Research map / Label-efficient and robust learning / OOD and anomaly detection
MoEAD: A Parameter-efficient Model for Multi-class Anomaly Detection
Shiyuan Meng, Wenchao Meng, Qihang Zhou, Shizhong Li, Weiye Hou, Shibo He
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 top-2 routing instead of top-1 in the MoE layer (ablation) | MoEAD · MVTec-AD | I-AUROC: 96.8 -> 97.7 | Improved |
| increase top-k beyond 2 (ablation) | MoEAD · MVTec-AD | I-AUROC: 97.7 -> 97.6 | No clear effect |
| increase decoder blocks from 4 to 6 or 8 (ablation) | MoEAD · MVTec-AD | I-AUROC: 97.7 -> 97.7 (6), 97.6 (8) | No clear effect |
| increase decoder blocks from 2 to 4 (ablation) | MoEAD · MVTec-AD | I-AUROC: 97.5 -> 97.7 | Improved |
| stack two MoE layers of n experts instead of one layer of 2n experts (ablation) | MoEAD · MVTec-AD (n=4) | I-AUROC: 97.3 -> 97.7 | Improved |
| stack two MoE layers of n experts instead of one layer of n^2 experts (ablation) | MoEAD · MVTec-AD (n=4) | I-AUROC: 97.6 -> 97.7 | No clear effect |
| increase experts per SMoE layer from 2 to 5 (ablation) | MoEAD · MVTec-AD | I-AUROC: 97.1 -> 97.9 | Improved |
| replace FFN with MoE layer | MoEAD · MVTec-AD | I-AUROC | Improved |
| share only attention or only MoE layer instead of both | MoEAD · MVTec-AD | model size | No clear effect |
Later papers that compared with this method
1 later accepted paper reports MoEAD as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Toward Long-Tailed Online Anomaly Detection through Class-Agnostic Concepts (ICCV 2025 · better on 3 of 3 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
- AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · similarity 0.87)
- ResAD: A Simple Framework for Class Generalizable Anomaly Detection (NeurIPS 2024 · similarity 0.86)
- Anomaly Heterogeneity Learning for Open-set Supervised Anomaly Detection (CVPR 2024 · similarity 0.86)
- Long-Tailed Anomaly Detection with Learnable Class Names (CVPR 2024 · similarity 0.86)
- Towards Training-free Anomaly Detection with Vision and Language Foundation Models (CVPR 2025 · similarity 0.86)
- A Comprehensive Augmentation Framework for Anomaly Detection (AAAI 2024 · similarity 0.86)
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.
