Assets not only belong to varied categories like animals, characters, and vehicles, but also include interiors and exteriors of large spaces that can be used, for example, to train robotic agents. In total, Objaverse assets were designed by over 150K artists which makes this dataset large and diversely sourced. Objaverse provides researchers with an incredible opportunity to benefit from an extensive and diverse 3D dataset with the potential to impact research progress across domains. This new dataset’s potential is vast and has already been demonstrated through the generative modeling of video-game-ready 3D assets and through the training of household robots to understand objects as described by natural language. In response, we on the PRIOR team at AI2 have developed Objaverse, a vast dataset with nearly 1 million annotated 3D objects, complete with descriptive captions, tags, and animations. However, existing 3D datasets, such as ShapeNet, are limited in scale and diversity, consisting of only 50,000 objects. Recent advancements in computer vision and natural language processing were made possible by the availability of massive datasets. While this dataset is primarily aimed at training AI, it also serves as a useful resource for game developers. Objaverse is a gigantic dataset of 3D models, mostly sourced from Sketchfab, under a variety of Creative Commons licenses.

There is a new massive resource for game developers seeking free 3D models, Objaverse.

Free 3d models for unity