Naver Labs has unveiled its artificial intelligence (AI) models “Dust3R-2” and “ANNY,” which further..
The 4th ‘AI for Robotics’ workshop will be held\nDuster 2′, which recognizes space and people in 3D, and \n3D body model ‘Annie’ will be introduced…”Leading the research on robot AI”
Naver Labs has unveiled its artificial intelligence (AI) models “Dust3R-2” and “ANNY,” which further expand the robot’s space and human understanding ability.
Naver Labs Europe announced on the 24th that it held the 4th “AI for Robotics” workshop for two days from the 20th to the 21st (local time) and introduced this content.
The biennial workshop from 2019 brings together researchers and experts from robotics and AI fields from around the world to discuss the present and future of AI technology for robots to understand and interact with the world like humans.
This year’s workshop was held under the theme of “Spatial AI.” More than 150 researchers, including Professor Andrew Davidson of Imperial College London, who first proposed the concept of spatial intelligence, attended and shared the latest research.
![Martin Humenberger, head of Naver Labs Europe, introduces the direction of spatial intelligence research of Naver Labs at the AI for Robotics workshop. [Picture = Naver]](https://wimg.mk.co.kr/news/cms/202511/24/news-p.v1.20251124.489ca8a8f71b455bba170af7ce472938_P1.png)
The next-generation model ‘Duster 2’ introduced this time is a version that integrates various derivative studies of Duster and considers commercialization. First unveiled in December 2023, Duster is a 3D vision model that reconstructs space in 3D with just one photo. Applying this to robots allows robots to quickly understand new spaces in three dimensions without pre-manufactured maps.
The 3D body model “Annie” introduced at the same time is a model that can express the human body in detail with various parameters. It uses anthropometric data from the MakeHuman community and the World Health Organization (WHO) demographics to reflect various human characteristics around the world, from children to the elderly, without personal information infringement. The model was released as an open source on the 6th of last month and is contributing to revitalizing the robot research ecosystem.
Naver Labs Europe, located in Grenoble, France, is leading spatial intelligence research with researchers from 26 countries around the world. It is evaluated as one of the first research institutes in the world to apply the foundation model to the field of robot decision optimization such as path search and task allocation.
Currently, robots are studying a “universal foundation model” that integrates various foundation models to understand the world. Through this, the goal is to allow robots to understand the social context and respond flexibly to complex environmental changes.
Naver Labs has been focusing on spatial intelligence and physical AI research since 2016, and has accumulated unprecedented experience in demonstrating technology and applying it to actual services on domestic and international testbeds, including Naver 1784 in Seongnam, Gyeonggi Province, said Lee Dong-hwan, leader of Naver Labs, who participated in the event. “We will continue to focus more on AI research and development for robots while actively sharing research results on the global stage.”
link
