RAFT: Recurrent All-Pairs Field Transforms for Optical Flow

Zachary Teed, Jia Deng
3/26/2020
cs.CV

Official code: princeton-vl/RAFT (4,117 stars, BSD-3-Clause, last commit 2025-08-24).

Abstract

We introduce Recurrent All-Pairs Field Transforms (RAFT), a new deep network architecture for optical flow. RAFT extracts per-pixel features, builds multi-scale 4D correlation volumes for all pairs of pixels, and iteratively updates a flow field through a recurrent unit that performs lookups on the correlation volumes. RAFT achieves state-of-the-art performance. On KITTI, RAFT achieves an F1-all error of 5.10%, a 16% error reduction from the best published result (6.10%). On Sintel (final pass), RAFT obtains an end-point-error of 2.855 pixels, a 30% error reduction from the best published result (4.098 pixels). In addition, RAFT has strong cross-dataset generalization as well as high efficiency in inference time, training speed, and parameter count. Code is available at https://github.com/princeton-vl/RAFT.

View on arXivView PDF

Code Implementations(1)

princeton-vl/RAFTOfficial100%
BSD-3-Clause
4,117683C++, CudaMar 27, 20201 years agoBSD-3-Clause

Similar Papers

FlowSeek: Optical Flow Made Easier with Depth Foundation Models and Motion Bases

88% match1 repo1502025

LiteFlowNet: A Lightweight Convolutional Neural Network for Optical Flow Estimation

87% match1 repo6342018

Rethinking RAFT for Efficient Optical Flow

87% match1 repo462024

Rethinking Optical Flow from Geometric Matching Consistent Perspective

87% match1 repo1002023

Leveraging Consistent Spatio-Temporal Correspondence for Robust Visual Odometry

87% match2024

Optical Flow Estimation using a Spatial Pyramid Network

86% match1 repo2592016

Cite this paper

@article{teed2020raft,
  title  = {RAFT: Recurrent All-Pairs Field Transforms for Optical Flow},
  author = {Zachary Teed and Jia Deng},
  year   = {2020},
  eprint = {2003.12039},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2003.12039}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2003.12039.svg)](https://codeofpaper.com/paper/2003.12039)

Discussion