Driving tests using actual vehicles have been conducted in order to verify automated vehicles. However, there are countless combinations of natural phenomena and traffic flows, and it is required to reduce the time and cost for collecting verification data and improve the efficiency. Therefore, Mitsubishi Heavy Industries, Ltd. (MHI) has developed Artificial Intelligence (AI) algorithm (related patents application) to convert a real image taken in a certain environment into an image of a different environment using deep learning for camera images, which play an important role in object recognition, in collaboration with the University of Tokyo. In this algorithm, semantic segmentation technology is combined with the conventional image conversion technology to improve the conversion accuracy of details. As a result, a large amount of data required for verification of automated vehicles can be generated in a short time, and the time and cost for collecting verification data can be reduced.
