transport logistic 2019

••• 7 ••• Innovationen Training data for autonomous driving Precise annotation of objects in image and video les due to articial intelligence A utonomous cars must per- ceive their environment true to reality. The correspond- ing algorithms are trained using a large number of image and video recordings. For the algo- rithm to recognize single image elements, such as a tree, a pe- destrian or a road sign, these are labeled. Labeling is improved and accelerated by computer sci- entist Philip Kessler, who studied at Karlsruhe Institute of Technol- ogy (KIT), and his colleague Marc Mengler. Troublesome and time-consuming “An algorithm learns by examples and the more examples exist, the better it learns,” Philip Kessler says. For this reason, automotive industry needs a large amount of video and image material in ma- chine learning for autonomous driving. So far, objects on the im- ages have been labeled manually by human staff. “The process is troublesome and time-consum- ing,” Kessler explains. He uses artificial intelligence to make la- beling up to ten times quicker and more precise. “Although image processing is highly automated in large parts, final quality control is made by humans. Combination of technol- ogy and human care is particularly important for safety-critical ac- tivities, such as autonomous driv- ing,” Kessler says. The labelings, also called anno- tations, in the image and video files have to agree with the real environment with pixel accu- racy. The better the quality of the processed image data, the better is the algorithm that uses these data for training. “As train- ing images cannot be supplied for all situations, such as accidents, we now also offer simulations based on real data,” Kessler says. Although he focuses on autono- mous driving, Kessler also plans to process image data for train- ing algorithms to detect tumors or to evaluate aerial photos in the future. Using processed images, algorithms learn to recognize the real en- vironment for autono- mous driving. Photo: understand.ai Verschiedene Partner aus dem Schweizer und dem europäischen Logistikumfeld treten gemeinsam auf. Gelebte Innovationskraft und gebündeltes Mobilitäts- und Transport-Know-how vereint in einem Auftritt. Mit von der Partie sind SBB Cargo, die Universität St. Gallen (HSG), TR Trans Rail, die Hörmann Gruppe, Transwaggon, Siemens Mobility, Voith, MEV Schweiz sowie Innofreight und ACTS. UNTER EINEM DACH: SWISSMOVERS BEWEGEN DIE SCHWEIZ transport logistic 2019 in München: Besuchen Sie uns im Aussenbereich, Stand FGL 804/1. www.swissmovers.org