The Ministry of Science and ICT said on Friday it had published a guideline and specification document allowing industry, academia and research institutes to jointly build and share training data for end-to-end (E2E) self-driving AI.
The global autonomous-driving industry is shifting rapidly toward the E2E approach, in which a single AI trained on vast volumes of data handles perception, judgment and control as one integrated process.
Leaders such as Waymo in the United States and Baidu in China have been expanding road testing and racing to amass ever-larger training datasets.
South Korean firms, by contrast, have built their data in isolation, leaving it hard to share because sensor placement and other specifications differ from vehicle to vehicle.
The new guideline covers the full data lifecycle, defining procedures for collection, processing, alignment, correction and labelling, and setting out sensor configurations, storage formats and methods for verifying raw data.
It was developed through the multi-ministry project for autonomous driving technology innovation, led by the Institute of Information & Communications Technology Planning & Evaluation, the autonomous-driving project group KADIF and the Electronics and Telecommunications Research Institute.
The ministry said it would work with the land ministry to deploy the standards in self-driving test cities building city-scale E2E datasets, refining the guideline as it goes.
Copyright ⓒ Aju Press All rights reserved.