Large-scale Point Cloud Semantic Segmentation with Superpoint Graphs
About
We propose a novel deep learning-based framework to tackle the challenge of semantic segmentation of large-scale point clouds of millions of points. We argue that the organization of 3D point clouds can be efficiently captured by a structure called superpoint graph (SPG), derived from a partition of the scanned scene into geometrically homogeneous elements. SPGs offer a compact yet rich representation of contextual relationships between object parts, which is then exploited by a graph convolutional network. Our framework sets a new state of the art for segmenting outdoor LiDAR scans (+11.9 and +8.8 mIoU points for both Semantic3D test sets), as well as indoor scans (+12.4 mIoU points for the S3DIS dataset).
Loic Landrieu, Martin Simonovsky• 2017
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Semantic segmentation | S3DIS (Area 5) | mIOU62.1 | 799 | |
| Semantic segmentation | SemanticKITTI (test) | mIoU20 | 335 | |
| Semantic segmentation | S3DIS (6-fold) | mIoU (Mean IoU)85.5 | 315 | |
| Semantic segmentation | S3DIS | mIoU62.1 | 88 | |
| Semantic segmentation | S3DIS (test) | mIoU58 | 47 | |
| Semantic segmentation | SemanticKITTI single-scan (test) | mIoU17.4 | 45 | |
| LiDAR Semantic Segmentation | SemanticKITTI sequences 11 to 21 (test) | Car IoU68.3 | 35 | |
| Semantic segmentation | Semantic3D reduced-8 (test) | mIoU76.2 | 33 | |
| Indoor Scene Segmentation | S3DIS (6-fold val) | mIoU (Category)62.1 | 32 | |
| 3D Semantic Segmentation | S3DIS Area 5 (test) | mIoU (%)58 | 32 |
Showing 10 of 27 rows