EUVIMEDEUVIMED entdecken
Crossref

Flow-Anything: Learning Real-World Optical Flow Estimation From Large-Scale Single-View Images

Yingping Liang, Ying Fu, Yutao Hu, Wenqi Shao, Jiaming Liu, Debing Zhang

IEEE Transactions on Pattern Analysis and Machine Intelligence · 2025-10 · 47 (10) · 8435-8452

DOI: 10.1109/tpami.2025.3576851 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025-10. Verfasst von Yingping Liang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 16-mal zitiert (Crossref) und nennt 95 Literaturangaben. Gefördert von: National Key R&D Program of China, National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Optical flow estimation is a crucial subfield of computer vision, serving as a foundation for video tasks. However, the real-world robustness is limited by animated synthetic datasets for training. This introduces domain gaps when applied to real-world applications and limits the benefits of scaling up datasets. To address these challenges, we propose Flow-Anything, a large-scale data generation framework designed to learn optical flow estimation from any single-view images in the real world. We employ two effective steps to make data scaling-up promising. First, we convert a single-view image into a 3D representation using advanced monocular depth estimation networks. This allows us to render optical flow and novel view images under a virtual camera. Second, we develop an Object-Independent Volume Rendering module and a Depth-Aware Inpainting module to model the dynamic objects in the 3D representation. These two steps allow us to generate realistic datasets for training from large-scale single-view images, namely FA-Flow Dataset. For the first time, we demonstrate the benefits of generating optical flow training data from large-scale real-world images, outperforming the most advanced unsupervised methods and supervised methods on synthetic datasets. Moreover, our models serve as a foundation model and enhance the performance of various downstream video tasks.

Förderung

  • National Key R&D Program of China
  • National Natural Science Foundation of China
  • Fundamental Research Funds for the Central Universities

Zitieren

APA

Liang, Y., Fu, Y., Hu, Y., Shao, W., Liu, J., & Zhang, D. (2025). Flow-Anything: Learning Real-World Optical Flow Estimation From Large-Scale Single-View Images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 47(10), 8435-8452. https://doi.org/10.1109/tpami.2025.3576851

Vancouver

Liang Y, Fu Y, Hu Y, Shao W, Liu J, Zhang D. Flow-Anything: Learning Real-World Optical Flow Estimation From Large-Scale Single-View Images. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2025;47(10):8435-8452. doi:10.1109/tpami.2025.3576851

BibTeX
@article{Liang2025,
  author = {Liang, Yingping and Fu, Ying and Hu, Yutao and Shao, Wenqi and Liu, Jiaming and Zhang, Debing},
  title = {Flow-Anything: Learning Real-World Optical Flow Estimation From Large-Scale Single-View Images},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year = {2025},
  volume = {47},
  number = {10},
  pages = {8435--8452},
  doi = {10.1109/tpami.2025.3576851}
}
Kostenlos starten →