Game developers now have access to a vast collection of free 3D models through Objaverse, a new resource. Objaverse is a colossal compilation of 3D models, predominantly obtained from Sketchfab, with various Creative Commons licenses. Although the primary purpose of this dataset is to train AI, it is also a valuable tool for game developers.
The availability of massive datasets has facilitated recent advancements in computer vision and natural language processing. However, the existing 3D datasets, such as ShapeNet, are limited in scale and diversity, containing only 50,000 objects. To address this issue, the PRIOR team at AI2 has developed Objaverse, a vast dataset that includes nearly 1 million annotated 3D objects with descriptive captions, tags, and animations. The potential of this new dataset is enormous, as evidenced by its use in generative modeling of video-game-ready 3D assets and training household robots to understand objects as described by natural language. With its extensive and diverse 3D dataset, Objaverse offers researchers an unparalleled opportunity to advance research progress across various domains.
The Objaverse collection boasts a vast and diverse dataset of assets, created by more than 150,000 artists. These assets span across different categories such as animals, characters, and vehicles, as well as interiors and exteriors of large spaces that can be utilized for training robotic agents. Additionally, the 3D objects vary greatly in visual appearance, including high poly, scans, point clouds, and support for physically based rendering.
Asset Ovi, an alternative 3D object search engine, is constructed using the Objaverse as its foundation and supplemented with other available sources of free 3D models, including Unity Asset Store’s free assets. Find the link to Asset Ovi below.
In the video below, take a glimpse at Asset Ovi in action and explore the vast array of free 3D models available in the Objaverse.
