Cerca nel sito
ICube   >   Agenda : Séminaire : Random Forest Based Tracking and Detection of 3D Objects in Depth Images

Séminaire : Random Forest Based Tracking and Detection of 3D Objects in Depth Images

30 giugno 2015
Strasbourg - IRCAD - amphi I

Le docteur Slobodan ILIC fera une présentation de ses travaux de recherche le mardi 30 Juin 2015 à 15h00 dans l'amphithéâtre I à l'IRCAD.

Titre : Random Forest Based Tracking and Detection of 3D Objects in Depth Images

Équipe : AVR

Résumé : We propose a novel algorithms, based on random forest, for tracking and detection of rigid objects in depth images using its 3D CAD model only. Our detector finds the object in the scene and estimates its pose and is used for initialisation and re-initialization of the tracker. Detector only makes use of the 3D CAD model and does not require real, labelled data or background modelling; it uses robust and efficient decision features within the forests; and, uses desirable and efficient pose parametrization that covers the full pose-space of the object and helps optimize the splitting in the trees' branches. The combination of these novel attributes results in an algorithm that consistently performs above 98% success rate on the standard dataset which significantly improves the state-of-the-art methods. This is coupled with the temporal tracking algorithm also based on random forest that uses depth images to estimate and track the 3D pose of a rigid object in real-time. Compared to the state of the art aimed at the same goal, our algorithm holds important attributes such as high robustness against holes and occlusion, low computational cost of both learning and tracking stages, and low memory consumption. In addition, the fast learning time enables us to extend our algorithm as a robust online tracker for model-free 3D objects under different viewpoints and appearance changes, as demonstrated by the experiments in terms of online 3D object tracking and 3D head pose estimation.

Bio : Slobodan Ilic is currently senior key expert research scientist at Siemens Corporate Technology in Munich, Perlach at the Sensor Technology Department. He is also a visiting researcher and lecturer at Computer Science Faculty of TUM and closely works with the Vision Group at the CAMP Chair, where he supervises a number of PhD students. From 2009 until end of 2013 he was senior researcher and leader of the Computer Vision Group of CAMP at TUM, and before that he was a senior researcher at Deutsche Telekom Laboratories in Berlin. In 2005 he obtained his PhD at EPFL in Switzerland under supervision of Pascal Fua. His research interests include: 3D reconstruction, deformable surface modelling and tracking, real-time object detection and tracking, human pose estimation and semantic segmentation.

À la une

Le dépôt des candidatures pour les postes d’enseignants-chercheur est ouvert. Les offres sont...

Flux RSS

Flux RSS