Naver logo. /Courtesy of Naver/News1 © News1 reporter Park Jun-bae

Naver Labs Europe unveiled next-generation AI models that expand robots' spatial understanding, "DUSt3R-2" and "ANNY," at an international conference in Grenoble, France. Naver Labs Europe said on the 24th that it hosted the "AI for Robotics" workshop on the 20th and 21st (local time) and showcased the two models.

First unveiled at the end of 2023, "DUSt3R" is a vision AI model that reconstructs 3D space from a single photo. It is evaluated as technology that enables robots to perceive their surroundings stereoscopically and move autonomously without maps. The new "DUSt3R-2" integrates derivative research to date and is a next-generation version aimed at commercialization, enabling even more precise 3D reconstruction in complex spaces.

Another new technology, "ANNY," is a 3D body model that can represent the human body in detail. Based on open-source human body data and WHO population statistics, it reflects diverse body types and racial characteristics from children to older adults while protecting personal information.

The workshop also discussed how spatial intelligence can advance robots' abilities to perceive, decide, and act. Lee Dong-hwan of Naver Labs said, "We will actively share our research achievements on the global stage and expand the AI ecosystem for robots."

※ This article has been translated by AI. Share your feedback here.