CanoVerse: 3D Object Scalable Canonicalization and Dataset for Generation and Pose
This work addresses the problem of arbitrary object rotations in 3D learning systems, which hinders pose-consistent generation and stable directional semantics, for researchers and developers working with 3D data.
This paper introduces CanoVerse, a large-scale canonical 3D dataset containing 320K objects across 1,156 categories, which is an order of magnitude larger than previous datasets. This dataset and its associated canonicalization framework enable improved 3D generation stability, precise cross-modal 3D shape retrieval, and zero-shot point-cloud orientation estimation for out-of-distribution data.
3D learning systems implicitly assume that objects occupy a coherent reference frame. Nonetheless, in practice, every asset arrives with an arbitrary global rotation, and models are left to resolve directional ambiguity on their own. This persistent misalignment suppresses pose-consistent generation, and blocks the emergence of stable directional semantics. To address this issue, we construct \methodName{}, a massive canonical 3D dataset of 320K objects over 1,156 categories -- an order-of-magnitude increase over prior work. At this scale, directional semantics become statistically learnable: Canoverse improves 3D generation stability, enables precise cross-modal 3D shape retrieval, and unlocks zero-shot point-cloud orientation estimation even for out-of-distribution data. This is achieved by a new canonicalization framework that reduces alignment from minutes to seconds per object via compact hypothesis generation and lightweight human discrimination, transforming canonicalization from manual curation into a high-throughput data generation pipeline. The Canoverse dataset will be publicly released upon acceptance. Project page: https://github.com/123321456-gif/Canoverse