Open3D: A Modern Library for 3D Data Processing
By Qian-Yi Zhou, Jaesik Park, Vladlen Koltun. arXiv.org, 2018
2,171 citations recorded by the source corpus.
Open3D is an open-source library that supports rapid development of software that deals with 3D data. The Open3D frontend exposes a set of carefully selected data structures and algorithms in both C++ and Python. The backend is highly optimized and is set up for parallelization. Open3D was developed from a clean slate with a small and carefully considered set of dependencies. It can be set up on different platforms and compiled from source with minimal effort. The code is clean, consistently styled, and maintained via a clear code review mechanism. Open3D has been used in a number of published research projects and is actively deployed in the cloud. We welcome contributions from the open-source community.
Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Geology
View Paper: https://www.semanticscholar.org/paper/2880ac931c11176aee6d42a7e7bb0703aacde3f9
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Aerospace Engineering
- Andreas Geiger
- Vladlen Koltun
- Christopher Bongsoo Choy
- Luca Carlone
- Michael J. Black
- Marc Pollefeys
- Raquel Urtasun
- Sanja Fidler
In Computer Vision and Pattern Recognition
- Luc Van Gool
- Jitendra Malik
- Dacheng Tao
- Xiaogang Wang
- Ming-Hsuan Yang
- Trevor Darrell
- Andreas Geiger
- Andrea Vedaldi
In Geology
- Vladlen Koltun
- Christopher Bongsoo Choy
- Andreas Geiger
- Wei Dong
- Yue Wang
- Justin M. Solomon
- Luca Carlone
- Jaesik Park
Citations in the corpus
- Kimera-Multi: Robust, Distributed, Dense Metric-Semantic SLAM for Multi-Robot Systems (2021, methodology): “The resulting semantically-labeled point clouds are then registered using the ICP [93] implementation in Open3D [94].”
- TEASER: Fast and Certifiable Point Cloud Registration (2020, methodology): “[122] Q.-Y. Zhou, J. Park, and V. Koltun, “Open3D: A modern library for 3D data processing,” 2018, arXiv:1801.09847.”
- Fully Convolutional Geometric Features (2019, methodology): “For all experiments, we used RANSAC [9] with early termination [38]. with RANSAC in Tab.”
- Tangent Convolutions for Dense Prediction in 3D (2018, positive): “All precomputations are implemented using Open3D [61].”
- RPM-Net: Robust Point Matching Using Learned Features (2020, methodology): “We use the implementations of ICP and FGR in Intel Open3D [44], and our own implementation of RPM.”
- 2D Gaussian Splatting for Geometrically Accurate Radiance Fields (2024, neutral): “We then analyze the contribution of the proposed components.”
- Deep Global Registration (2020, methodology): “Recall TE (cm) RE (deg) Time (s) FGR [50] 0.2% 40.7 1.02 1.42 RANSAC [35] 34.2% 25.9 1.39 1.37 FCGF [9] 98.2%”
- Resolving 3D Human Pose Ambiguities With 3D Scene Constraints (2019, methodology): “For each sequence a human annotator annotates 3 correspondences between the 3D scene Ms and the point cloud P t to get an initial estimate of Tc, which is then refined using ICP [9, 75].”
- Kimera: an Open-Source Library for Real-Time Metric-Semantic Localization and Mapping (2019, methodology): “Moreover, we provide Jupyter Notebooks to visualize intermediate VIO statistics ( e.g., quality of the feature tracks, IMU preintegration errors), as well as to automatically assess the quality of the 3D reconstruction using Open3D [72].”
- NeRD: Neural Reflectance Decomposition from Image Collections (2020, methodology): “For meshing, we use the Open3D [73] implementation of the Poisson surface reconstruction algorithm [29] using the normals from NeRD.”
- A comprehensive survey on point cloud registration (2021, neutral): “Average Recall Thresholds ICP(p2point)[120] 6.04 TE(0.3m),RE(15 ◦ ) ICP(p2plane)[120] 6.59 TE(0.3m),RE(15 ◦ ) Super4PCS[68] 21.6 TE(0.3m),RE(15 ◦ ) GO-ICP[106] 22.9 TE(0.3m),RE(15 ◦ ) FGR[119] 42.7 TE(0.3m),RE(15 ◦ ) RANSAC [86] 66.1 TE(0.3m),RE(15 ◦ ) SpinImage [46] 34 rmseP( 0 .”
- UniSim: A Neural Closed-Loop Sensor Simulator (2023, methodology): “We first estimate per-point normals from 200 nearest neighbors, and then we downsample the points into 4 cm 3 voxels [103].”
- HOI4D: A 4D Egocentric Dataset for Category-Level Human-Object Interaction (2022, methodology): “In the process of static scene segmentation labeling, given an original RGB-D video with 2D motion segmentation masks, we remove all of the masked objects from single frame point clouds and then reconstruct the static point cloud [10,53] of the indoor scene containing only objects whose position has”
- GET3D: A Generative Model of High Quality 3D Textured Shapes Learned from Images (2022, methodology): “To compute the metrics based on LFD, we convert the output point clouds (10k points) to a mesh representation using Open3D [37] implementation of Poisson surface reconstruction [20].”
- Grasp'D: Differentiable Contact-rich Grasp Synthesis for Multi-fingered Hands (2022, methodology): “We create a Poisson reconstruction [53] of this point cloud using the Open3D library [96] with a depth of 5.”
- D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features (2020, methodology): “During testing, instead of using voxel downsample, we use uniform downsample i.e. uniform down sample in Open3D [39] implementation, by which the density variations of input point clouds is enlarged.”
- KISS-ICP: In Defense of Point-to-Point ICP – Simple, Accurate, and Robust Registration If Done the Right Way (2022, background): “Most voxelization approaches, however, select the center of each occupied voxel to downsample the point cloud [25], [36].”