Senior Computer Vision Engineer at MobiLysis, developing Computer Vision solutions for large-scale mobility monitoring and analysis from drone imagery.

Postdoctoral Researcher at VITA Lab, EPFL, working on generic Deep Learning, learning human skeleton keypoints or using them for human behavior understanding, and pedestrian analysis for autonomous driving, while also helping and advising on other diverse activities of the lab.Industrial collaborations with Valeo.ai and Dartfish.Main highlights:• Published 10 scientific papers in international conferences/journals (ICCV, NeurIPS, ICRA, T-ITS, TR-C, RA-L, TRISTAN) on the topics of Deep Learning techniques, visual understanding using human skeletons, and pedestrian analysis for autonomous driving;• Outperformed internal and public methods for detection of hockey players with skeletons (>90%) for an Innosuisse project with Dartfish;• Collaborated with Valeo.ai on scaling multi-task models to recognize up to 32 pedestrian attributes, including actions and attention;• Supervised 4 PhD students and 17 semester-long Master-level projects;• Reviewed paper submissions for international conferences/journals (CVPR, ICCV, WACV, TPAMI, IJCV, CVIU, RA-L).

Postdoctoral Researcher at MLIA Lab, Sorbonne Université, working on adding depth privileged information into Unsupervised Domain Adaptation from synthetic data for autonomous driving applications.Main highlights:• Published 1 scientific paper on training Deep Learning models with synthetic data for autonomous driving;• Supervised 1 PhD student in collaboration with Valeo.ai.

Industrial PhD (CIFRE PhD program) carried out at Thales LAS France as a Research Engineer in collaboration with Sorbonne Université. The industrial context of the PhD was about the detection of vehicles in aerial infrared images using Deep Learning.Main highlights:• Completed PhD partly at company, with scientific research topics inspired from relevant industrial problems;• Designed a pipeline for object detection using Deep Learning adapted to the industrial problem, and integrated it into a framework existing at Thales;• Advised on the methodology for diverse Machine Learning questions.

Final year internship at ENSTA ParisTech.Topic: Detection of Small Targets in Aerial Infrared Images.Six-month internship at Sagem DS with the aim of improving the performance of the detection of small vehicles on the VeDAI (Vehicle Detection in Aerial Imagery) database, by using Deep Learning. The main difficulties arose from the reduced sizes of targets (a few dozens of pixels) and from the relatively low number of examples of each class.Main highlights:• Completed a study of the state of the art in object detection, with a focus on Deep Learning and Transfer Learning;• Extracted features from deep Convolutional Neural Networks learned on large databases with the Caffe library using C++, and integrated them within an existing detection pipeline on VeDAI;• Optimized and further developed the whole detection pipeline, leading to a +16% improvement of the state of the art on VeDAI;• Received "André Blanc-Lapierre" award from SEE (French sibling of IEEE) for the best final year internship in STEM within Paris region.Under the supervision of Farid Oudyi.

Second year internship at ENSTA ParisTech.Topic: Coordination Control of Service Robots.Three-month internship at the Laboratory for Intelligent Robots & Vision System (IRVS), Kyushu University, within the frame of Robot Town, a long-term project aiming at the insertion of robots in daily life.Main highlights:• Initiated the integration of a new Robotics motion planning framework (ROS MoveIt!, C++) in the laboratory to control various robots;• Applied it to the cooperation between two mobile robots: a mobile base Pioneer 3-DX and a humanoid robot SmartPal V.Under the supervision of Ryo Kurazume.

First year internship at ENSTA ParisTech.Three-week Japanese seminar followed by an intensive one-month manual work experience in a Japanese hotel. Work included room cleaning and restaurant serving.