Robust reconstruction of indoor scenes

By Sungjoon Choi, Qian-Yi Zhou, Vladlen Koltun. 2015

704 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Environmental Engineering, Geology, 3D Vision & Pose

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

In Computer Vision and Pattern Recognition

In Environmental Engineering

In Geology

In 3D Vision & Pose

Citations in the corpus

ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (2017, background): “The semantic seg-mentations are more than an order of magnitude larger than any previous RGB-D dataset.”
OctNetFusion: Learning Depth Fusion from Data (2017, background): “This has led to the creation of large databases of real-world 3D content [7, 10, 42, 43], enabling progress in a variety of areas including 3D modeling [28, 54], 3D reconstruction [6, 52], 3D recognition [18, 44] and 3D scene understanding [16, 42].”
Shape Completion Using 3D-Encoder-Predictor CNNs and Shape Synthesis (2016, background): “Since the introduction of commodity range sensors such as the Microsoft Kinect, RGB-D scanning has gained a huge momentum in both ofﬂine and real-time contexts [26, 3, 28, 42, 4, 7].”
A Large Dataset of Object Scans (2016, methodology): “We also picked three interesting large objects and reconstructed them using a high-fidelity reconstruction pipeline that performs loop closure detection and global optimization [1].”
NICE-SLAM: Neural Implicit Scalable Encoding for SLAM (2021, methodology): “For reference we also show the 3D reconstruction using the offline tool Redwood [10] in Open3D [69].”
Open3D: A Modern Library for 3D Data Processing (2018, methodology): “Since there is no standard depth image format, we have implemented depth image support for multiple datasets including NYU [19], TUM [20], SUN3D [21], and Redwood [5].”
Colored Point Cloud Registration Revisited (2017, methodology): “[4] without our enhancements (referred to as CZK).”
Learning Where to Look: Data-Driven Viewpoint Set Selection for 3D Scenes (2017, background): “There is a recent explosion in generation of synthetic training data for many vision tasks including tracking, object recognition, semantic segmentation, pose estimation, and optical ﬂow among others [4, 27, 12, 14, 13, 22, 30, 33, 29].”
Scan2CAD: Learning CAD Model Alignment in RGB-D Scans (2018, background): “Very recent RGB-D reconstruction frameworks combine efficient scene representations with global pose estimation [5], and can even perform online updates with global loop closures [8].”
School of Electrical and Computer Engineering Oklahoma State University , Stillwater , OK 74078 , USA (methodology): “In contrast, the main intention of the researchers in the computer vision community is dense modeling and enhancing the quality and fidelity of the reconstructed models [22, 29, 19, 4].”
Deep Global Registration (2020, methodology): “For multi-way registration, we use the simulated Augmented ICL-NUIM dataset [6, 17] for quantitative trajectory results, and Indoor LiDAR RGB-D dataset [31] and Stanford RGB-D dataset [6] for qualitative registration visualizations.”
Online Reconstruction of Indoor Scenes from RGB-D Streams (2016, methodology): “Model Reconstruction: Upon termination of scanning, we perform pose graph optimization to refine the trajectory and handle loop closures.”
3DMatch: Learning the Matching of Local 3D Geometry in Range Scans (2016, methodology): “[6] uses a fairly liberal threshold at τ = 0.”
Deferred neural rendering (2019, background): “In particular in the context of 3D reconstruction with commodity RGB-D sensors, researchers have made significant progress, enabling robust tracking [Choi et al. 2015; Dai et al. 2017; Izadi et al. 2011; Newcombe et al. 2011; Whelan et al. 2016] and large-scale 3D scene capture [Chen et al. 2013;…”
Occupancy Anticipation for Efficient Exploration and Navigation (2020, methodology): “Noisy actuation and sensing: In the 2020 challenge, RGB-D sensing noise is simulated artificially by using a Gaussian noise-model for the RGB sensor and the Redwood noise model [3] for the depth sensor.”
NeuralRecon: Real-Time Coherent 3D Reconstruction from Monocular Video (2021, methodology): “To achieve real-time 3D reconstruction that is suitable for interactive applications, the reconstruction process needs to be incremental and the input images should be processed sequentially in local fragments [40].”
Graduated Non-Convexity for Robust Spatial Perception: From Non-Minimal Solvers to Global Outlier Rejection (2019, methodology): “nimal solvers. Indeed, only a few specialized methods for spatial perception have used GNC. Particularly, Zhou et al. [39] develop a method for PREG, using Horn’s or Arun’s methods [3], [4] (see also [40]). Local Methods. In contrast to global methods, local methods require an initial guess. In the ”
TEASER: Fast and Certifiable Point Cloud Registration (2020, background): “The computer vision community has investigated the rotation search problem in the context of point cloud registration [17], [20], image stitching [11], motion estimation and 3-D reconstruction [4], [5].”
Real-Time Large-Scale Dense 3D Reconstruction with Loop Closure (2016, methodology): “In the works of [26, 25], submaps are created entirely based on a temporal criterion: every k frames a new subvolume is created, and coincidentally the default is k = 50 in both works.”