InteriorGS dataset from Qunhe Technology tops the global open-source list.

Reporters learned that recently, the open-source dataset InteriorGS released by Qunhe Technology has topped the trend list of the world's largest AI open-source community HuggingFace. The Nemotron-Post dataset launched by NVIDIA closely follows, and another 3D dataset InteriorAgent open-sourced by Qunhe has also made the list. It is understood that the 3D Gaussian semantic dataset InteriorGS, which topped the list for the first time, introduces 3D Gaussian technology into the field of AI space training, and combines its self-developed spatial large model capabilities, making the dataset both realistic and semantic, becoming the world's first large-scale 3D dataset suitable for intelligent agents to freely move within. The InteriorGS dataset consists of 1000 Gaussian scenes, covering more than 80 environmental types, with over 554,000 object labels belonging to 755 categories. Each object is equipped with a 3D bounding box and semantic annotation.