AI Paper Atlas

Research map / Image generation / Super-resolution

Global Learnable Attention for Single Image Super-Resolution

Jian-Nan Su, Min Gan, Guangyong Chen, Jia-Li Yin, C. L. Philip Chen

TPAMI 2022 · Image generation › Super-resolution · approach: Efficient lightweight super-resolution · 65 citations (OpenAlex)

Read the paper

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 learnable similarity scoring s_l to fixed dot product in GLA (no LFFB) (ablation)DLSN · Set14 (x2)PSNR (dB): 33.33 -> 33.49Improved
add learnable similarity scoring s_l on top of fixed dot product (with LFFB) (ablation)DLSN · Set14 (x2)PSNR (dB): 33.65 -> 33.77Improved
add LFFB local features fusion block (ablation)DLSN · Set14 (x2)PSNR (dB): 33.49 -> 33.77Improved
add GLA to LFFB (ablation)DLSN · Set14 (x2)PSNR (dB): 33.54 -> 33.77Improved
increase hashing rounds in training and testing from h=1 to h=4 (ablation)DLSN · Set14 (x2)PSNR (dB): 33.60 -> 33.78Improved
use h=3 instead of h=4 hashing rounds (ablation)DLSN · Set14 (x2)PSNR (dB): 33.77 -> 33.78No clear effect
enlarge bucket size beyond l=128 (ablation)DLSN · Set14 (x2)PSNR (dB)Hurt
integrate GLA into FSRCNNFSRCNN · SISR benchmarksPSNR (dB): +0.13dBImproved
integrate GLA into EDSREDSR · SISR benchmarksPSNR (dB): +0.14dBImproved
integrate GLA into RCANRCAN · SISR benchmarksPSNR (dB): +0.17dBImproved
DLSN vs NLSN at x3 on Urban100DLSN · Urban100 (x3)PSNR (dB): +0.29dBImproved
DLSN (C=128) vs NLSN with fewer parametersDLSN · Manga109 (x4)PSNR (dB): +0.38dBImproved

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