Deep Point Cloud Normal Estimation via Triplet Learning (Demonstration)

Abstract

In this demonstration paper, we show the technical details of our proposed triplet learning-based point cloud normal estimation method. Our network architecture consists of two phases: (a) feature encoding to learn representations of local patches, and (b) normal estimation that takes the learned representations as input to regress normals. We are motivated that local patches on isotropic and anisotropic surfaces respectively have similar and distinct normals, and these separable representations can be learned to facilitate normal estimation. Experiments show that our method preserves sharp features and achieves good normal estimation results especially on computer-aided design (CAD) shapes.