Diffusion-based point cloud generation with smoothness constraints

Yukun Li, Liping Liu

Abstract: Diffusion models have been popular for point cloud generation tasks. Existing works utilize the forward diffusion process as a discrete Markov Chain to convert the original point distribution into a noise distribution (e.g., standard Gaussian distribution) and learn the reverse diffusion process to recover the target point distribution from the noise distribution. However, the diffusion process can produce samples with non-uniform points on the surface without consideration of the point cloud geometric feature. To alleviate the problem, we propose a novel diffusion-based framework for point cloud generation and incorporate the local smoothness constraint into the generation process. Experiments demonstrate that the proposed model is not only capable of generating realistic shapes but also generating more uniform point clouds, outperforming multiple state-of-the-art methods.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Applications (eg, speech processing, computer vision, NLP)