Research map / Label-efficient and robust learning / OOD and anomaly detection
GeneralAD: Anomaly Detection Across Domains by Attending to Distorted Features
Luc P.J. Sträter, Mohammadreza Salehi, Efstratios Gavves, Cees G.M. Snoek, Yuki M. Asano
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use GeneralAD (patch-attention discriminator on DINOv2 features) instead of Transformaly | semantic and near AD · Aircraft-FGVC | image-level AUROC: 84.0 -> 94.6 | Improved |
| use GeneralAD instead of SimpleNet on logical anomalies | industrial AD · MVTec-LOCO | image-level AUROC: 77.6 -> 84.9 | Improved |
| use GeneralAD instead of SimpleNet on MVTec-AD | industrial AD · MVTec-AD | image-level AUROC: 99.6 -> 99.2 | No clear effect |
| use GeneralAD instead of FITYMI on Stanford-Cars | near AD · Stanford-Cars | image-level AUROC: 90.8 -> 87.3 | Hurt |
| swap DINOv2 backbone for ViT-B/16 in GeneralAD (ablation) | GeneralAD · CIFAR-10 | image-level AUROC: 99.1 -> 93.0 | Hurt |
| swap DINOv2 backbone for DINO-B/8 in GeneralAD (ablation) | GeneralAD · Aircraft-FGVC | image-level AUROC: 93.4 -> 75.2 | Hurt |
| replace backbone of SimpleNet with DINOv2-B (ablation) | SimpleNet · MVTec-AD | image-level AUROC: 99.6 -> 97.7 | Hurt |
| replace backbone of Transformaly with DINOv2-B (ablation) | Transformaly · MVTec-AD | image-level AUROC: 87.9 -> 80.9 | Hurt |
| use Noise All Patches instead of Noise Random Patches for semantic data (ablation) | GeneralAD · CIFAR-10 | image-level AUROC: 96.6 -> 99.3 | Improved |
| add attention shuffle distortion on semantic data (ablation) | GeneralAD · Aircraft-FGVC | image-level AUROC: 94.6 -> 74.5 | Hurt |
| add attention shuffle distortion for logical anomalies (ablation) | GeneralAD · MVTec-LOCO | image-level AUROC: 84.1 -> 84.9 | Improved |
| use Noise Random Patches instead of Noise All Patches for MVTec-AD (ablation) | GeneralAD · MVTec-AD | image-level AUROC: 99.0 -> 99.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 3 on most shared results.
- SimpleNet — SimpleNet: A Simple Network for Image Anomaly Detection and Localization (CVPR 2023) · better on 3 of 4 shared results
- Recontrast — ReContrast: Domain-Specific Anomaly Detection via Contrastive Reconstruction (NeurIPS 2023) · better on 1 of 4 shared results
- PANDA — PANDA: Adapting Pretrained Features for Anomaly Detection and Segmentation (CVPR 2021) · better on 0 of 1 shared result
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
- Towards Training-free Anomaly Detection with Vision and Language Foundation Models (CVPR 2025 · similarity 0.90)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.90)
- AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models (AAAI 2024 · similarity 0.90)
- Long-Tailed Anomaly Detection with Learnable Class Names (CVPR 2024 · similarity 0.89)
- Toward Generalist Anomaly Detection via In-context Residual Learning with Few-shot Sample Prompts (CVPR 2024 · similarity 0.89)
- AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · similarity 0.89)
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.
