NAVER LABS Europe has released Anny, a parametric 3D body model, as an open-source project.
Anny can represent the human body using a variety of parameters such as age, height, weight, and muscle mass, and allows researchers to easily generate body shapes suited to their study needs through an intuitive slider interface.
Most existing 3D human body models have been trained on expensive scan datasets based primarily on the body types of Western adults, making it difficult to reflect diversity in age, gender, and ethnicity. In addition, because scan data is biometric information, it raises ethical and privacy concerns. Anny takes a different approach; it uses anthropometric data from the MakeHuman community and demographic statistics from the WHO, enabling it to represent a wide range of human characteristics from children to the elderly without infringing on personal data.
The accompanying Anny-One dataset includes about 800,000 synthetic human images created by combining diverse poses, clothing, lighting, and backgrounds. This helps AI vision models learn to recognize people more fairly and accurately.
A robot using the Anny body model to reconstruct 3D shapes of adults and children while navigating
NAVER LABS Europe is currently using Anny to study how robots can perceive and interact naturally with humans. By making this model open-source, the team hopes more researchers and developers in computer vision and robotics will advance technologies that help machines better understand people and apply them across fields such as robotics, vision AI, AR/VR, healthcare, and gaming.
“Technology that helps robots better understand people should be open to everyone.”
Related Articles