AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Encoding Structure-Texture Relation with P-Net for Anomaly Detection in Retinal Images

Kang Zhou, Yuting Xiao, Jianlong Yang, Jun Cheng, Wen Liu, Weixin Luo, Zaiwang Gu, Jiang Liu, Shenghua Gao

ECCV 2020 · Label-efficient and robust learning › OOD and anomaly detection · approach: Unified multi-class anomaly detection · 154 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
add domain adaptation to the structure extractor (ablation)P-Net · RESC (OCT)AUC: 0.8152 -> 0.8518Improved
add domain adaptation to the structure extractor (ablation)P-Net · iSee (fundus)AUC: 0.5914 -> 0.7196Improved
feed both image and structure to the reconstruction generator instead of image only (ablation)P-Net · RESC (OCT)AUC: 0.8219 -> 0.8518Improved
feed both image and structure instead of structure only (ablation)P-Net · iSee (fundus)AUC: 0.6914 -> 0.7196Improved
add structure consistency loss (ablation)P-Net · RESC (OCT)AUC: 0.8518 -> 0.9288Improved
add structure consistency loss (ablation)P-Net · iSee (fundus)AUC: 0.7196 -> 0.7245Improved
combine image and structure differences (lambda_f=0.8) instead of image difference only (ablation)P-Net · RESC (OCT)AUC: 0.8481 -> 0.9288Improved
combine both differences (lambda_f=0.8) instead of structure difference only (ablation)P-Net · RESC (OCT)AUC: 0.9084 -> 0.9288Improved
use more reconstruction cycles at test time (1 to 4) (ablation)P-Net · RESC (OCT)AUC: 0.9288 -> 0.9380Improved
train with 2 cycles instead of 1 (ablation)P-Net · RESC (OCT)AUC: 0.9288 -> 0.8935Hurt
use Canny edges as structure on MVTec AD texture categoriesP-Net · MVTec AD texturesAUCMixed
P-Net vs GANomaly on AMD sub-classP-Net · iSee AMDAUC: 0.5713 -> 0.5688Hurt

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