AI Paper Atlas

Research map / Image generation / Super-resolution

Anchored Regression Networks Applied to Age Estimation and Super Resolution

Eirikur Agustsson, Radu Timofte, Luc Van Gool

ICCV 2017 · Image generation › Super-resolution · approach: Real-world super-resolution · 60 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 DEX classification layer with ARN regression layer (101 anchors)DEX pipeline · ChaLearn LAP validation (DEX IMDB-WIKI pretraining)MAE (years): 3.252 -> 3.153Improved
replace DEX output layer with ARN (10 anchors)DEX pipeline · ChaLearn LAP validation (DEX IMDB-WIKI pretraining)MAE (years): 3.505 -> 3.237Improved
replace DEX output layer with ARN (101 anchors) starting from ImageNet VGG-16DEX pipeline · ChaLearn LAP validation (VGG-16 ImageNet pretraining)MAE (years): 5.965 -> 4.947Improved
replace DEX with ARN (101 anchors)DEX pipeline · MORPH 2MAE (years): 3.25 -> 3.00Improved
replace DEX with ARN when using IMDB-WIKI pretrained modelDEX pipeline · MORPH 2 (IMDB-WIKI pretrained)MAE (years): 2.68 -> 2.63Improved
use 10 anchors instead of 101 when starting from ImageNet VGG-16 (ablation)ARN · ChaLearn LAP validation (VGG-16 ImageNet pretraining)MAE (years): 4.947 -> 4.516Improved
reduce DEX output neurons below 10 (ablation)DEX · ChaLearn LAP validationMAE vs number of outputsHurt
vary number of regressors in ARN (ablation)ARN · ChaLearn LAP validationMAE vs number of regressorsNo clear effect
replace A+ with ARN (1024 regressors)A+ · Set14 x3 SRPSNR (dB): 29.13 -> 29.31Improved
replace A+ with ARN (1024 regressors)A+ · Set5 x3 SRPSNR (dB): 32.59 -> 32.89Improved
learn anchors jointly by gradients instead of closed-form global solution over fixed anchors (ablation)ARN · Set5 x3 SRPSNR (dB): 32.69 -> 32.89Improved
stack layers (Deep ARN, 7 x 16 regressors) instead of single ARN layer with 1024 regressors (ablation)ARN · Set14 x3 SRPSNR (dB): 29.31 -> 29.37Improved

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