Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

WAFT: Warping-Alone Field Transforms for Optical Flow

About

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}.

Yihan Wang, Jia Deng• 2025

Related benchmarks

TaskDatasetResultRank
Optical Flow EstimationKITTI 2015 (train)
Fl-epe1.15
446
Optical FlowSintel (train)
AEPE (Clean)1.28
200
Optical Flow EstimationSintel Final (test)
EPE2.02
133
Optical Flow EstimationSintel clean (test)
EPE0.95
120
Optical Flow EstimationKITTI 2015 (test)
Fl-all3.56
108
Optical FlowKITTI (train)
Fl-all0.129
84
Optical Flow EstimationKITTI 2015
Fl-all3.31
60
Optical FlowSintel Clean
EPE0.94
59
Optical FlowSintel Final
EPE2.02
59
Point TrackingDAVIS TAP-Vid--
52
Showing 10 of 17 rows

Other info

Follow for update