SynFlow: Scaling Up LiDAR Scene Flow Estimation with Synthetic Data

2026-04-10

AI Summary

This work addresses the scarcity of high-quality LiDAR scene flow annotations in real-world datasets, which severely limits the generalization of 3D motion perception models. To overcome this challenge, the authors propose a motion-centric synthetic data learning paradigm and introduce SynFlow, a dedicated generation pipeline that yields the SynFlow-4k dataset—comprising 4,000 sequences (approximately 940k frames), 34 times larger than existing real-world benchmarks. Departing from conventional emphasis on photorealistic sensor fidelity, the approach prioritizes modeling motion priors, demonstrating strong domain invariance and zero-shot transferability. Models trained solely on this synthetic data match supervised performance on nuScenes and surpass state-of-the-art methods by 31.8% on TruckScenes; with fine-tuning on merely 5% of real labels, they outperform models trained on full real datasets.

Technology Category

• Computer Vision: Motion & Tracking
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Large Multimodal Models (LMMs)

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
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Reliable 3D dynamic perception requires models that can anticipate motion beyond predefined categories, yet progress is hindered by the scarcity of dense, high-quality motion annotations. While self-supervision on unlabeled real data offers a path forward, empirical evidence suggests that scaling unlabeled data fails to close the performance gap due to noisy proxy signals. In this paper, we propose a shift in paradigm: learning robust real-world motion priors entirely from scalable simulation. We introduce SynFlow, a data generation pipeline that generates large-scale synthetic dataset specifically designed for LiDAR scene flow. Unlike prior works that prioritize sensor-specific realism, SynFlow employs a motion-oriented strategy to synthesize diverse kinematic patterns across 4,000 sequences ($\sim$940k frames), termed SynFlow-4k. This represents a 34x scale-up in annotated volume over existing real-world benchmarks. Our experiments demonstrate that SynFlow-4k provides a highly domain-invariant motion prior. In a zero-shot regime, models trained exclusively on our synthetic data generalize across multiple real-world benchmarks, rivaling in-domain supervised baselines on nuScenes and outperforming state-of-the-art methods on TruckScenes by 31.8%. Furthermore, SynFlow-4k serves as a label-efficient foundation: fine-tuning with only 5% of real-world labels surpasses models trained from scratch on the full available budget. We open-source the pipeline and dataset to facilitate research in generalizable 3D motion estimation. More detail can be found at https://kin-zhang.github.io/SynFlow.

Problem

• LiDAR scene flow
• motion annotation scarcity
• 3D dynamic perception
• domain generalization
• synthetic data

Innovation

• Synthetic Data
• LiDAR Scene Flow
• Domain Invariance
• Zero-shot Generalization
• Motion Prior