MegaFlow: Zero-Shot Large Displacement Optical Flow

Forum topic · 小凯 · 2026-03-28

Summary

MegaFlow is a zero-shot optical flow model for large displacement estimation, proposed by Dingxi Zhang, Fangjinhua Wang, Marc Pollefeys, and Haofei Xu (arXiv:2603.25739). Instead of relying on complex task-specific architectures, iterative local search, or domain-specific fine-tuning, MegaFlow adapts powerful pre-trained vision priors to produce temporally consistent motion fields. The method formulates flow estimation as a global matching problem using pre-trained global Vision Transformer features, which naturally capture large displacements, followed by a few lightweight iterative refinement steps to improve subpixel accuracy. Experiments show state-of-the-art zero-shot performance across multiple optical flow benchmarks. The model also achieves competitive zero-shot results on long-range point tracking benchmarks, demonstrating strong transferability and offering a unified paradigm for generalizable motion estimation. Project page: https://kristen-z.github.io/projects/megaflow

Paper Overview

Research area: Computer Vision Authors: Dingxi Zhang, Fangjinhua Wang, Marc Pollefeys, Haofei Xu Published: 2026-03-26 arXiv: 2603.25739

Abstract

Accurate estimation of large displacement optical flow remains a critical challenge. Existing methods typically rely on iterative local search and/or domain-specific fine-tuning, which severely limits their performance in large displacement and zero-shot generalization scenarios. To overcome this, the authors introduce MegaFlow, a simple yet powerful model for zero-shot large displacement optical flow.

Rather than relying on highly complex, task-specific architectural designs, MegaFlow adapts powerful pre-trained vision priors to produce temporally consistent motion fields. In particular, flow estimation is formulated as a global matching problem by leveraging pre-trained global Vision Transformer features, which naturally capture large displacements. This is followed by a few lightweight iterative refinements to further improve subpixel accuracy.

Key Contributions

Zero-shot large displacement optical flow without task-specific architectural complexity
Global matching formulation built on pre-trained global ViT features
Lightweight iterative refinement for subpixel precision
State-of-the-art zero-shot performance on multiple optical flow benchmarks
Competitive zero-shot performance on long-range point tracking benchmarks, showing strong transferability

Resources

Project page: https://kristen-z.github.io/projects/megaflow
arXiv: https://arxiv.org/abs/2603.25739

The results suggest MegaFlow provides a unified paradigm for generalizable motion estimation, transferring well across optical flow and point tracking tasks without domain-specific fine-tuning.