Learning Rotation-Invariant Representations of Point Clouds Using Aligned Edge Convolutional Neural Networks

被引:13
|
作者
Zhang, Junming [1 ]
Yu, Ming-Yuan [1 ]
Vasudevan, Ram [1 ]
Johnson-Roberson, Matthew [1 ]
机构
[1] Univ Michigan, Ann Arbor, MI 48109 USA
关键词
D O I
10.1109/3DV50981.2020.00030
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Point cloud analysis is an area of increasing interest due to the development of 3D sensors that are able to rapidly measure the depth of scenes accurately. Unfortunately, applying deep learning techniques to perform point cloud analysis is non-trivial due to the inability of these methods to generalize to unseen rotations. To address this limitation, one usually has to augment the training data, which can lead to extra computation and require larger model complexity. This paper proposes a new neural network called the Aligned Edge Convolutional Neural Network (AECNN) that learns a feature representation of point clouds relative to Local Reference Frames (LRFs) to ensure invariance to rotation. In particular, features are learned locally and aligned with respect to the LRF of an automatically computed reference point. The proposed approach is evaluated on point cloud classification and part segmentation tasks. This paper illustrates that the proposed technique outperforms a variety of state of the art approaches (even those trained on augmented datasets) in terms of robustness to rotation without requiring any additional data augmentation.
引用
收藏
页码:200 / 209
页数:10
相关论文
共 50 条
  • [41] Rotation-Invariant Local-to-Global Representation Learning for 3D Point Cloud
    Kim, Seohyun
    Park, Jaeyoo
    Han, Bohyung
    [J]. ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 33, NEURIPS 2020, 2020, 33
  • [42] Orthogonal Transforms For Learning Invariant Representations In Equivariant Neural Networks
    Singh, Jaspreet
    Singh, Chandan
    Rana, Ankur
    [J]. 2023 IEEE/CVF WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV), 2023, : 1523 - 1530
  • [43] Point-based Attention Convolutional Neural Networks for Point Clouds Semantic Segmentation
    Li, Ying
    Li, Qing
    [J]. ACM International Conference Proceeding Series, 2022, : 1642 - 1647
  • [44] Correspondence matching in unorganized 3D point clouds using Convolutional Neural Networks
    Pujol-Miro, Alba
    Casas, Josep R.
    Ruiz-Hidalgo, Javier
    [J]. IMAGE AND VISION COMPUTING, 2019, 83-84 : 51 - 60
  • [45] Residual learning with annularly convolutional neural networks for classification and segmentation of 3D point clouds
    Hassan, R.
    Fraz, M. M.
    Rajput, A.
    Shahzad, M.
    [J]. NEUROCOMPUTING, 2023, 526 : 96 - 108
  • [46] Learning Deep Graph Representations via Convolutional Neural Networks
    Ye, Wei
    Askarisichani, Omid
    Jones, Alex
    Singh, Ambuj
    [J]. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2022, 34 (05) : 2268 - 2279
  • [47] Cy-CNN: cylinder convolution based rotation-invariant neural network for point cloud registration
    Zhao, Hengwang
    Liang, Zhidong
    He, Yuesheng
    Wang, Chunxiang
    Yang, Ming
    [J]. SCIENCE CHINA-INFORMATION SCIENCES, 2023, 66 (05)
  • [48] Cy-CNN: cylinder convolution based rotation-invariant neural network for point cloud registration
    Hengwang ZHAO
    Zhidong LIANG
    Yuesheng HE
    Chunxiang WANG
    Ming YANG
    [J]. Science China(Information Sciences), 2023, 66 (05) : 73 - 87
  • [49] Learning and Transferring Mid-Level Image Representations using Convolutional Neural Networks
    Oquab, Maxime
    Bottou, Leon
    Laptev, Ivan
    Sivic, Josef
    [J]. 2014 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2014, : 1717 - 1724
  • [50] RIP-NeRF: Learning Rotation-Invariant Point-based Neural Radiance Field for Fine-grained Editing and Compositing
    Wang, Yuze
    Wang, Junyi
    Qu, Yansong
    Qi, Yue
    [J]. PROCEEDINGS OF THE 2023 ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA RETRIEVAL, ICMR 2023, 2023, : 125 - 134