AI Paper Atlas

Research map / Image generation / Image restoration

DNF: Decouple and Feedback Network for Seeing in the Dark

Xin Jin, Ling-Hao Han, Zhen Li, Chun-Le Guo, Zhi Chai, Chongyi Li

CVPR 2023 · Image generation › Image restoration · approach: Low-light image enhancement · 71 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
decouple RAW denoising from color restoration using clean RAW supervision (ablation)DNF · SID SonyPSNR: 30.48 -> 30.62Improved
replace clean RAW supervision with sRGB supervision (ablation)DNF · SID SonyPSNR: 30.20 -> 30.62Hurt
replace feedback with a direct single-stage RAW-to-sRGB cascade (ablation)DNF · SID SonyPSNR: 30.16 -> 30.62Hurt
replace feature-level feedback with image-level multi-stage cascading (ablation)DNF · SID SonyPSNR: 30.32 -> 30.62Hurt
use gated fusion modules for feedback features (ablation)DNF · SID SonyPSNR: 30.40 -> 30.62Improved
add gating to the feedback fusion module (ablation)DNF · SID SonyPSNR: 30.35 -> 30.62Improved
use GFM instead of SKFF for feature fusion (ablation)DNF · SID SonyPSNR: 30.37 -> 30.62Improved
use a global residual shortcut during RAW denoising (ablation)DNF CID block · SID SonyPSNR: 30.26 -> 30.32Improved
remove the global residual shortcut during color restoration (ablation)DNF CID block · SID SonyPSNR: 30.29 -> 30.32Improved
use the residual switch mechanism with a shared RAW encoder (ablation)DNF · SID SonyPSNR: 30.32 -> 30.62Improved
use DNF rather than a feature-level multi-stage framework with separate RAW encoders (ablation)DNF · SID SonyPSNR: 30.46 -> 30.62Improved
use DNF rather than MCRRAW-based LLIE · SID SonyPSNR: 29.65 -> 30.62Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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