Learning Descriptor Networks for 3D Shape Synthesis and Analysis (2018-04-02T00:00:00.000000Z)

TL;DR

A 3D shape descriptor network, which is a deep convolutional energy-based model, for modeling volumetric shape patterns and can be used to train a 3D generator network via MCMC teaching.

Abstract

This paper proposes a 3D shape descriptor network, which is a deep convolutional energy-based model, for modeling volumetric shape patterns. The maximum likelihood training of the model follows an "analysis by synthesis" scheme and can be interpreted as a mode seeking and mode shifting process. The model can synthesize 3D shape patterns by sampling from the probability distribution via MCMC such as Langevin dynamics. The model can be used to train a 3D generator network via MCMC teaching. The conditional version of the 3D shape descriptor net can be used for 3D object recovery and 3D object super-resolution. Experiments demonstrate that the proposed model can generate realistic 3D shape patterns and can be useful for 3D shape analysis.

Authors

Y. Wu

2 papers

Song-Chun Zhu

17 papers

Wenguan Wang

11 papers

TL;DR

Abstract

Authors

References38 items

Cooperative Learning of Energy-Based Model and Latent Variable Model via MCMC Teaching

Introspective Neural Networks for Generative Modeling

Introspective Classification with Convolutional Nets

Dynamic Edge-Conditioned Filters in Convolutional Neural Networks on Graphs

Wasserstein GAN

Improving Generative Adversarial Networks with Denoising Feature Matching

Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

Deep Learning 3D Shape Surfaces Using Geometry Images

Cooperative Training of Descriptor and Generator Networks

Inducing wavelets into random fields via generative boosting

VConv-DAE: Deep Volumetric Shape Learning Without Object Labels

Volumetric and Multi-view CNNs for Object Classification on 3D Data

Learning a Predictable and Generative Vector Representation for Objects

A Theory of Generative ConvNet

ShapeNet: An Information-Rich 3D Model Repository

Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks

Learning FRAME Models Using CNN Filters

DeepPano: Deep Panoramic Representation for 3-D Shape Recognition

VoxNet: A 3D Convolutional Neural Network for real-time object recognition

Multi-view Convolutional Neural Networks for 3D Shape Recognition

Adam: A Method for Stochastic Optimization

Learning Sparse FRAME Models for Natural Image Patterns

3D ShapeNets: A deep representation for volumetric shapes

Auto-Encoding Variational Bayes

A probabilistic model for component-based shape synthesis

Herding dynamical weights to learn

Learning Generative Models via Discriminative Approaches

Pattern Theory: From Representation to Inference

Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Statistical Modeling and Conceptualization of Visual Patterns

A Morphable Model For The Synthesis Of 3D Faces

Filters, Random Fields and Maximum Entropy (FRAME): Towards a Unified Theory for Texture Modeling

An algorithm and data structure for 3D object synthesis using surface patch intersections

GENERATIVE ADVERSARIAL NETS

Exploiting the PANORAMA Representation for Convolutional Neural Network Classification and Retrieval

A Tutorial on Energy-Based Learning

Initializing mode seeking : For i = 1 ,..., ˜ n , generate Z i ∼ N ( 0 , I d ) , and generate ˆ Y i = g ( Z i ; α ( t ) )+ ε i

Mode seeking : For i = 1 ,..., ˜ n , starting from ˆ Y i , run l steps of Langevin dynamics to obtain ˜ Y i , each step following equation (4

Field of Study

Journal Information

Name

Page