Saved in:
Bibliographic Details
Main Authors: Wang, Yihan, Deng, Jia
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.21526
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914309687214080
author Wang, Yihan
Deng, Jia
author_facet Wang, Yihan
Deng, Jia
contents We introduce Warping-Alone Field Transforms (WAFT), a simple and effective method for optical flow. WAFT is similar to RAFT but replaces cost volume with high-resolution warping, achieving better accuracy with lower memory cost. This design challenges the conventional wisdom that constructing cost volumes is necessary for strong performance. WAFT is a simple and flexible meta-architecture with minimal inductive biases and reliance on custom designs. Compared with existing methods, WAFT ranks 1st on Spring, Sintel, and KITTI benchmarks, achieves the best zero-shot generalization on KITTI, while being 1.3-4.1x faster than existing methods that have competitive accuracy (e.g., 1.3x than Flowformer++, 4.1x than CCMR+). Code and model weights are available at \href{https://github.com/princeton-vl/WAFT}{https://github.com/princeton-vl/WAFT}.
format Preprint
id arxiv_https___arxiv_org_abs_2506_21526
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle WAFT: Warping-Alone Field Transforms for Optical Flow
Wang, Yihan
Deng, Jia
Computer Vision and Pattern Recognition
We introduce Warping-Alone Field Transforms (WAFT), a simple and effective method for optical flow. WAFT is similar to RAFT but replaces cost volume with high-resolution warping, achieving better accuracy with lower memory cost. This design challenges the conventional wisdom that constructing cost volumes is necessary for strong performance. WAFT is a simple and flexible meta-architecture with minimal inductive biases and reliance on custom designs. Compared with existing methods, WAFT ranks 1st on Spring, Sintel, and KITTI benchmarks, achieves the best zero-shot generalization on KITTI, while being 1.3-4.1x faster than existing methods that have competitive accuracy (e.g., 1.3x than Flowformer++, 4.1x than CCMR+). Code and model weights are available at \href{https://github.com/princeton-vl/WAFT}{https://github.com/princeton-vl/WAFT}.
title WAFT: Warping-Alone Field Transforms for Optical Flow
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.21526