Weakly Supervised Semantic Segmentation in 3DGraph-Structured Point Clouds of Wild Scenes

本資料は2021年7月18日に社内共有資料として展開していたものをWEBページ向けにリニューアルした内容になります。

■目次

Introduction

  • Point clouds
  • Task: semantic segmentation of point clouds
  • Idea: weakly supervised learning with 2D projected image
  • Note: Projection of 3D pointcloud onto 2D image

Method

  • Model architecture
  • Perspective rendering

Experiments

  • Datasets
  • Metrics
  • Comparison with other fully-supervised methods
  • Inference samples for SUNCG dataset
  • Inference samples for S3DIS dataset

Conclusion

■Point clouds

■Task: semantic segmentation of point clouds

■Idea: weakly supervised learning with 2D projected image

■Note: Projection of 3D pointcloud onto 2D image

■Model architecture

■Perspective rendering

■Datasets

■Metrics

■Comparison with other fully-supervised methods

■Inference samples for SUNCG dataset

■Inference samples for S3DIS dataset

■Conclusion

■ダウンロード

Weakly Supervised Semantic Segmentation in 3DGraph-Structured Point Clouds of Wild Scenes.pdf