CanoVerse: 3D Object Scalable Canonicalization and Dataset for Generation and Pose
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866915842306867200 |
|---|---|
| author | Jin, Li Yang, Yuchen Chen, Weikai Wang, Yujie Hao, Dehao Jia, Tanghui Yin, Yingda Hu, Zeyu Zhang, Runze Luo, Keyang Yuan, Li Quan, Long Wang, Xin Qin, Xueying |
| author_facet | Jin, Li Yang, Yuchen Chen, Weikai Wang, Yujie Hao, Dehao Jia, Tanghui Yin, Yingda Hu, Zeyu Zhang, Runze Luo, Keyang Yuan, Li Quan, Long Wang, Xin Qin, Xueying |
| contents | 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 |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_07144 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | CanoVerse: 3D Object Scalable Canonicalization and Dataset for Generation and Pose Jin, Li Yang, Yuchen Chen, Weikai Wang, Yujie Hao, Dehao Jia, Tanghui Yin, Yingda Hu, Zeyu Zhang, Runze Luo, Keyang Yuan, Li Quan, Long Wang, Xin Qin, Xueying Computer Vision and Pattern Recognition 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 |
| title | CanoVerse: 3D Object Scalable Canonicalization and Dataset for Generation and Pose |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.07144 |