华东师范大学学报(自然科学版) ›› 2024, Vol. 2024 ›› Issue (2): 108-118.doi: 10.3969/j.issn.1000-5641.2024.02.012

• 计算机科学 • 上一篇    

基于组对比学习的弱监督三维点云语义分割方法

郑智鸿, 宋海川*()   

  1. 1. 华东师范大学 计算机科学与技术学院, 上海 200062
  • 收稿日期:2023-02-01 出版日期:2024-03-25 发布日期:2024-03-18
  • 通讯作者: 宋海川 E-mail:hcsong@cs.ecnu.edu.cn

Group contrastive learning for weakly-supervised 3D point cloud semantic segmentation

Zhihong ZHENG, Haichuan SONG*()   

  1. 1. School of Computer Science and Technology, East China Normal University, Shanghai 200062, China
  • Received:2023-02-01 Online:2024-03-25 Published:2024-03-18
  • Contact: Haichuan SONG E-mail:hcsong@cs.ecnu.edu.cn

摘要:

三维点云语义分割方法, 是三维视觉环境感知中的重要任务, 被广泛应用于自动驾驶、增强现实、机器人等领域. 然而, 大多数语义分割方法工作在全监督的模式下, 为数据标注带来了极大的压力, 为了解决对于大规模点云标注数据的依赖问题, 许多工作基于有标签数据训练生成伪标签进一步迭代训练模型, 但未考虑到错误伪标签累积所导致的确认偏差. 针对该问题, 本文提出了一种基于组对比学习的弱监督三维点云语义分割方法, 在从伪标签中选择的正例组与负例组之间构造对比学习, 令伪标签之间相互竞争, 减少错误伪标签的梯度贡献, 从而缓解确认偏差. 实验结果表明, 本文所提出的方法在S3DIS、ScanNet-V2、Semantic3D等3个公开数据集上, 相较于目前最优方法均取得了更优的精度.

关键词: 弱监督学习, 三维点云, 语义分割, 对比学习

Abstract:

Three-dimensional point cloud semantic segmentation is an essential task for 3D visual perception and has been widely used in autonomous driving, augmented reality, and robotics. However, most methods work under a fully-supervised setting, which heavily relies on fully annotated datasets. Many weakly-supervised methods have utilized the pseudo-labeling method to retrain the model and reduce the labeling time consumption. However, the previous methods have failed to address the conformation bias induced by false pseudo labels. In this study, we proposed a novel weakly-supervised 3D point cloud semantic segmentation method based on group contrastive learning, constructing contrast between positive and negative sample groups selected from pseudo labels. The pseudo labels will compete with each other within the group contrastive learning, reducing the gradient contribution of falsely predicted pseudo labels. Results on three large-scale datasets show that our method outperforms state-of-the-art weakly-supervised methods with minimal labeling annotations and even surpasses the performance of some classic fully-supervised methods.

Key words: weakly-supervised learning, 3D point cloud, semantic segmentation, contrastive learning

中图分类号: