Test-Time Augmentation for 3D Point Cloud Classification and Segmentation

被引:0
|
作者
Tuan-Anh Vu [1 ]
Sarkar, Srinjay [2 ]
Zhang, Zhiyuan [3 ]
Binh-Son Hua [4 ]
Yeung, Sai-Kit [1 ]
机构
[1] Hong Kong Univ Sci & Technol, Hong Kong, Peoples R China
[2] VinAI Res, Hanoi, Vietnam
[3] Singapore Management Univ, Singapore, Singapore
[4] Trinity Coll Dublin, Dublin, Ireland
关键词
D O I
10.1109/3DV62453.2024.00153
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Data augmentation is a powerful technique to enhance the performance of a deep learning task but has received less attention in 3D deep learning. It is well known that when 3D shapes are sparsely represented with low point density, the performance of the downstream tasks drops significantly. This work explores test-time augmentation (TTA) for 3D point clouds. We are inspired by the recent revolution of learning implicit representation and point cloud upsampling, which can produce high-quality 3D surface reconstruction and proximity-to-surface, respectively. Our idea is to leverage the implicit field reconstruction or point cloud upsampling techniques as a systematic way to augment point cloud data. Mainly, we test both strategies by sampling points from the reconstructed results and using the sampled point cloud as test-time augmented data. We show that both strategies are effective in improving accuracy. We observed that point cloud upsampling for test-time augmentation can lead to more significant performance improvement on downstream tasks such as object classification and segmentation on the ModelNet40, ShapeNet, ScanObjectNN, and SemanticKITTI datasets, especially for sparse point clouds.
引用
收藏
页码:1543 / 1553
页数:11
相关论文
共 50 条
  • [1] Geometry Sharing Network for 3D Point Cloud Classification and Segmentation
    Xu, Mingye
    Zhou, Zhipeng
    Qiao, Yu
    [J]. THIRTY-FOURTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, THE THIRTY-SECOND INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE AND THE TENTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2020, 34 : 12500 - 12507
  • [2] Multi-Modal Continual Test-Time Adaptation for 3D Semantic Segmentation
    Cao, Haozhi
    Xu, Yuecong
    Yang, Jianfei
    Yin, Pengyu
    Yuan, Shenghai
    Xie, Lihua
    [J]. 2023 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2023), 2023, : 18763 - 18773
  • [3] Time and Memory Efficient 3D Point Cloud Classification
    Ullah, Shan
    Qayyum, Usman
    Choudhry, Aadil Jaleel
    [J]. PROCEEDINGS OF 2019 16TH INTERNATIONAL BHURBAN CONFERENCE ON APPLIED SCIENCES AND TECHNOLOGY (IBCAST), 2019, : 521 - 525
  • [4] 3D Point Cloud Segmentation: A survey
    Anh Nguyen
    Le, Bac
    [J]. PROCEEDINGS OF THE 2013 6TH IEEE CONFERENCE ON ROBOTICS, AUTOMATION AND MECHATRONICS (RAM), 2013, : 225 - 230
  • [5] Deep 3D point cloud classification and segmentation network based on GateNet
    Liu, Hui
    Tian, Shuaihua
    [J]. VISUAL COMPUTER, 2024, 40 (02): : 971 - 981
  • [6] Structure-Aware Convolution for 3D Point Cloud Classification and Segmentation
    Wang, Lei
    Liu, Yuxuan
    Zhang, Shenman
    Yan, Jixing
    Tao, Pengjie
    [J]. REMOTE SENSING, 2020, 12 (04)
  • [7] Leaves Segmentation in 3D Point Cloud
    Gelard, William
    Herbulot, Ariane
    Devy, Michel
    Debaeke, Philippe
    McCormick, Ryan F.
    Truong, Sandra K.
    Mullet, John
    [J]. ADVANCED CONCEPTS FOR INTELLIGENT VISION SYSTEMS (ACIVS 2017), 2017, 10617 : 664 - 674
  • [8] Deep 3D point cloud classification and segmentation network based on GateNet
    Hui Liu
    Shuaihua Tian
    [J]. The Visual Computer, 2024, 40 (2) : 971 - 981
  • [9] EB-LG Module for 3D Point Cloud Classification and Segmentation
    Chen, Jintao
    Zhang, Yan
    Ma, Feifan
    Tan, Zhuangbin
    [J]. IEEE ROBOTICS AND AUTOMATION LETTERS, 2023, 8 (01) : 160 - 167
  • [10] Training- and Test-Time Data Augmentation for Hyperspectral Image Segmentation
    Nalepa, Jakub
    Myller, Michal
    Kawulok, Michal
    [J]. IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, 2020, 17 (02) : 292 - 296