Sebastien Grand
Ph.D. Student
IMT Mines Albi
Hi there, I am Sébastien Grand!
I'm a 1st year PhD student at IMT Mines Albi
working on robust and adaptive Deep Multimodal Fusion.
Previously, I received a M.S in Computer Science and Mathematics, a double diploma from
INP ENSEEIHT and
INSA Toulouse in 2023.
Then, I worked for a year at EPSI Radar at developing Computer Vision classification models for radar data.
Education
IMT Mines Albi
Ph.D. Student in Machine Learning and Computer Vision
Advisor: Aurélie Montarnal, Ph.D., Bruno Mériaux, Ph.D., Frédérick Benaben, Ph.D.
INP ENSEEIHT
M.S. in Computer Science and Mathematics
Highest thesis grade
University of Toulouse 3 Paul Sabatier
B.S. in Electrical Engineering
Top of the class
Publications
Coming 2025
Micro-Doppler Based Human Activity Detection under Adverse Outdoor Conditions using Vision Models
Sébastien Grand, Charles Piffault, Guillaume Pouget Bruno Meriaux, Aurelie Montarnal, Frédérick Benaben
Experience
Research Engineer — EPSI Radar
Manager: Bruno Mériaux, Ph.D
Classification of human activities vs. environment, based on micro-Doppler radar signals, leveraging Convolutional Neural Networks (CNNs) for robust spectrum-based classification. • Developed an ML for human activity detection based on micro-Doppler spectrum classification using Convolutional Neural Networks (CNNs). • Improvement in detection performance compared with the previous model. • Development of the inference pipeline in C++ using ONNXruntime for further deployment.
Research Intern (Master Thesis) — Hensoldt Analytics
Advisor: Jonathan Kobold, Ph.D
• Generated robust motion maps based on videos captured by UAV cameras to feed an object detector and improve its performance at detecting drones. • Developed motion compensation algorithm using Panoramic View building (Image stitching), motion extraction algorithm using Robust PCA, and addition of the motion map to the RGB frame to improve drone detection. • Modified YOLOX (object detector) architecture to add the motion map as a new feature. • Enhanced motion extraction algorithm and improved object detector performance with the addition of the motion map as a new feature.
Portfolio
Generating Robust Motion Maps to Improve Object Detector Performance
This report presents the work I realised during my master thesis at Hensoldt Analytics, in Munich, Germany
Lectures
IMT Mines Albi, Master 1 - option Data Science, Introduction to Machine Learning, 39h
Data Analysis Course, Advanced Techniques for Statistical Data Analysis, 10h
Production Management Workshop, Fundamentals of Production Systems, 15h