Working on Few-shot Learning problems under the supervision of Prof Vincent Gripon and Associate Professors Bastien Pasdeloup and Giulia Lioi.

Unsupervised Cross-Domain adaptation for Medical Localization. During 6 months, I studied different unsupervised or weekly supevised techniques in Deep Learning to transfer disc and vertebrates localization tasks from different modalities. Given a domain A (MRI) where discs are clearer and a domain B (Computed Tomography - CT) where vertebrates are brighter, I explored multiple approaches to localize both tasks on both modalities.

As an Intern I implementend, tested and compared different algorithms for estimating the human skeleton using Human Pose Estimation models applied on images from a Poppy Robot’s Camera for physiotherapy exercises.All my work is described in this link : http://wiki-robot.enstb.org/doku.php?id=comparatif_openpose_kinectAnd GitHub Project in this link : https://github.com/ybendou/HumanPoseEstimation-Use-and-Comparaison.git