AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

Deep Degradation Prior for Low-Quality Image Classification

Yang Wang, Yang Cao, Zheng-Jun Zha, Jing Zhang, Zhiwei Xiong

CVPR 2020 · Image generation › Autoregressive and latent generation · approach: Learning-based image quality assessment · 59 citations (OpenAlex)

Check 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
plug the FDM (deep degradation prior) into a pre-trained VGG16 trained on only 10 foggy CUB-C imagesVGG16 · ImageNet-C fogaccuracy: 37% -> 55%Improved
plug the FDM into VGG16 trained on 10 motion-blur imagesVGG16 · ImageNet-C motion bluraccuracy: 24% -> 38%Improved
reduce FDM training data from all 11788 images to 1% (ablation)FDM · ImageNet-C Fog-3, Contrast-3, Brightness-3accuracy: drops by 5%Hurt
use FDM instead of dehaze-then-classify two-stage methodsVGG16, AlexNet · ImageNet-C fog, 5 levelsaccuracy: more than 30%Improved
apply contrast-enhancement two-stage methods (HE, CVC, WAHE, LDR, OpenCE) before classifyingVGG16, AlexNet · ImageNet-C contrast levels 1-2accuracyHurt
use FDM instead of contrast enhancementVGG16, AlexNet · ImageNet-C contrast, 5 levelsaccuracyImproved
apply HDR methods (LIME, HQEC, SRIE, MRINP) before classifyingVGG16, AlexNet · ImageNet-C brightnessaccuracyHurt
use FDM instead of HDR enhancementVGG16, AlexNet · ImageNet-C brightnessaccuracyImproved

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