Event
CorrelationFlow Achieves Training-Free LiDAR Scene Flow Estimation via BEV Correlation
1 reports · 1 sources
The paper proposes CorrelationFlow, which reformulates LiDAR scene flow estimation as a correlation maximization problem between bird's-eye-view (BEV) images, replacing learnable parameters with a sequence of image processing operations to achieve training-free estimation. The method won second place in the unsupervised track of the Argoverse 2 Scene Flow Challenge (CVPR 2026). The test set covers five datasets, multiple vehicle platforms and environments, and results show its performance can match or even surpass learning-based methods. The source code will be open-sourced.
Report timeline
- CorrelationFlow: A Training-Free LiDAR Scene Flow Estimation Method Based on BEV Correlation
The paper proposes CorrelationFlow, which reformulates LiDAR scene flow estimation as a correlation maximization problem between Bird's-Eye View (BEV) images, replacing learnable parameters with a chain of image processing operations. This training-free method won second place in the unsupervised track of the Argoverse 2 Scene Flow Challenge (CVPR 2026). The test set covers five datasets and a variety of vehicle platforms and environments, and results show its performance can match or even surpass learning-based methods. The source code will be open-sourced.
arXiv cs.RO ·