AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

UPFlow: Upsampling Pyramid for Unsupervised Optical Flow Learning

Kunming Luo, Chuan Wang, Shuaicheng Liu, Haoqiang Fan, Jue Wang, Jian Sun

CVPR 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 89 citations (OpenAlex)

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use UPFlow rather than ARFlowunsupervised optical flow estimation · KITTI 2012 testEPE: 1.8 -> 1.4Improved
Use UPFlow rather than UFlowunsupervised optical flow estimation · KITTI 2015 testF1-all: 11.13% -> 9.38%Improved
Add SGU to the model with CL, BDWL, and ARL (ablation)UPFlow · KITTI 2015 trainEPE ALL: 2.64 -> 2.56Improved
Add PDL to the model with CL, BDWL, and ARL (ablation)UPFlow · KITTI 2015 trainEPE ALL: 2.64 -> 2.61Improved
Add SGU to the model with CL, BDWL, and ARL (ablation)UPFlow · Sintel Clean trainEPE ALL: (2.61) -> (2.46)Improved
Add PDL to the model with CL, BDWL, and ARL (ablation)UPFlow · Sintel Final trainEPE ALL: (3.17) -> (2.93)Improved
Add both SGU and PDL (ablation)UPFlow with CL, BDWL, and ARL · Sintel Clean trainEPE ALL: (2.61) -> (2.33)Improved
Add SGU interpolation flow and interpolation map (ablation)SGU-FM · KITTI 2012EPE: 1.35 -> 1.27Improved
Add the SGU interpolation map (ablation)SGU-M · KITTI 2012EPE: 1.33 -> 1.27Improved
Replace bilinear upsampling with SGU (ablation)UPFlow · KITTI 2012EPE: 1.36 -> 1.27Improved
Replace no pyramid loss with PDL (ablation)UPFlow · Sintel Clean train x1EPE: (2.46) -> (2.33)Improved
Replace PDL with pyramid unsupervised loss at intermediate resolution (ablation)UPFlow · Sintel Clean train x1EPE: (2.33) -> (2.49)Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy