Point cloud completion via structured feature maps using a feedback network

TL;DR

This paper introduces FSNet and IFNet, novel modules for point cloud completion that leverage structured feature maps and self-correction to improve the recovery of detailed 3D structures from partial data.

Contribution

The paper presents a new structured feature map module and a self-correcting upsampling network for enhanced point cloud completion.

Findings

Outperforms state-of-the-art methods on ShapeNet, MVP, and KITTI datasets.

Effectively captures both global structure and local details.

Improves point cloud density and uniformity through self-correction.

Abstract

In this paper, we tackle the challenging problem of point cloud completion from the perspective of feature learning. Our key observation is that to recover the underlying structures as well as surface details, given partial input, a fundamental component is a good feature representation that can capture both global structure and local geometric details. We accordingly first propose FSNet, a feature structuring module that can adaptively aggregate point-wise features into a 2D structured feature map by learning multiple latent patterns from local regions. We then integrate FSNet into a coarse-tofine pipeline for point cloud completion. Specifically, a 2D convolutional neural network is adopted to decode feature maps from FSNet into a coarse and complete point cloud. Next, a point cloud upsampling network is used to generate a dense point cloud from the partial input and the coarse intermediate output. To efficiently exploit local structures and enhance point distribution uniformity, we propose IFNet, a point upsampling module with a self-correction mechanism that can progressively refine details of the generated dense point cloud. We have conducted qualitative and quantitative experiments on ShapeNet, MVP, and KITTI datasets, which demonstrate that our method outperforms state-of-theart point cloud completion approaches.