DiffPMAE: Diffusion Masked Autoencoders for Point Cloud Reconstruction

Abstract

Point cloud streaming is increasingly getting popular, evolving into the norm for interactive service delivery and the future Metaverse. However, the substantial volume of data associated with point clouds presents numerous challenges, particularly in terms of high bandwidth consumption and large storage capacity. Despite various solutions proposed thus far, with a focus on point cloud compression, upsampling, and completion, these reconstruction-related methods continue to fall short in delivering high fidelity point cloud output. As a solution, in , we propose an effective point cloud reconstruction architecture. Inspired by self-supervised learning concepts, we combine Masked Autoencoder and Diffusion Model to remotely reconstruct point cloud data. By the nature of this reconstruction process, can be extended to many related downstream tasks including point cloud compression, upsampling and completion. Leveraging ShapeNet-55 and ModelNet datasets with over 60000 objects, we validate the performance of exceeding many state-of-the-art methods in terms of autoencoding and downstream tasks considered. Our source code is available at : https://github.com/TyraelDLee/DiffPMAE