Kakao Mobility Unveils Complete Autonomous Driving AI Dataset

## Kakao Mobility Accelerates Future Mobility Innovation by Releasing AI Autonomous Driving Dataset Amidst intense competition in autonomous driving technology, a core driver of future mobility, Kakao Mobility is injecting new vitality into...

admin

Feb 28, 2025 - 00:00

0 712

## Kakao Mobility Accelerates Future Mobility Innovation by Releasing AI Autonomous Driving Dataset Amidst intense competition in autonomous driving technology, a core driver of future mobility, Kakao Mobility is injecting new vitality into the domestic autonomous driving ecosystem by releasing valuable resources for the development of AI-based autonomous driving technology. This 'AI Training Autonomous Driving Dataset,' released through the Korea Electronics Technology Institute (ETRI)'s 'AI Nanum' platform, is expected to lay the groundwork for technological commercialization and drive the growth of the overall related industry. **National-Level Strategic Vision, Realized by Kakao Mobility's Technology** This dataset release is one of the key achievements of the , ambitiously promoted by the Ministry of Science and ICT and the Korea Agency for Driving Innovation (KADIF). Kakao Mobility has participated in this significant national project, making diverse efforts with the goal of implementing an advanced 'Level 4 (Lv.4)' autonomous driving system. Specifically, by successfully developing technology that efficiently generates, systematically manages, and automatically distributes vast amounts of 'fused autonomous driving data' collected from real road environments to where it's needed, the company has laid the foundation for data-driven autonomous driving technology innovation. **The Value of Differentiated Data, Optimized for Domestic Road Environments** The dataset presented by Kakao Mobility holds special value because it is optimized for domestic road environments more than any other data. This is because it is composed of vivid information acquired through various sensors installed along major domestic roads and autonomous vehicles directly operated by Kakao Mobility that have traversed real roads. A vast amount of 150,000 cases across 10 types of data contains essential elements for autonomous driving AI models to learn complex and diverse scenarios they may encounter in real driving situations. This data contributes to enhancing AI's perception and judgment capabilities in two key areas. First, it provides accurate perception capabilities for '3D dynamic objects' that move constantly, such as people, vehicles, and bicycles. This plays a critical role in enabling autonomous vehicles to understand their surroundings three-dimensionally, preemptively detect potential hazards, and plan driving routes safely and flexibly. Second, it assists in the accurate recognition and judgment of '2D static objects' that are fixed but essential for driving, such as traffic lights and road signs. The ability to quickly and accurately interpret numerous pieces of information in a complex urban environment is a core competence that determines the reliability of autonomous driving systems. **Significant Leap in AI Performance Proven by Data** The true value of a dataset is revealed when it is applied to actual AI models. ETRI's meticulous validation results clearly demonstrate the innovative changes brought about by this dataset. AI models utilizing this data showed an improvement in object detection and recognition performance by a remarkable 2-8% compared to existing models. This goes beyond a mere numerical improvement, signifying a groundbreaking increase in the reliability and safety of autonomous driving systems in real-world environments they may encounter. Particularly noteworthy is the fact that this dataset showed excellent effectiveness in improving AI performance for 'sparse data'. AI perception capabilities were significantly enhanced in specific situations with high actual accident risk, such as urban night traffic congestion or pedestrian traffic lights, which are relatively rare in general driving data. The ability to respond to these 'Edge Cases' is considered the final puzzle piece for the commercialization of autonomous driving technology, proving that this dataset played a decisive role in elevating the level of domestic autonomous driving technology by one step. **The Future of Autonomous Driving to Be Paved Through Openness and Sharing** Jang Seong-wook, Head of Kakao Mobility's Future Mobility Research Lab, stated that this dataset release is "expected to become a cornerstone for accelerating the commercialization and development of domestic autonomous driving technology." This demonstrates Kakao Mobility's firm commitment to contributing to the growth of the overall domestic autonomous driving industry, beyond mere technology development. Jeong Gwang-bok, Head of KADIF, also emphasized the importance of open innovation through data sharing, stating, "We hope it will serve as a stepping stone for the growth of related academia and startups, and further contribute to the advancement of AI autonomous driving technology." Kakao Mobility's release of this AI training autonomous driving dataset will be a significant milestone towards a future where domestic autonomous driving technology moves beyond simple research and safely and conveniently integrates into citizens' lives on real roads. The advancement of AI technology, based on rich and precise data, is expected to greatly contribute to building a safer and more efficient mobility environment, and further strengthen South Korea's future industrial competitiveness. Kakao Mobility will continue to open new horizons in the era of autonomous driving through continuous technological innovation and openness.