SemiautomatictrackingbasedlabelingtoolforautomotiveapplicationsITS2015智能交通世界大会ITS智慧城市社区人工智能AI物联网IT报告课件教案.pdfVIP

22nd ITS World Congress, Bordeaux, France, 5–9 October 2015 Paper number ITS-2643 Semi-automatic tracking-based labeling tool for automotive applications Andoni Cortés, Orti Senderos, Nerea Aranjuelo, Marcos Nieto and Oihana Otaegui Vicomtech-IK4, Intelligent Transport Systems and Engineering Dpt., Paseo Mikeletegi 57, 20009, Donostia-San Sebastián, Spain. {acortes,osenderos,naranjuelo,mnieto,ootaegui}@ Abstract The trend toward smart cars continues to build momentum in the automotive industry. As vision based sensors proliferate in the vehicles the quantity and variety of data will become overwhelming and difficult to be processed effectively. Computer vision-based approaches can be trained and evaluated with such data sources once adequately labeled. However, accurately annotating entities in video is labor intensive and an expensive task. As the quantity of available video grows, traditional solutions to this task are unable to scale to meet the needs of sectors like automotive or security. We present a semi-automatic multi-purpose annotation tool which reduces the manual annotation effort by enabling the user to verify automatically generated annotations, rather than annotating from scratch. This tool has been successfully used in a variety of example applications in the domain of Advanced Driver Assistance Systems. Keywords: Massive labelling, video annotation, video summarization Introduction In-Vehicle Infotainment (IVI) and Advanced Driver Assistance System (ADAS) technologies are becoming increasingly relevant as they promote improved passenger experiences and increased s