Objaverse

Objaverse is a large dataset of high-fidelity 3D objects introduced by researchers from UT Austin, NVIDIA, and others in 2022. It contains 800K+ annotated 3D models covering over 10 million object annotations across diverse semantic categories. Each object includes multiple 3D file formats (GLB, OBJ, USDZ), high-resolution textures, and comprehensive metadata including category labels, licensing information, source attribution, and physical properties. The dataset aggregates models from various sources including Sketchfab, turbosquid, and community contributions. Objaverse has quickly become an essential resource for 3D vision, embodied AI, and simulation research. It is used by FoundationPose, OpenVLA, and various 3D foundation models for training and evaluation. The dataset is designed to fill the gap of large-scale 3D data in the AI community, analogous to the role of ImageNet in 2D vision.