Michigan researchers have created a new 3D dataset that could transform how household robots understand their surroundings. Called 3D-GRAND, this AI-generated collection includes over 40,000 virtual rooms with millions of detailed descriptions, helping robots interpret 3D spaces using text. Early tests show it outperforms previous models by a wide margin. Find out more about this breakthrough and the team presenting their work at the upcoming CVPR Conference in the full article.
