FoldingNet: Point Cloud Auto-encoder via Deep Grid Deformation

TL;DR

FoldingNet introduces a novel deep auto-encoder for point clouds that uses a folding-based decoder to effectively reconstruct 3D shapes from 2D grids, enhancing unsupervised learning and classification accuracy.

Contribution

The paper presents a new folding-based decoder architecture that significantly reduces parameters while improving point cloud reconstruction and representation quality.

Findings

Achieves low reconstruction errors for complex objects

Outperforms benchmarks in classification accuracy

Uses only 7% of parameters compared to fully-connected decoders

Abstract

Recent deep networks that directly handle points in a point set, e.g., PointNet, have been state-of-the-art for supervised learning tasks on point clouds such as classification and segmentation. In this work, a novel end-to-end deep auto-encoder is proposed to address unsupervised learning challenges on point clouds. On the encoder side, a graph-based enhancement is enforced to promote local structures on top of PointNet. Then, a novel folding-based decoder deforms a canonical 2D grid onto the underlying 3D object surface of a point cloud, achieving low reconstruction errors even for objects with delicate structures. The proposed decoder only uses about 7% parameters of a decoder with fully-connected neural networks, yet leads to a more discriminative representation that achieves higher linear SVM classification accuracy than the benchmark. In addition, the proposed decoder structure is shown, in theory, to be a generic architecture that is able to reconstruct an arbitrary point cloud from a 2D grid. Our code is available at http://www.merl.com/research/license#FoldingNet