samir-abh/HARP-NeXt

This repository contains model weights for HARP-NeXt: High-Speed and Accurate Range-Point Fusion Network for 3D LiDAR Semantic Segmentation, a lightweight, high-performing, and embedded 3D point cloud architecture.

In HARP-NeXt, we first propose a novel point cloud pre-processing methodology that significantly reduces computational overhead. Then, we design the Conv-SE-NeXt feature extraction block to efficiently capture representations without deep layer stacking per network stage. We also employ a multi-scale range-point fusion backbone that leverages information at multiple abstraction levels to preserve essential geometric details, thereby enhancing accuracy.

Paper & Resources

• Paper: HARP-NeXt: High-Speed and Accurate Range-Point Fusion Network for 3D LiDAR Semantic Segmentation
• Arxiv: https://arxiv.org/abs/2510.06876
• Codebase: https://github.com/SamirAbouHaidar/HARP-NeXt

Models

We release the pretrained model weights for the nuScenes and SemanticKITTI benchmarks we reported in our paper.

Citation

@INPROCEEDINGS{abouhaidar2025harpnext,
  author={Abou Haidar, Samir and Chariot, Alexandre and Darouich, Mehdi and Joly, Cyril and Deschaud, Jean-Emmanuel},
  booktitle={2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  title={HARP-NeXt: High-Speed and Accurate Range-Point Fusion Network for 3D LiDAR Semantic Segmentation}, 
  year={2025},
  pages={5946-5953},
  doi={10.1109/IROS60139.2025.11246557}}