Deep Point Cloud Reconstruction

Abstract

Point cloud obtained from 3D scanning is often sparse, noisy, and irregular. To cope with these issues, recent studies have been separately conducted to densify, denoise, and complete inaccurate point cloud. In this paper, we advocate that jointly solving these tasks leads to significant improvement for point cloud reconstruction. To this end, we propose a deep point cloud reconstruction network consisting of two stages: 1) a 3D sparse stacked-hourglass network as for the initial densification and denoising, 2) a refinement via transformers converting the discrete voxels into 3D points. In particular, we further improve the performance of transformer by a newly proposed module called amplified positional encoding. This module has been designed to differently amplify the magnitude of positional encoding vectors based on the points' distances for adaptive refinements. Extensive experiments demonstrate that our network achieves state-of-the-art performance among the recent studies in the ScanNet, ICL-NUIM, and ShapeNetPart datasets. Moreover, we underline the ability of our network to generalize toward real-world and unmet scenes.

arXiv commentICLR 2022 accepted

References in corpus (8)

• Adam: A Method for Stochastic Optimization
• Open3D: A Modern Library for 3D Data Processing
• NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis
• ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes
• Submanifold Sparse Convolutional Networks
• On the Spectral Bias of Neural Networks
• DPSNet: End-to-end Deep Plane Sweep Stereo
• Total Denoising: Unsupervised Learning of 3D Point Cloud Cleaning

Cited by in corpus (2)

• SuperPC: A Single Diffusion Model for Point Cloud Completion, Upsampling, Denoising, and Colorization
• PointMixer: MLP-Mixer for Point Cloud Understanding