CorrelationFlow: A Training-Free Geometric Approach for LiDAR Scene Flow Estimation

2026-07-31

AI Summary

Existing LiDAR scene flow methods commonly fail in sparse, long-range, or high-speed motion scenarios due to their reliance on shared architectures and self-supervised losses. This work proposes the first training-free, purely geometric framework that reformulates scene flow estimation as a connected-component labeling and correlation maximization problem on bird’s-eye-view occupancy images, enabling efficient motion estimation from a single frame pair through sparse boundary keypoint matching. By eschewing deep learning entirely and relying solely on classical geometric and computer vision techniques, the method achieves second place among unsupervised approaches on the Argoverse 2 2026 multi-domain test set and exhibits the mildest performance degradation at long ranges, significantly outperforming existing unsupervised methods in such challenging conditions.

Abstract

LiDAR scene flow estimation has settled into a monoculture: nearly all recent methods share the same feed-forward architecture and the same family of self-supervised losses, inheriting each other's assumptions, and each other's blind spots. When those assumptions fail, as they do for sparse, distant, or fast-moving objects, every method built on them fails together, and adding parameters or simulated training data does not fix what the formulation itself gets wrong. This paper takes the opposite path. We present CorrelationFlow, a training-free geometric framework that reduces scene flow to two textbook operations: connected-component labeling and correlation maximization on bird's-eye-view occupancy images. Objects are isolated as spatio-temporal connected components, their motions recovered as correlation peaks, and the resulting velocities propagated to all member points. However, this dense correlation evaluates every candidate displacement of every cluster and requires a window of past sweeps; therefore, we develop a sparse counterpart that operates on a single sweep pair by matching lightweight occupancy descriptors at boundary key points. Because nothing is trained, nothing is inherited: on the multi-domain test set of the Argoverse 2 2026 Scene Flow Challenge, spanning five datasets with heterogeneous sensors and platforms, CorrelationFlow ranked second among unsupervised methods and degrades most gracefully at long range, where the shared assumptions of learned methods break down. Our results suggest that a substantial share of the scene flow problem is solvable by classical computer vision, and that progress may require questioning the formulation, not scaling it.

Problem

• LiDAR scene flow
• geometric approach
• training-free
• correlation maximization
• unsupervised estimation

Innovation

• training-free
• geometric approach
• correlation maximization
• connected-component labeling
• LiDAR scene flow