Semantic segmentation of point clouds, aiming to assign each point a semantic category, is critical to 3D scene understanding. Although significant advances in recent years, most of the existing methods still suffer from either the object-level misclassification or the boundary-level ambiguity. In this paper, we present a robust semantic segmentation network by deeply exploring the geometry of point clouds, dubbed GeoSegNet. Our GeoSegNet consists of a multi-geometry-based encoder and a boundary-guided decoder. In the encoder, we develop a new residual geometry module from multi-geometry perspectives to extract object-level features. In the decoder, we introduce a contrastive boundary learning module to enhance the geometric representation of boundary points. Benefiting from the geometric encoder–decoder modeling, GeoSegNet infers the segmentation of objects effectively while making the intersections (boundaries) of two or more objects clear. GeoSegNet achieves a significant performance with 64.9% mIoU on the challenging S3DIS dataset (Area 5) and 70.2% mIoU on S3DIS sixfold. Experiments show obvious improvements of GeoSegNet over its competitors in terms of the overall segmentation accuracy and object boundary clearness. Code is available at https://github.com/Chen-yuiyui/GeoSegNet .
Bibliographical notePublisher Copyright:
© 2023, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.
- Contrastive boundary learning
- Point cloud semantic segmentation
- Residual geometry module