NUC-Net: Non-uniform Cylindrical Partition Network for Efficient LiDAR Semantic Segmentation

News

• NUC-Net was accepted to IEEE Transactions on Circuits and Systems for Video Technology (TCSVT) in 2025.

Project Overview

NUC-Net is a network based on non-uniform cylindrical partitioning for efficient LiDAR semantic segmentation. This method overcomes the high computational cost of traditional -based approaches by adaptively partitioning point cloud data, significantly improving the efficiency and accuracy of LiDAR point cloud processing in large-scale environments.

Updates

• Non-uniform Cylindrical Partition: An adaptive partitioning method to better capture the geometric structure of LiDAR point clouds. (Completed)
• Instance Augmentation: A data augmentation strategy designed to enhance the robustness of instance-level segmentation. (Coming soon)

Features

• Non-uniform cylindrical partitioning strategy
• Faster and more accurate processing of LIDAR point cloud
• A few lines of code are enough to bring a significant improvement.
• Our method demonstrates strong generalizability and significantly enhances the performance of PolarNet and Cylinder3D as shown in the Figure below. The results presented in the table are obtained with test-time augmentation.

Installation

-Please refer to the following link: https://github.com/xinge008/Cylinder3D

Citation

-If you find NUC-Net useful to your research, please cite our work:

@article{wang2025nuc,
  title={NUC-Net: Non-uniform Cylindrical Partition Network for Efficient LiDAR Semantic Segmentation},
  author={Wang, Xuzhi and Feng, Wei and Kong, Lingdong and Wan, Liang},
  journal={IEEE Transactions on Circuits and Systems for Video Technology},
  year={2025},
  publisher={IEEE}
}