Research map / Vision-language and multimodal / CLIP and zero-shot
DeCo: Zero-Shot Anomaly Generation through Decoupling and Recoupling
Shilei Zeng, Xurui Li, Yaohan Tang, Yu 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 AP-Routing to unified LoRA baseline (ablation) | DeCo · MVTec AD downstream detection | P-AP: 62.7 -> 63.2 | Improved |
| add QK-Routing after AP-Routing (ablation) | DeCo · MVTec AD downstream detection | P-AP: 63.2 -> 66.7 | Improved |
| add PI-Flow (ablation) | DeCo · MVTec AD downstream detection | P-AP: 66.7 -> 67.6 | Improved |
| add PCC (ablation) | DeCo · MVTec AD downstream detection | P-AP: 67.6 -> 68.0 | Improved |
| use full-rank A-LoRA instead of rank 64 (ablation) | DeCo · MVTec AD downstream detection | P-AP: 65.2 -> 68.0 | Improved |
| add AP-Routing text token split (vs w/o Routing) (ablation) | DeCo · MVTec AD downstream detection | P-AP: 66.4 -> 68.0 | Improved |
| add A-Routing (vs w/o A-Routing) (ablation) | DeCo · MVTec AD downstream detection | I-AUC: 96.9 -> 98.0 | Improved |
| add P-Routing (vs w/o P-Routing) (ablation) | DeCo · MVTec AD downstream detection | P-AP: 67.4 -> 68.0 | Improved |
| restrict A-LoRA to Q only (vs Symmetric) (ablation) | DeCo · MVTec AD downstream detection | P-AP: 63.1 -> 68.0 | Improved |
| restrict A-LoRA to Q only (vs Hybrid) (ablation) | DeCo · MVTec AD downstream detection | P-AP: 64.3 -> 68.0 | Improved |
| add PI-Flow loss weight 0.1 (vs 0) (ablation) | DeCo · MVTec AD downstream detection | P-AP: 67.2 -> 68.0 | Improved |
| raise PI-Flow weight from 0.1 to 1.0 (ablation) | DeCo · MVTec AD downstream detection | P-AP: 68.0 -> 67.4 | Hurt |
Closest papers by meaning
- Dual-Interrelated Diffusion Model for Few-Shot Anomaly Image Generation (CVPR 2025 · similarity 0.89)
- Exploring Intrinsic Normal Prototypes within a Single Image for Universal Anomaly Detection (CVPR 2025 · similarity 0.89)
- AnomalyDiffusion: Few-Shot Anomaly Image Generation with Diffusion Model (AAAI 2024 · similarity 0.88)
- AnomalyNCD: Towards Novel Anomaly Class Discovery in Industrial Scenarios (CVPR 2025 · similarity 0.88)
- Unseen Visual Anomaly Generation (CVPR 2025 · similarity 0.87)
- MuSc: Zero-Shot Industrial Anomaly Classification and Segmentation with Mutual Scoring of the Unlabeled Images (ICLR 2024 · similarity 0.87)
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.
