Self-Supervised Diffusion-Based Scene Flow Estimation and Motion Segmentation with 4D Radar

This is the official repository of the RadarSFEMOS, a self-supervised method for 4D radar scene flow estimation.

Citation

If you find our work useful, please consider citing:

@article{liu2025ral,
  author={Liu, Yufei and Chen, Xieyuanli and Wang, Neng and Andreev, Stepan and Dvorkovich, Alexander and Fan, Rui and Lu, Huimin},
  journal={IEEE Robotics and Automation Letters}, 
  title={Self-Supervised Diffusion-Based Scene Flow Estimation and Motion Segmentation With 4D Radar}, 
  year={2025},
  volume={10},
  number={6},
  pages={5895-5902},
}

Abstract

We propose a novel self-supervised framework that exploits denoising diffusion models to effectively handle radar noise inputs and predict point-wise scene flow and motion status simultaneously. To extract key features from the raw input, we design a transformer-based feature encoder tailored to address the sparsity of 4D radar data. Additionally, we generate self-supervised segmentation signals by exploiting the discrepancy between robust rigid ego-motion estimates and scene flow predictions, thereby eliminating the need for manual annotations. Experimental evaluations on the View-of-Delft (VoD) dataset and TJ4DRadSet demonstrate that our method achieves state-of-the-art performance for both radar-based SFE and MOS.

License

This project is free software made available under the MIT License. For details see the LICENSE file.