AI Paper Atlas

Research map / Image generation / Autoregressive and latent generation

InvAD: Inversion-based Reconstruction-Free Anomaly Detection with Diffusion Models

Shunsuke Sakai, Xiangteng He, Chunzhi Gu, Leonid Sigal, Tatsuhito Hasegawa

CVPR 2026 · Image generation › Autoregressive and latent generation · approach: Reconstruction-based diffusion image detection · 7 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 multi-step inversion (instead of single-step) (ablation)InvAD with FDM · MVTecADmAD: 71.0 -> 83.7Improved
add inversion to feature-space diffusion (ablation)InvAD · MVTecADmAD: 57.3 -> 83.7Improved
add feature-space diffusion to single-step inversion (ablation)InvAD · MVTecADmAD: 44.9 -> 71.0Improved
add NLL+Diff scoring instead of NLL alone at S=1000 (ablation)InvAD · MVTecADimage AU-ROC: 89.1 -> 95.4Improved
add Diff term to NLL at S=3 (ablation)InvAD scoring · MVTecADimage AU-ROC: 96.1 -> 99.0Improved
add NLL term to Diff at S=1000 (ablation)InvAD scoring · MVTecADimage AU-ROC: 82.0 -> 95.4Improved
use inversion instead of reconstruction at S=3 (ablation)diffusion AD · MVTecADimage AU-ROC: 64.9 -> 99.0Improved
increase diffusion steps S from 3 to 1000 (ablation)InvAD · MVTecADimage AU-ROC: 99.0 -> 95.4Hurt
plug InvAD into DiADDiAD · MVTecADimage AU-ROC / FPS: 97.2 -> 98.2; 0.1 -> 88.1Improved
plug InvAD into MDMMDM · MVTecADimage AU-ROC / FPS: 91.9 -> 98.2; 2.2 -> 63Improved
use EfficientNet-B4 instead of ViT-B feature extractor (ablation)InvAD · MVTecADimage AU-ROC: 94.2 -> 99.0Improved
replace DiT-gigant with MLP diffusion network (ablation)InvAD · MVTecADimage AU-ROC: 99.0 -> 97.2No clear effect

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