ABO: Dataset and Benchmarks for Real-World 3D Object Understanding

TL;DR

ABO is a comprehensive dataset combining real-world images, metadata, and detailed 3D models of household objects, enabling advancements in 3D reconstruction, material estimation, and object retrieval.

Contribution

This paper introduces ABO, a large-scale dataset with complex geometries and materials, along with benchmarks for real-world 3D object understanding tasks.

Findings

Current state-of-the-art methods are challenged by ABO's complexity.

Benchmarks reveal gaps in 3D reconstruction and material estimation.

ABO facilitates progress in bridging real and virtual 3D understanding.

Abstract

We introduce Amazon Berkeley Objects (ABO), a new large-scale dataset designed to help bridge the gap between real and virtual 3D worlds. ABO contains product catalog images, metadata, and artist-created 3D models with complex geometries and physically-based materials that correspond to real, household objects. We derive challenging benchmarks that exploit the unique properties of ABO and measure the current limits of the state-of-the-art on three open problems for real-world 3D object understanding: single-view 3D reconstruction, material estimation, and cross-domain multi-view object retrieval.