Research map / Vision-language and multimodal / CLIP and zero-shot
WinCLIP: Zero-/Few-Shot Anomaly Classification and Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 WinCLIP instead of CLIP-AC | CLIP anomaly classification · MVTec-AD, 0-shot | AUROC: 74.0 -> 91.8 | Helped |
| Use WinCLIP instead of CLIP-AC | CLIP anomaly classification · VisA, 0-shot | AUROC: 59.3 -> 78.1 | Helped |
| Use WinCLIP+ instead of PatchCore | Few-normal-shot anomaly classification · MVTec-AD, 1-shot | AUROC: 83.4 -> 93.1 | Helped |
| Use WinCLIP+ instead of PatchCore | Few-normal-shot anomaly classification · VisA, 1-shot | AUROC: 79.9 -> 83.8 | Helped |
| Use WinCLIP instead of MaskCLIP | Zero-shot anomaly segmentation · MVTec-AD, 0-shot | pAUROC: 63.7 -> 85.1 | Helped |
| Use WinCLIP instead of MaskCLIP | Zero-shot anomaly segmentation · VisA, 0-shot | pAUROC: 60.9 -> 79.6 | Helped |
| Use WinCLIP+ instead of PatchCore | Few-normal-shot anomaly segmentation · MVTec-AD, 1-shot | pAUROC: 92.0 -> 95.2 | Helped |
| Use WinCLIP+ instead of PatchCore | Few-normal-shot anomaly segmentation · VisA, 1-shot | pAUROC: 95.4 -> 96.4 | Helped |
| Add an anomalous-state prompt alongside the normal prompt (ablation) | CLIP anomaly classification · MVTec-AD, 0-shot | AUROC: 34.2 -> 74.0 | Helped |
| Add a state-word ensemble (ablation) | Two-class CLIP anomaly classification · MVTec-AD, 0-shot | AUROC: 74.0 -> 89.8 | Helped |
| Add a prompt-template ensemble (ablation) | CLIP anomaly classification with state ensemble · MVTec-AD, 0-shot | AUROC: 89.8 -> 90.8 | Helped |
| Add multi-crop prediction (ablation) | CLIP anomaly classification with compositional prompt ensemble · MVTec-AD, 0-shot | AUROC: 90.8 -> 91.8 | Helped |
Closest papers by meaning
- AA-CLIP: Enhancing Zero-Shot Anomaly Detection via Anomaly-Aware CLIP (CVPR 2025 · similarity 0.89)
- MuSc: Zero-Shot Industrial Anomaly Classification and Segmentation with Mutual Scoring of the Unlabeled Images (ICLR 2024 · similarity 0.88)
- AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection (ICLR 2024 · similarity 0.88)
- Few Shot Part Segmentation Reveals Compositional Logic for Industrial Anomaly Detection (AAAI 2024 · similarity 0.87)
- AnomalyGPT: Detecting Industrial Anomalies Using Large Vision-Language Models (AAAI 2024 · similarity 0.86)
- Zero-Shot Learning in Industrial Scenarios: New Large-Scale Benchmark, Challenges and Baseline (AAAI 2025 · 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.