Dense Non-rigid Shape Correspondence Using Random Forests

By Emanuele Rodolà, Samuel Rota Bulò, Thomas Windheuser, Matthias Vestner, Daniel Cremers. 2014

185 citations recorded by the source corpus.

Subfields: Computational Mechanics, 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 Computational Mechanics

In Computer Vision and Pattern Recognition

Citations in the corpus

Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs (2016, background): “For this purpose, correspondence is cast as a labelling problem, where one tries to label each vertex of a given query shape X with the index of a corresponding point on some reference shape Y [38, 32, 7].”
Learning shape correspondence with anisotropic convolutional neural networks (2016, methodology): “The first attempts were aimed at learning local shape descriptors [15, 5, 27], and shape correspondence [20].”
Dense Human Body Correspondences Using Convolutional Networks (2015, methodology): “When both source and target shapes are full models, an automatic mapping between shapes with considerable deformations becomes possible as shown in [20, 25, 33, 10].”
Deep Functional Maps: Structured Prediction for Dense Shape Correspondence (2017, methodology): “We compare with the most recent state of the art in learning-based shape correspondence: random forests (RF) [36], anisotropic diffusion descriptors (ADD) [10], geodesic CNNs (GCNN) [29], anisotropic CNNs (ACNN) [9], and the very recent MoNet model [31].”
Deformable Shape Completion with Graph Convolutional Autoencoders (2017, background): “Dense non-rigid shape correspondence [23, 11, 33, 47, 8] is a fundamental challenge as it is an enabler for many high level tasks like pose or texture transfer across surfaces.”
Learning class‐specific descriptors for deformable shapes using localized spectral convolutional networks (2015, unknown): “…processing trend of learning invariant structure rather than trying to hand-craft them, several learning frameworks have been proposed in the geometry processing community as well, for applications such as correspondence [RRBW ∗ 14], retrieval [LBBC14], labelling and segmentation [KHS10, HSG13].”
Anisotropic Laplace-Beltrami Operators for Shape Analysis (2014, unknown): “Computing the LB operator directly on meshes with approaches such as [18, 15, 20], numerous works exploit the information contained in its eigen-decomposition to build a new representation of a shape, either by explicit formulas [23, 24, 1] or by learning [13, 22].”
Convolutional neural networks on surfaces via seamless toric covers (2017, methodology): “We compared our algorithm to a baseline random forest classifier [Breiman 2001] (using maltab’s implementation) which was recently shown to be a successful classifier for shape analysis problems [Rodolà et al. 2014].”
Image Matching from Handcrafted to Deep Features: A Survey (2020, methodology): “Rodolà et al. (2014) proposed a learning method that enables thewave kernel descriptor to recognize a broader class of deformations from the example set by using the random forest classifier.”
Anisotropic Diffusion Descriptors (2016, methodology): “In the second experiment, we reproduced the settings of Rodolà et al. [R ∗ 14] on the FAUST dataset, replacing their random forest with our ADD neural network model.”
Geometric deep learning (2016, methodology): “The text in the course materials is primarily based on previously published work including [71, 56, 86, 100, 10, 59, 97, 12, 11] among several others.”
Self-supervised Learning of Dense Shape Correspondence (2018, methodology): “Early examples include approaches based on decision forests and metric learning [29, 37, 17]; more recently, several papers have proposed an adaptation of deep learning models to non-Euclidean domains, achieving dramatic improvement.”
Geodesic Convolutional Neural Networks on Riemannian Manifolds (2015, methodology): “[37] on the FAUST dataset, replacing their random forest with a GCNN architecture GCNN3 containing three convolutional layers (input: 150-dimensional geometry vectors, LIN16+ReLU, GC32+AMP+ReLU, GC64+AMP+ReLU, GC128+AMP+ReLU, LIN256, LIN6890).”
Efficient Globally Optimal 2D-to-3D Deformable Shape Matching (2016, background): “A recent line of works builds upon the functional representation [18, 19, 22, 12, 23], where a point-wise map is replaced by a linear map between function spaces.”
3D-CODED: 3D Correspondences by Deep Deformation (2018, methodology): “Several neural network based methods have recently been developed to this end to analyze meshes [36,27,7,31] or depth maps [46].”
Deformation-driven shape correspondence via shape recognition (2017, background): “For these shapes, there have been recent e orts attempting to learn intrinsic invariants to compute correspondences [Masci et al. 2015; Rodolà et al. 2014].”