值得一读的论文：深度学习与三维点云

近年来深度学习方法应用在三维点云处理上较好的论文,mark一下。慢慢看。/(ㄒoㄒ)/~~
github已经有总结的比较好的repository了，不如直接来看这个awesome point cloud analysis，大佬有一直在维护，非常贴心！

Summary

比较新的一篇综述(arxiv 2019.12)： Deep Learning for 3D Point Clouds: A Survey 论文链接

PointNet family

• PointNet
  Charles R Q , Su H , Kaichun M , et al. PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2017.
  论文链接 & tensorflow源码 & pytorch源码
  成功将深度学习应用于原始点云数据的处理，Shared MLP + max pool（symmetric function），通过使用对称函数（max pool/average pool，论文中有通过实验说明max pool效果更好一些）来解决点云的置换不变性，使用T-Net来解决点云的旋转不变性。由于主要是逐点进行特征提取的，并没有考虑相邻点的空间关系，此外也没有考虑尺度不变性、点云密度不均的问题。
• PointNet++
  Qi C R , Yi L , Su H , et al. PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space[J]. 2017.
  论文链接 & tensorflow源码 & pytorch源码
  在PointNet的基础上考虑了相邻点的信息，更好的提取了局部特征，并考虑到了点云密度不均的问题。Sampling+Grouping+PointNet
• Frustum PointNets
  Qi C R , Liu W , Wu C , et al. Frustum PointNets for 3D Object Detection from RGB-D Data[J]. 2017.
  用PointNet的思想搭建的物体识别的框架，效果也不错。
• Flownet3D
  Liu, Xingyu, Qi, Charles R, Guibas, Leonidas J. FlowNet3D: Learning Scene Flow in 3D Point Clouds[J].
  论文链接
  根据两帧点云数据进行光流估计，也用PointNet完成了其中的一些操作。

Regular Processing

• VoxelNet
  Zhou, Yin, Tuzel, Oncel. VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection[J].
  将点云划分到网格中，然后用3DCNN做后续的工作。
• SPLATNet
  Su et al. SPLATNet: Sparse Lattice Networks for Point Cloud Processing CVPR2018
  论文链接 & pytorch源码
  将点云映射到晶格空间做双边卷积，相关的工作有Kiefei et al. Permutohedral Lattice CNNs. ICLR 2015, Jampani et al. Bilateral Neural Networks. CVPR 2016.
• PCNN
  Matan Atzmon, Haggai Maron, Yaron Lipman. Point Convolutional Neural Networks by Extension Operators. SIGGRAPH 2018.
  论文链接 & tensorflow源码
  将点云映射到体积空间而不存在任何的近似和离散
• PointCNN
  Li, Yangyan, Bu, Rui, Sun, Mingchao, Wu, Wei, Di, Xinhan, & Chen, Baoquan. . Pointcnn: convolution on ꭓ-transformed points.
  论文链接 & tensorflow源码 & pytorch源码 & mxnet
  设计了$\kappa$-Conv Operation
• PointSIFT
  Jiang M , Wu Y , Zhao T , et al. PointSIFT: A SIFT-like Network Module for 3D Point Cloud Semantic Segmentation[J]. 2018.
  论文链接 & tensorflow源码 & pytorch源码
  对于每一个局部邻域找到中心点，以中心点为原点建立坐标系，然后分方向进行卷积，此外还设计了尺度编码
• SO-Net
  Li J , Chen B M , Lee G H . SO-Net: Self-Organizing Network for Point Cloud Analysis[J]. 2018.
  论文链接 & pytorch源码
  Self-Organization Map + MLP
• Pointwise CNN
  Hua, Binh-Son, Tran, Minh-Khoi, Yeung, Sai-Kit. Pointwise Convolutional Neural Networks[J].
  论文链接 & tensorflow源码
  想法很简单，但是好像效果海星
• Kd-Net
  Roman Klokov，Victor Lempitsky. Escape from Cells: Deep Kd-Networks for the Recognition of 3D Point Cloud Models.
  论文链接 & pytorch源码链接
  把数据结构引入到点云的深度学习中
• FCPN
  Rethage, Dario, Wald, Johanna, Sturm, Jürgen et al. Fully-Convolutional Point Networks for Large-Scale Point Clouds[J].
  3D grid+PointNet+3DCNN

Graph-based Modeling

• DGCNN
  Wang Y , Sun Y , Liu Z , et al. Dynamic Graph CNN for Learning on Point Clouds[J]. 2018.
  论文链接 & tensorflow源码链接
  设计了动态图卷积网络
• RGCNN
  Gusi Te, Wei Hu, Zongming Guo, Amin Zheng. RGCNN: Regularized Graph CNN for Point Cloud Segmentation. MM 2018.
  论文链接
• GAN for Point Cloud
  Diego Valsesia, Giulia Fracastoro, Enrico Magli. Learning Localized Generative Models for 3D Point Clouds via Graph Convolution. ICLR 2019
  采用了GAN
• LocalSpecGCN
  Wang C , Samari B , Siddiqi K . Local Spectral Graph Convolution for Point Set Feature Learning[J]. 2018.
  论文链接 & tensorflow源码链接
  在谱域进行卷积
• Supperpoint GCN
  Landrieu L , Simonovsky M . Large-scale Point Cloud Semantic Segmentation with Superpoint Graphs[J]. CVPR 2018.
  论文链接 & pytorch源码链接
  针对大型点云
• Deep GCNs
  Guohao Li, Matthias Müller, Ali Thabet, Bernard Ghanem. DeepGCNs: Can GCNs Go as Deep as CNNs?. ICCV 2019 Oral
  论文链接 & 源码链接
  加入了残差卷积和密集卷积的思想，把网络做的更深

Convolution Kernel

• KC-Net
  Shen Y , Feng C , Yang Y , et al. Mining Point Cloud Local Structures by Kernel Correlation and Graph Pooling[J]. 2017.
  论文链接 & caffe源码 & pytorch源码
• KPConv
  Hugues et al. KPConv: Flexible and Deformable Convolution for Point Clouds. arXiv 2019
  论文链接 & 源码链接
• MCCNN
  源码链接
• PointConv
  PointConv: Deep Convolutional Networks on 3D Point Clouds
  论文链接
• Geo-CNN
  Lan S , Yu R , Yu G , et al. Modeling Local Geometric Structure of 3D Point Clouds using Geo-CNN[J]. 2018.
  论文链接
• Tangent Conv
  Tatarchenko et al. Tangent Convolutions for Dense Prediction in 3D
  论文链接
• PointWeb
  PointWeb: Enhancing Local Neighborhood Features for Point Cloud Processing. CVPR 2019
  论文链接 & pytorch源码 & b站效果展示
• A-CNN
  A-CNN: Annularly Convolutional Neural Networks on Point Clouds. CVPR 2019
  论文链接 & tensorflow源码

Robustness

• Li et al. Discrete Rotation Equivariance. ICRA 2019
• Cohen et al. Group Equivariant CNN. ICML 2016.
• Chen et al. ClusterNet. CVPR 2019
• Esteves et al. SO(3) Equivariant ECCV 2018
• Cohen et al. Spherical CNNs. ICLR 2018
• Rao et al. SFCNN CVPR 2019

另附：

• Point Cloud 2019 CVPR、ICCV、arXiv 的一些（好多）论文的链接和源码链接
• awesome point cloud analysis