Research map / Detection and segmentation / Open-set and few-shot detection
LogiCo: A Unified Framework for Logical and Structural Anomaly Detection
Ximiao Zhang, Min Xu, Xiuzhuang Zhou
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 |
|---|---|---|---|
| add CRN (component-level reconstruction) to SRN (ablation) | LogiCo · MVTec-LOCO | logical I-AUC: 75.34 -> 89.43 | Improved |
| add SRN to CRN (ablation) | LogiCo · MVTec-LOCO | structural I-AUC: 84.59 -> 94.94 | Improved |
| add cross-attentive reconstruction to SRN (ablation) | LogiCo · MVTec-LOCO | structural sPRO: 81.98 -> 84.71 | Improved |
| add SMD (segmentation map discriminator) (ablation) | LogiCo · MVTec-LOCO | logical I-AUC: 90.05 -> 96.53 | Improved |
| add SMD (segmentation map discriminator) (ablation) | LogiCo · MVTec-LOCO | logical sPRO: 68.08 -> 63.63 | Hurt |
| use all four synthesis strategies instead of single (ablation) | LogiCo · MVTec-LOCO | logical I-AUC: 93.41 -> 96.53 | Improved |
| use final layer only instead of early layers (ablation) | LogiCo · MVTec-LOCO | I-AUC: 93.33 -> 96.28 | Improved |
| increase image size 256 to 512 with DINOv3 ViT-B (ablation) | LogiCo · MVTec-LOCO | I-AUC: 94.78 -> 96.28 | Improved |
| scale backbone from ViT-B to ViT-L (ablation) | LogiCo · MVTec-LOCO | I-AUC: 96.28 -> 95.80; sPRO 72.90 -> 73.88 | Mixed |
| use SAM-refined component segmentation | LogiCo · MVTec-LOCO | average sPRO: 72.9 -> 80.4 | Improved |
| compare LogiCo vs structural SOTA Dinomaly on MVTec-AD | LogiCo · MVTec-AD | I-AUC: 99.6 vs 99.6 | No clear effect |
| compare LogiCo vs Dinomaly on VisA | LogiCo · VisA | I-AUC: 98.9 -> 98.5 | Hurt |
Closest papers by meaning
- Few Shot Part Segmentation Reveals Compositional Logic for Industrial Anomaly Detection (AAAI 2024 · similarity 0.90)
- Towards Training-free Anomaly Detection with Vision and Language Foundation Models (CVPR 2025 · similarity 0.87)
- LogicAD: Explainable Anomaly Detection via VLM-based Text Feature Extraction (AAAI 2025 · similarity 0.86)
- Towards Generalizable 3D Anomaly Detection via Relational Inconsistency Modeling (NeurIPS 2026 · similarity 0.85)
- AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · similarity 0.85)
- Dinomaly: The Less Is More Philosophy in Multi-Class Unsupervised Anomaly Detection (CVPR 2025 · similarity 0.85)
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.
