With the capacity of modeling long-range dependencies in sequential data, transformers have shown remarkable performances in a variety of generative tasks such as image, audio, and text generation. Yet, taming them in generating less structured and voluminous data formats such as high-resolution point clouds have seldom been explored due to ambiguous sequentialization processes and infeasible computation burden. In this paper, we aim to further exploit the power of transformers and employ them for the task of 3D point cloud generation. The key idea is to decompose point clouds of one category into semantically aligned sequences of shape compositions, via a learned canonical space. These shape compositions can then be quantized and used to learn a context-rich composition codebook for point cloud generation. Experimental results on point cloud reconstruction and unconditional generation show that our model performs favorably against state-of-the-art approaches. Furthermore, our model can be easily extended to multi-modal shape completion as an application for conditional shape generation. The source code and trained models can be found at https://github.com/AnjieCheng/CanonicalVAE.
|Title of host publication||Computer Vision – ECCV 2022 - 17th European Conference, Proceedings|
|Editors||Shai Avidan, Gabriel Brostow, Moustapha Cissé, Giovanni Maria Farinella, Tal Hassner|
|Publisher||Springer Science and Business Media Deutschland GmbH|
|Number of pages||16|
|Publication status||Published - 2022|
|Event||17th European Conference on Computer Vision, ECCV 2022 - Tel Aviv, Israel|
Duration: 2022 Oct 23 → 2022 Oct 27
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||17th European Conference on Computer Vision, ECCV 2022|
|Period||22/10/23 → 22/10/27|
Bibliographical noteFunding Information:
Acknowledgments. The MOST, Taiwan under Grants 110-2634-F-002-051, MOST Joint Research Center for AI Technology, All Vista Healthcare, and NSF CAREER grant 1149783. We thank National Center for High-performance Computing (NCHC) for providing computational and storage resources.
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
All Science Journal Classification (ASJC) codes
- Theoretical Computer Science
- Computer Science(all)