AI Paper Atlas

Research map / Video / Video segmentation and tracking

Efficient Video Deblurring Guided by Motion Magnitude

Yusheng Wang, Yunfan Lu, Ye Gao, Lin Wang, Zhihang Zhong, Yinqiang Zheng, Atsushi Yamashita

ECCV 2022 · Video › Video segmentation and tracking · approach: Spike and neuromorphic deblurring

Read the paperCheck your idea against 137,217 papers

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
add MMAM with MMP (ablation)MMP-RNN · GOPROPSNR: 32.33 -> 32.64Improved
add motion magnitude loss (ablation)MMP-RNN · GOPROPSNR: 32.25 -> 32.33Improved
add transmission of non-deblurred features (ablation)MMP-RNN · GOPROPSNR: 32.06 -> 32.25Improved
add MMAM with MMP (ablation)MMP-RNN · BSDPSNR: 32.16 -> 32.79Improved
add motion magnitude loss (ablation)MMP-RNN · BSDPSNR: 32.07 -> 32.16Improved
add transmission of non-deblurred features (ablation)MMP-RNN · BSDPSNR: 32.03 -> 32.07Improved
add estimated MMP prior (ablation)MMP-RNN small · GOPROPSNR: 30.17 -> 30.48Improved
use ground-truth MMP instead of estimated MMP (ablation)MMP-RNN small · GOPROPSNR: 30.48 -> 30.54Improved
use center-frame flow magnitude as prior instead of none (ablation)MMP-RNN small · GOPROPSNR: 30.17 -> 30.47Improved
replace MMP and MMAM with spatial self-attention (ablation)MMP-RNN small · GOPROPSNR: 30.48 -> 30.37Hurt
normalize each ground-truth MMP map to 0-1 (ablation)MMP-RNN small · GOPROPSNR: 30.54 -> 30.41Hurt

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