NAVER LABS Europe Wins Long-term Visual Localization Challenge at CVPR 2020

NAVER LABS Europe ranked first in the Autonomous Vehicle Challenge of the Long-term Visual Localization Under Changing Conditions Workshop at CVPR 2020. Moreover, NAVER LABS Europe recorded excellent results in other categories—coming in second in the Local Feature Challenge and fourth in the Handheld Devices Challenge. During this workshop, there were three categories for VL-related challenges, and NAVER LABS Europe was the only company that participated in all the tasks, displaying outstanding outcomes for each of them.

Precise location recognition technology is essential for autonomous driving vehicles and service robots, and serves as a catalyst for creating new location-based services. Visual localization (VL) technology, which enables accurate location recognition from a single photograph under various conditions such as daytime, nighttime, seasonal changes and different shooting angles, allows continuous measurement in indoor spaces and city centers dense with buildings where GPS is not available. VL also provides excellent accuracy, thereby attracting research and development in recent years throughout the world. In this regard, NAVER LABS Europe unveiled a deep learning-based algorithm named R2D2 at CVPR last year, which was appraised as a major transition in the existing VL technology. Furthermore, NAVER LABS Europe was able to meet all the different conditions and, at the same time, achieve accomplishments in a number of VL challenges at this year’s CVPR, largely because of the newly unveiled VL pipeline KAPTURE, in addition to its R2D2 technology. KAPTURE has been independently developed by NAVER LABS Europe to integrate datasets collected in various ways into a single format for efficient management.

The Long-term Visual Localization Workshop will be held online on the 14th and 15th (local time).

Related Articles