PCN: Point Completion Network

TL;DR

The paper introduces PCN, a novel deep learning model that directly completes partial 3D point clouds into dense, realistic shapes without relying on shape assumptions or annotations, improving shape completion in vision and robotics.

Contribution

PCN is the first method to operate directly on raw point clouds for shape completion without structural assumptions or annotations, enabling fine-grained and robust 3D reconstructions.

Findings

PCN produces dense, complete point clouds with realistic structures.

It performs well on various levels of incompleteness and noise.

Effective on LiDAR scans from the KITTI dataset.

Abstract

Shape completion, the problem of estimating the complete geometry of objects from partial observations, lies at the core of many vision and robotics applications. In this work, we propose Point Completion Network (PCN), a novel learning-based approach for shape completion. Unlike existing shape completion methods, PCN directly operates on raw point clouds without any structural assumption (e.g. symmetry) or annotation (e.g. semantic class) about the underlying shape. It features a decoder design that enables the generation of fine-grained completions while maintaining a small number of parameters. Our experiments show that PCN produces dense, complete point clouds with realistic structures in the missing regions on inputs with various levels of incompleteness and noise, including cars from LiDAR scans in the KITTI dataset.