MVP-Net: Multiple View Pointwise Semantic Segmentation of Large-Scale Point Clouds
This work addresses a critical efficiency problem for autonomous driving systems that rely on real-time 3D environment perception, though it is incremental as it builds on existing pointwise segmentation methods.
The paper tackles the efficiency bottleneck of KNN-based neighbor searching in pointwise semantic segmentation of large-scale point clouds by proposing MVP-Net, which uses space filling curves and multi-rotation methods to achieve 11 times faster inference than RandLA-Net while maintaining the same accuracy on the SemanticKITTI dataset.
Semantic segmentation of 3D point cloud is an essential task for autonomous driving environment perception. The pipeline of most pointwise point cloud semantic segmentation methods includes points sampling, neighbor searching, feature aggregation, and classification. Neighbor searching method like K-nearest neighbors algorithm, KNN, has been widely applied. However, the complexity of KNN is always a bottleneck of efficiency. In this paper, we propose an end-to-end neural architecture, Multiple View Pointwise Net, MVP-Net, to efficiently and directly infer large-scale outdoor point cloud without KNN or any complex pre/postprocessing. Instead, assumption-based space filling curves and multi-rotation of point cloud methods are introduced to point feature aggregation and receptive field expanding. Numerical experiments show that the proposed MVP-Net is 11 times faster than the most efficient pointwise semantic segmentation method RandLA-Net and achieves the same accuracy on the large-scale benchmark SemanticKITTI dataset.