AI Research Manager @ Dassault Systèmes
Corporate Research, Velizy, France.
2015 - avril 2018
Mixing the power of deep machine learning with 3D computer graphics/vision. Some realisations:
{ Design and implementation of an RGB-D scene parsing component (semantic segmentation) for an indoor
home staging application.
{ Design and implementation of a Learning-based Single-view Body Shape Regression from RGB-D data
(Patent).
{ Design and implementation of a 3D Sensor Simulation component for RGB-D synthetic database generation
(Patent).
{ Design and implementation of a 3D Parametric Body Model Learning component for a magic mirror
application.
{ Design and implementation of a 3D Object Feature Learning and Indexing component for CAD search
retrieval application.
2012-2014 Co-founder and CTO @ SmartMeUp, Grenoble, France.
A startup company focusing on machine learning applications and major attention in face analysis.
Some technical missions:
{ Implementation in Python of SmartMeUp’s Machine Learning toolbox: feature extraction, data preprocessing,
clustering, classification and regression algorithms for fast applications experimentation and results reporting.
{ Research, design and implementation of SmartMeUp’s SDK (Python, C++) for video real-time face detection
and attributes recognition (mood, gender, age, 3d pose, ..).
{ Development of an ICAO standards checker (frontal 3d pose, neutral expression, closed mouth, open eyes,
no hats, ..) for Photomaton company (photobooth).
{ Development of a state of the art real-time face recognition module for Netatmo IoT company.
{ Experimentation of a Deep Convolutionel Neural Networks (CNN) on GPU for large scale image classification
(ImageNet dataset, 1.4M of images with 1000 object classes).
{ Design and implementation of a license plate detection and recognition system.
2010-2012 AI Research Engineer @ INRIA Research Center, Grenoble, France.
Within the Quaero european research program:
{ Implementation of a C/Python package for image classification based on Fisher Vectors and Spatial Pyramid
(Sift, PCA, GMM, linear SVM...)
{ Design and implementation of a two modalities (Video and Audio) event detection system.
Participation to TRECVID Media Event Detection challenge (MED’11): Experimentation on
tens of thousands of videos using a cluster of computers, under strong time constraints (C, Python, MATLAB,
dense trajectories, MFCC, Bag of Features(BoF), SVM..). Ranked 8/20
{ Implementation of a Human Action Recognition algorithm based on dense trajectories (C, Python, KTH
and Hollywood2 datasets, BoF , Fisher vectors, SVM ..)
{ Experimentation and implementation of Split/Merge EM algorithm for Gaussian Mixture Model (GMM)
training (C, Expectation-Maximization, component splitting, Kmeans...)
{ Experimentation and implementation of image clustering algorithm for image retrieval results improvement
(C, Python, Gist, Kmeans, Kernel PCA Kmeans, Hierarchichal clustering ..)
{ Implementation of an image classification algorithm based on BoF descriptors and SVM classifier(C, Python,
Pascal’07 database, Chi Kernel, intersection Kernel, Fisher Kernel, linear Kernel ..).
Feb-Oct 2010 Computer Vision Research Engineer @ THALES TTS, Cergy, France.
Real-Time 3D head pose and facial features estimation using monocular images.
{ State of the Art: Active Appearance Model, Active Contour Model, Gradient descent algorithm, Particle
filter tracking..
{ Implementation/Experimentation in Matlab
{ C++ coding and optimization in GPU (CUDA Toolkit)