Research researches focus on Deep Convolutional Neural Networks and Geometry Based Models for Computer Vision and Perception Systems.

Concerning Geometry, we are interested in 3D localisation and dense reconstruction from heterogeneous videos provided by different cameras acquiring orme or less trajectories at different times (days or seasons).

Concerning Machine Learning, we propose Deep Models that deal with several domains and heterogeneous annotations (from image classification to fine instance segmentation) for multi object detection, segmentation and tracking into mono or multi camera video sequences . We are also involved into robustness estimation of the output of Neural Networks.