AI Paper Atlas

Research map / Image generation / Super-resolution

Uncertainty-Driven Loss for Single Image Super-Resolution

Qian Ning, Weisheng Dong, Xin Li, Jinjian Wu, Guangming Shi

NeurIPS 2021 · Image generation › Super-resolution · approach: Efficient lightweight super-resolution · 32 citations (OpenAlex)

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
replace L1 loss with uncertainty-estimation loss LEU alone (ablation)EDSR-S x4 · Urban100PSNR (dB): 24.71 -> 24.21Hurt
replace L1 loss with LESU alone (ablation)EDSR-S x4 · Urban100PSNR (dB): 24.71 -> 24.27Hurt
add uncertainty-driven loss LUDL stage with LESU pretraining (full method)EDSR-S x4 · Urban100PSNR (dB): 25.25 -> 25.54Improved
use LESU (Jeffrey's prior sparsity) instead of LEU at step 1 (ablation)LUDL EDSR-S x4 · Urban100PSNR (dB): 25.49 -> 25.54Improved
weight loss by uncertainty instead of error map (ablation)EDSR-S x4 · Urban100PSNR (dB): 25.40 -> 25.54Improved
weight loss by uncertainty instead of HR gradient map (ablation)EDSR-S x4 · Urban100PSNR (dB): 25.42 -> 25.54Improved
use log scaling instead of linear scaling for uncertainty weights (ablation)LUDL EDSR-S x4 · Urban100PSNR (dB): 25.54 -> 25.57No clear effect
use exp(s) instead of linear scaling (ablation)LUDL EDSR-S x4 · Urban100PSNR (dB): 25.54 -> 25.53No clear effect
use LUDL instead of GRAMEDSR-S x4 · Urban100PSNR (dB): 24.81 -> 25.54Improved
use LUDL instead of GRAMDPDNN x4 · Urban100PSNR (dB): 25.63 -> 26.09Improved
apply GRAM instead of original lossEDSR-S x4 · Urban100PSNR (dB): 25.25 -> 24.81Hurt
apply LUDL under blur-downsampling degradationEDSR-S x4 · Urban100PSNR (dB): 25.77 -> 25.95Improved

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