Reconstruction-Aware Prior Distillation for Semi-supervised Point Cloud Completion

TL;DR

RaPD is a semi-supervised point cloud completion method that reduces reliance on large paired datasets by learning semantic priors and employing a two-stage training process, improving performance on incomplete real-world data.

Contribution

Introduces RaPD, a novel semi-supervised approach for point cloud completion that leverages unpaired data and a two-stage training scheme to reduce dataset annotation requirements.

Findings

RaPD outperforms previous methods on multiple datasets.

Effective use of unpaired data improves completion quality.

Reduces need for large paired datasets in real-world scenarios.

Abstract

Real-world sensors often produce incomplete, irregular, and noisy point clouds, making point cloud completion increasingly important. However, most existing completion methods rely on large paired datasets for training, which is labor-intensive. This paper proposes RaPD, a novel semi-supervised point cloud completion method that reduces the need for paired datasets. RaPD utilizes a two-stage training scheme, where a deep semantic prior is learned in stage 1 from unpaired complete and incomplete point clouds, and a semi-supervised prior distillation process is introduced in stage 2 to train a completion network using only a small number of paired samples. Additionally, a self-supervised completion module is introduced to improve performance using unpaired incomplete point clouds. Experiments on multiple datasets show that RaPD outperforms previous methods in both homologous and heterologous scenarios.