Towards Internet-scale multi-view stereo

By Yasutaka Furukawa, Brian Curless, S. M. Seitz, Richard Szeliski. 2010

868 citations recorded by the source corpus.

Subfields: Computer Graphics and Computer-Aided Design, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Graphics and Computer-Aided Design

In Computer Vision and Pattern Recognition

Citations in the corpus

Deep Level Sets: Implicit Surface Representations for 3D Shape Inference (2019, background): “problems. 1. Introduction In recent years, the ﬁeld of 3D reconstruction has achieved great progress trying to tackle many categories of problems such as structure from motion [14], multi-view stereo [11] and reconstruction from a single image [5]. The application domain includes, but is not limited”
Unsupervised Learning of Depth and Ego-Motion from Video (2017, background): “Structure from motion The simultaneous estimation of structure and motion is a well studied problem with an established toolchain of techniques [12, 50, 38].”
Fast Single Shot Detection and Pose Estimation (2016, methodology): “We are able to cheaply and easily collect labels for all images by ﬁrst sparsely reconstructing each scene using COLMAP [16], and then densely reconstructing the scene using CMVS [4, 5].”
High Accuracy and Visibility-Consistent Dense Multiview Stereo (2012, neutral): “Furukawa et al. [17] partitioned the cameras in view clusters to run a multiview stereo of choice for each cluster.”
First-Person Activity Forecasting with Online Inverse Reinforcement Learning (2016, methodology): “Figure 2: Sparse SLAM points (2a) and offline dense reconstruction (2b) using [5] for two of our dataset environments.”
Building Rome on a Cloudless Day (2010, methodology): “outperforms [4,5] by running on a single PC within less than two hours for 64,000 images.”
Straightforward reconstruction of 3D surfaces and topography with a camera: Accuracy and geoscience application (2012, methodology): “To permit matching with large image collections, a pre-processing step, CMVS (Clustering Views for MultiView Stereo) [ Furukawa et al. , 2010] is integrated into the software.”
Creating Large-Scale City Models from 3D-Point Clouds: A Robust Approach with Hybrid Representation (2012, neutral): “Another interesting challenge is to adapt our approach to point clouds generated from internet photo collections (Agar-wal et al., 2009; Frahm et al., 2010; Furukawa et al., 2010) which contain more outliers and have spatial distributions highly heterogeneous.”
Let There Be Color! Large-Scale Texturing of 3D Reconstructions (2014, background): “[8] produce and render point clouds with very limited resolution, which is especially apparent in close-ups.”
Block-NeRF: Scalable Large Scene Neural View Synthesis (2022, background): “This process presents its own scaling difﬁculties [17].”
Maplab: An Open Framework for Research in Visual-Inertial Mapping and Localization (2017, methodology): “Also includes an interface to CMVS/PMVS2 [39].”
A review of 3D reconstruction techniques in civil engineering and their applications (2018, background): “CMVS removes the redundant images of the scene and clusters the resulting images to generate clustered images [29,68].”
Dense Object Reconstruction with Semantic Priors (2013, neutral): “the drawbacks of traditional multiview stereo by incorporating semantic information in the form of learned category-level shape priors and object detection.”
VastGaussian: Vast 3D Gaussians for Large Scene Reconstruction (2024, neutral): “The following works [17, 19] produce a dense point cloud or triangle mesh from the SfM output based on multi-view stereo (MVS).”
BigBIRD: A large-scale 3D database of object instances (2014, neutral): “3D model construction [16], [17], [18].”
DeepMVS: Learning Multi-view Stereopsis (2018, background): “ and photometric error measures. Recent advances include robust neighbor view selection [24], incorporation of visibility consistency [10], and clustering-based techniques for efﬁcient reconstruction [13,8]. Recently, Schonberger et al. present a MVS system — COLMAP [¨ 33] — that jointly estimates d”
Elastic Fragments for Dense Scene Reconstruction (2013, background): “Substantial progress on dense scene reconstruction from photographs and video sequences has been made, despite the ambiguity of photometric cues [20, 26, 21, 6, 15, 17].”
State of the art in high density image matching (2014, methodology): “…advanced by recent developments in computer vision (Hartley and Zisserman, 2003; Goesele et al., 2007; Pollefeys et al., 2008; Snavely et al., 2008; Furukawa et al., 2010; Shan et al., 2013), is presently offering additional automated procedures for both image orientation (Remondino et al., 2012)…”