Joint 3D Scene Reconstruction and Class Segmentation

By Christian Häne, Christopher Zach, Andrea Cohen, Roland Angst, Marc Pollefeys. 2013

247 citations recorded by the source corpus.

Subfields: Aerospace Engineering, 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 Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Learning a Multi-View Stereo Machine (2017, methodology): “Learning has been used for multi-view reconstruction in the form of shape priors for objects [2, 9, 58, 20, 27, 52], or semantic class speciﬁc surface priors for scenes [22, 17, 45].”
OctNetFusion: Learning Depth Fusion from Data (2017, background): “Noise reduction can be achieved using variational techniques which integrate local smoothness assumptions [2, 19, 55] into the formulation.”
Semantic Scene Completion from a Single Depth Image (2016, neutral): “[9] and Blaha et al .”
SUN3D: A Database of Big Spaces Reconstructed Using SfM and Object Labels (2013, background): “While there are also several recent works [3, 6, 25, 19, 8] that combine recognition and 3D reconstruction, they do not use human annotations: instead, they jointly recognize objects and solve for 3D models.”
Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age (2016, neutral): “The second reason is that Gaussian elimination is performed on a linearized version of the problem, hence approaches as DDF-SAM [74] require good linearization points and complex bookkeeping to ensure consistency of the linearization points across the robots.”
Discrete optimization of ray potentials for semantic 3D reconstruction (2015, methodology): “Recently, also approaches which jointly reason about geometry and semantic labels have been proposed [11].”
Pulling Things out of Perspective (2014, background): “In these approaches, the rather weak source of mutual information utilized is the distribution of height [14] or surface normals [9] of different semantic classes.”
A Symmetry Prior for Convex Variational 3D Reconstruction (2016, methodology): “…volumetric 3D reconstruction framework which has been used in various settings, e.g. depth map fusion [29], 3D reconstruction [9,27], multi-label semantic reconstruction [5], and spatio-temporal reconstruction [17], with anisotropic regularization [10,5,18] or connectivity constraints [19].”
3DMV: Joint 3D-Multi-View Prediction for 3D Semantic Scene Segmentation (2018, background): “Typically, these approaches operate on a single RGB image; however, one can easily formulate the analogous task in 3D on a per-voxel basis [5, 13, 21, 34, 40, 41], which is a common scenario in the context of 3D scene reconstruction.”
Comparative Evaluation of Hand-Crafted and Learned Local Features (2017, methodology): “Next, we evaluate the performance on the South Building dataset [15], which consists of 128 highly overlapping images with mostly repetitive scene structure captured by the same camera in a struc-”
Learning Where to Classify in Multi-view Semantic Segmentation (2014, background): “Recently, [33, 34] combined the creation of geometry with the semantic labelling implicitly evaluating all data redundantly.”
Deep learning in remote sensing: a review (2017, methodology): “[171]), where 3-D reconstruction is performed with a volumetric method.”
Displets: Resolving stereo ambiguities using object knowledge (2015, background): “While the mutual benefits of recognition and reconstruction have been shown using simple priors and multiple views [15,29], little work has addressed the binocular stereo problem in this context with notable exceptions [33,48,61].”
Multimodal Classification of Remote Sensing Images: A Review and Future Directions (2015, methodology): “But the interesting part still lies ahead since: 1) semantic logic can be integrated in CRF models, as shown in [110], where the authors improve 3-D scene reconstruction by including class posterior probabilities and spatial logic in the interaction term; 2) by encoding higher order information in t”
Patches, Planes and Probabilities: A Non-Local Prior for Volumetric 3D Reconstruction (2016, background): “, by constraining the set of plausible geometries [2, 11, 19] or by modeling class specific normal distributions [20,21].”
Towards unified depth and semantic prediction from a single image (2015, methodology): “Several methods try to getter better semantic segmentation results given RGB-D data [28] [31] [12] [13], while others take the predicted semantic labels to estimate depth [23] [15].”
Multi-Label Semantic 3D Reconstruction Using Voxel Blocks (2016, methodology): “If we represent the scene as a voxel grid where all voxels have equal size, as in [10], we need to store an indicator variable for every label at every voxel.”
The ApolloScape Open Dataset for Autonomous Driving and Its Application (2018, methodology): “Our work is also related to joint reconstruction, pose estimation and parsing [25], [66] through embedding 2D-3D consistency.”
Large-Scale Semantic 3D Reconstruction: An Adaptive Multi-resolution Model for Multi-class Volumetric Labeling (2016, methodology): “Our hierarchical model is a direct extension of the fixedgrid convex labeling method [17] and emerges naturally as the optimal adaptive extension of that scheme, i.”