DoGFlow: Self-Supervised LiDAR Scene Flow via Cross-Modal Doppler Guidance

2025-08-25

AI Summary

Existing 3D scene flow estimation methods heavily rely on scarce, costly manual annotations, while current self-supervised approaches suffer from degraded performance under long-range and adverse weather conditions. To address these limitations, this paper proposes a radar-guided self-supervised LiDAR scene flow estimation framework. Our key innovation is the first introduction of a cross-modal pseudo-label transfer mechanism: high-confidence motion pseudo-labels are generated from 4D radar Doppler velocity measurements and precisely transferred to the LiDAR point cloud domain via dynamic-aware association and ambiguity-aware propagation. Furthermore, we integrate temporal LiDAR-radar joint optimization with self-supervised point cloud matching to achieve accurate 3D motion estimation without ground-truth annotations. On the MAN TruckScenes dataset, our method achieves over 90% of the performance of a fully supervised baseline using only 10% real annotations, demonstrating substantial improvements in labeling efficiency and generalization capability.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Unsupervised & Self-Supervised Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Accurate 3D scene flow estimation is critical for autonomous systems to navigate dynamic environments safely, but creating the necessary large-scale, manually annotated datasets remains a significant bottleneck for developing robust perception models. Current self-supervised methods struggle to match the performance of fully supervised approaches, especially in challenging long-range and adverse weather scenarios, while supervised methods are not scalable due to their reliance on expensive human labeling. We introduce DoGFlow, a novel self-supervised framework that recovers full 3D object motions for LiDAR scene flow estimation without requiring any manual ground truth annotations. This paper presents our cross-modal label transfer approach, where DoGFlow computes motion pseudo-labels in real-time directly from 4D radar Doppler measurements and transfers them to the LiDAR domain using dynamic-aware association and ambiguity-resolved propagation. On the challenging MAN TruckScenes dataset, DoGFlow substantially outperforms existing self-supervised methods and improves label efficiency by enabling LiDAR backbones to achieve over 90% of fully supervised performance with only 10% of the ground truth data. For more details, please visit https://ajinkyakhoche.github.io/DogFlow/

Problem

• Self-supervised LiDAR scene flow estimation without manual annotations
• Overcoming performance gap in long-range and adverse weather scenarios
• Reducing reliance on expensive human-labeled ground truth data

Innovation

• Self-supervised LiDAR scene flow estimation
• Cross-modal Doppler guidance from radar
• Real-time motion pseudo-labels transfer