Cascaded Refinement Network for Point Cloud Completion

TL;DR

This paper introduces a cascaded refinement network with a coarse-to-fine approach for detailed 3D point cloud completion, effectively preserving existing details and generating missing parts with high fidelity.

Contribution

It proposes a novel cascaded refinement network and patch discriminator for improved point cloud completion, capturing local details and complex distributions more effectively.

Findings

Outperforms state-of-the-art methods on multiple datasets

Preserves existing partial details accurately

Generates high-fidelity missing parts

Abstract

Point clouds are often sparse and incomplete. Existing shape completion methods are incapable of generating details of objects or learning the complex point distributions. To this end, we propose a cascaded refinement network together with a coarse-to-fine strategy to synthesize the detailed object shapes. Considering the local details of partial input with the global shape information together, we can preserve the existing details in the incomplete point set and generate the missing parts with high fidelity. We also design a patch discriminator that guarantees every local area has the same pattern with the ground truth to learn the complicated point distribution. Quantitative and qualitative experiments on different datasets show that our method achieves superior results compared to existing state-of-the-art approaches on the 3D point cloud completion task. Our source code is available at https://github.com/xiaogangw/cascaded-point-completion.git.