, it is possible to develop virtualcoaches. Depending on the field, the virtual coach will have more or less skills, butan evaluation according to precise criteria is always mandatory. Providing feedbackon mistakes is also essential for the learning of a novice.In this thesis, different solutions for developing the most effective virtualcoaches are proposed.First of all, and as mentioned above, it is necessary to evaluate the gestures. Fromthis point of view, a first part consisted in understanding the stakes of automaticgesture analysis, in order to develop an automatic evaluation algorithm that is asefficient as possible. Subsequently, two algorithms for automatic quality evaluationare proposed. These two algorithms, based on deep learning, were then tested ontwo different gestures databases in order to evaluate their genericity.Once the evaluation has been carried out, it is necessary to provide relevantfeedback to the learner on his errors. In order to maintain continuity in the workcarried out, this feedback is also based on neural networks and deep learning.A method has been developed based on neural network explanability methods. Itallows to go back to the moments of the gestures when errors were made accordingto the evaluation model. Finally, coupled with semantic segmentation, this methodmakes it possible to indicate to learners which part of the gesture was badlyperformed, and to provide them with statistics and a learning curve. Show less

  • Research Engineer

    Oct 2021 - Sept 2022

  • Research and Teaching Associate

    Oct 2020 - Oct 2021

  • PHD Student

    Oct 2017 - Oct 2020
• 

  Inria

  Feb 2023 - now

  Research Engineer