Objaverse-XL: A Universe of 10M+ 3D Objects

TL;DR

Objaverse-XL is a large-scale, diverse 3D object dataset with over 10 million items, designed to advance 3D vision tasks by enabling training on extensive data, leading to improved zero-shot generalization.

Contribution

This work introduces Objaverse-XL, the largest and most diverse 3D object dataset to date, facilitating significant progress in 3D vision through large-scale training.

Findings

Training Zero123 on Objaverse-XL improves zero-shot generalization.

Utilizing over 100 million multi-view images enhances 3D model performance.

The dataset enables new possibilities for 3D vision research.

Abstract

Natural language processing and 2D vision models have attained remarkable proficiency on many tasks primarily by escalating the scale of training data. However, 3D vision tasks have not seen the same progress, in part due to the challenges of acquiring high-quality 3D data. In this work, we present Objaverse-XL, a dataset of over 10 million 3D objects. Our dataset comprises deduplicated 3D objects from a diverse set of sources, including manually designed objects, photogrammetry scans of landmarks and everyday items, and professional scans of historic and antique artifacts. Representing the largest scale and diversity in the realm of 3D datasets, Objaverse-XL enables significant new possibilities for 3D vision. Our experiments demonstrate the improvements enabled with the scale provided by Objaverse-XL. We show that by training Zero123 on novel view synthesis, utilizing over 100 million multi-view rendered images, we achieve strong zero-shot generalization abilities. We hope that releasing Objaverse-XL will enable further innovations in the field of 3D vision at scale.