AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

DiffusionAD: Norm-Guided One-Step Denoising Diffusion for Anomaly Detection

Hui Zhang, Zheng Wang, Dan Zeng, Zuxuan Wu, Yu–Gang Jiang

TPAMI 2025 · Image generation › Autoregressive and latent generation · approach: Reconstruction-based diffusion image detection · 197 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 auto-encoder reconstruction with diffusion noise-to-norm reconstruction (ablation)DiffusionAD segmentation pipeline · VisAImage AUROC: 90.6 -> 96.8Improved
replace auto-encoder reconstruction with diffusion noise-to-norm reconstruction (ablation)DiffusionAD segmentation pipeline · VisAPixel PRO: 76.1 -> 92.5Improved
use one-step denoising instead of 400-step iterative denoising (ablation)DiffusionAD · VisAImage AUROC: 96.8 -> 98.0Improved
use one-step denoising instead of 400-step iterative denoising (ablation)DiffusionAD · VisAFPS: 0.09 FPS -> 26.7 FPSImproved
use one-step denoising instead of DDIM-30 denoising (ablation)DiffusionAD · VisAImage AUROC: 96.6 -> 98.0Improved
add norm-guided denoising on top of one-step denoising (ablation)DiffusionAD · VisAImage AUROC: 98.0 -> 98.8Improved
add norm-guided denoising on top of one-step denoising (ablation)DiffusionAD · VisAPixel PRO: 94.7 -> 96.0Improved
add norm-guided denoising on top of one-step denoising (ablation)DiffusionAD · VisAPixel AP: 44.2 -> 45.8Improved
add norm-guided denoising on top of one-step denoising (ablation)DiffusionAD · VisAFPS: 26.7 -> 23.5Hurt
use iterative-400 inference after one-step training (train-inference mismatch) (ablation)DiffusionAD · VisAImage AUROC: 98.0 -> 96.8Hurt
train segmentation on iterative-400 denoised outputs and infer iteratively (ablation)DiffusionAD · VisAImage AUROC: 98.0 -> 98.3Improved
raise noise scale t to 500 or more for one-step denoising (ablation)DiffusionAD · VisAImage AUROC: decrease at t >= 500Hurt

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