/Survey_3D_DL

A survey on 3D Deep Learning

A Survey on 3D Deep Learning

  • Dataset
  • Paper
  • Tutorial
  • Packages

Background

Data Representation

The 3D data can be represented in the following forms:

  • multi-view RGB(D) images
  • volumetric
  • polygonal mesh
  • point cloud
  • primitive-based CAD models

Dataset

Packages

  • Geometric Deep Learning Extension Library for PyTorch (Github)

  • Pytorch implementation of Graph Convolution Networks & Graph Attention Convolutional Networks (Github)

  • PyTorch implementation of Graph Convolutional Networks for semi-supervised classification (Github, paper, blog)

  • Point cloud semantic segmentation via Deep 3D Convolutional Neural Network (Github, slides)

Tutorial

  • A Tutorial on 3D Deep Learning (CVPR 2017) (web, video)

  • Machine Learning for 3D Data (Stanford CS468 - Spring 2017) (web)

  • Machine Learning for 3D Data (CSE291-I00 - Winter 2018)(web)

  • Data-Driven Shape Analysis and Processing (original version, latest version)

  • Polygon Mesh Processing (web)

    A free online book about mesh representation

  • Udacity Interactive 3D Graphics web

Papers

Feature based methods

  • [2015] 3D Mesh Labeling via Deep Convolutional Neural Networks (paper)

  • 3D Shape Segmentation via Shape Fully Convolutional Networks (paper)

  • 3D Mesh Segmentation via Multi-branch 1D Convolutional Neural Networks (paper)

Octree

  • [2017] O-CNN: Octree-based Convolutional Neural Networks for 3D Shape Analysis (paper, code)

Projection

  • [2017] 3D Shape Segmentation with Projective Convolutional Networks (paper, web, code)