NAVER LABS has secured first place in major challenges at ECCV 2024 (European Conference on Computer Vision), the world’s top computer vision conference. One of the most prestigious international conferences in the fields of computer vision and image processing, ECCV is held every two years to give top researchers from around the globe an opportunity to present their latest findings and introduce new technological trends.
At ECCV 2024, NAVER LABS demonstrated its exceptional capabilities in spatial intelligence by winning first place in two challenges: “Map-Free Visual Re-localization,” a technology that identifies locations using visual information without a map, and “Benchmark for 6D Object Pose Estimation,” which accurately estimates the 3D position and pose of objects. These achievements are a testament to the cutting-edge technology that NAVER LABS possesses in the field of spatial intelligence.
Secured First Place in the Map-Free Visual Re-localization Challenge
NAVER LABS Europe achieved top results in the Map-Free Visual Re-localization Challenge, which focuses on accurately estimating location without the use of high-precision 3D maps. The key to this achievement was NAVER’s advanced technology system “MASt3R,” which reconstructs 2D images in 3D, making it possible to understand complex environments and extract geometric information.
Visual Localization (VL) typically estimates a location in situations where a 3D map is present. However, in disaster zones or construction sites—where creating 3D maps is challenging—or in locations that are too small or fragmented for efficient pre-mapping, technology capable of recognizing locations without maps is essential. In this challenge, MASt3R demonstrated its effectiveness by quickly analyzing images and understanding complex environments to accurately estimate the location values for a single reference image even in new environments without maps.
▲ MASt3R demo for the Map-Free Visual Re-localization dataset
Ranked First for Two Consecutive Years in the Benchmark for the 6D Object Pose Estimation Challenge
NAVER LABS has won the “Benchmark for 6D Object Pose Estimation” challenge for two consecutive years. This technology is essential for smartphones and robots to accurately recognize the three-dimensional positions of objects in their surroundings.
In this challenge, which focuses on accurately determining the position of objects within images, NAVER LABS achieved first place in two categories: “Track 1: Model-Based 6D Localization of Unseen Objects,” which estimates the pose of objects that were not part of the training data, and “Track 2: Model-Based 6D Detection of Unseen Objects,” which involves detecting objects that were not used in training and then estimating their position and pose. In both challenges, NAVER LABS demonstrated world-class technical capabilities by delivering the most accurate object position measurements using only camera images while maintaining fast processing speeds.
▲ NAVER LABS technology demo for quickly measuring the position and pose of unseen objects from a single RGB image
NAVER LABS has achieved remarkable things in the field of spatial intelligence, and we gained international recognition by winning the Visual Localization Challenge at CVPR in both 2019 and 2020. To add to our achievements at this year’s ECCV, we will continue to dedicate ourselves to research in spatial intelligence that can be applied to the physical world in our everyday lives.