The research paper about “Street-View Change Detection with Deconvolutional Networks” which was collaboratively signed by CVC Researcher German Ros, was one of the finalists in 2016 Robotics: Science and Systems Conference
CVC Researcher German Ros from the team of ADAS joined 2016 Robotics: Science and Systems Conference with his collaborative research paper which was chosen one of the finalists for “The Best Systems Paper Award” competing against world class research institutes like MIT. The conference took place in the University of Michigan, USA between 18th and 22nd of June.
In the research, German Ros (CVC), Pablo F. Alcantarilla (iRobot Corporation, London), Simon Stent (University of Cambridge), Roberto Arroyo (University of Alcala, Madrid) and Riccardo Gherardi (Toshiba Research Europe Ltd, UK) worked together in order to propose a solution for more frequent and efﬁcient updates in the large-scale maps used in autonomous vehicle navigation.
The method proposed in the research chains a multi-sensor fusion SLAM and fast dense 3D reconstruction pipeline, which provide coarsely registered image pairs to a deep deconvolutional network for pixel-wise change detection. To train and evaluate our network they introduced a new urban change detection data-set which is an order of magnitude larger than existing datasets and contains challenging changes due to seasonal and lighting variations. The method also outperforms existing literature on this dataset, which was made available to the community, and an existing panoramic change detection dataset, demonstrating its wide applicability.
Massachusetts Institute of technology (MIT) was chosen the best research in the conference where many institutes and research companies from different research fields participated around the world.
The research is publicly available on this link.