Hardware Design and Accurate Simulation of Structured-Light Scanning for Benchmarking of 3D Reconstruction AlgorithmsDownload PDF

18 Aug 2021, 19:50 (modified: 15 Jan 2022, 20:43)NeurIPS 2021 Datasets and Benchmarks Track (Round 2)Readers: Everyone
Keywords: Structured-Light Scanning, Structured Light Scanning, 3D Scanning, Surface Reconstruction, 3D Scanning Benchmark, Range Scan Completion, Shape Completion, Scan Denoising
TL;DR: We co-develop a 3D structured-light scanning hardware setup together with a corresponding light transport simulation. This provides an ideal framework for developing and benchmarking data-driven algorithms in the area of 3D reconstruction.
Abstract: Images of a real scene taken with a camera commonly differ from synthetic images of a virtual replica of the same scene, despite advances in light transport simulation and calibration. By explicitly co-developing the Structured-Light Scanning (SLS) hardware and rendering pipeline we are able to achieve negligible per-pixel difference between the real image and the synthesized image on geometrically complex calibration objects with known material properties. This approach provides an ideal test-bed for developing and evaluating data-driven algorithms in the area of 3D reconstruction, as the synthetic data is indistinguishable from real data and can be generated at large scale by simulation. We propose three benchmark challenges using a combination of acquired and synthetic data generated with our system: (1) a denoising benchmark tailored to structured-light scanning, (2) a shape completion benchmark to fill in missing data, and (3) a benchmark for surface reconstruction from dense point clouds. Besides, we provide a large collection of high-resolution scans that allow to use our system and benchmarks without reproduction of the hardware setup on our website: https://geometryprocessing.github.io/scanner-sim
Supplementary Material: zip
URL: https://geometryprocessing.github.io/scanner-sim
Contribution Process Agreement: Yes
Dataset Url: https://geometryprocessing.github.io/scanner-sim
License: Our datasets are available under the CC BY 4.0 license except for the 3D models of 7 colored 3D printed objects which are licensed by their respective creators under various Creative Commons licenses as listed above. Our code is published partly under the BSD 3-Clause Clear license and partly under the GPL 3.0 license. The parts are clearly separated in our code repository and marked with the respective license. The rendering engine we used and modified is GPL-licensed, therefore the simulation framework is licensed under GPL and the calibration and physical scanner software is licensed under BSD 3-Clause Clear.
Author Statement: Yes
10 Replies

Loading