[论文] GenOpticalFlow: A Generative Approach to Unsupervised Optical Flow Lea...

## 论文概要 **研究领域**: CV **作者**: Yixuan Luo, Feng Qiao, Zhexiao Xiong, Yanjing Li, Nathan Jacobs **发布时间**: 2026-03-23 **arXiv**: [2603.22270](https://arxiv.org/abs/2603.22270) ## 中文摘要 光流估计是计算机视觉中的一个基础问题，但对昂贵的真值标注的依赖限制了监督方法的可扩展性。虽然无监督和半监督方法缓解了这一问题，但它们往往受到基于亮度恒定和平滑假设的不可靠监督信号的影响，导致在复杂真实世界场景中的运动估计不准确。为克服这些限制，我们提出了GenOpticalFlow，这是一个新颖的框架，能够在无需人工标注的情况下合成大规模、完美对齐的帧-光流数据对用于监督光流训练。具体而言，我们的方法利用预训练的深度估计网络生成伪光流，作为条件输入用于训练下一帧生成模型，以生成高保真、像素对齐的后续帧。这一过程使得创建大量具有精确运动对应关系的高质量合成数据成为可能。此外，我们提出了一种不一致像素过滤策略，识别并移除生成帧中的不可靠像素，有效提升了在真实世界数据集上的微调性能。 ## 原文摘要 Optical flow estimation is a fundamental problem in computer vision, yet the reliance on expensive ground-truth annotations limits the scalability of supervised approaches. Although unsupervised and semi-supervised methods alleviate this issue, they often suffer from unreliable supervision signals based on brightness constancy and smoothness assumptions, leading to inaccurate motion estimation in complex real-world scenarios. To overcome these limitations, we introduce GenOpticalFlow, a novel framework that synthesizes large-scale, perfectly aligned frame-flow data pairs for supervised optical flow training without human annotations. Specifically, our method leverages a pre-trained depth estimation network to generate pseudo optical flows, which serve as conditioning inputs for a next-fram... --- *自动采集于 2026-03-25* #论文 #arXiv #CV #小凯