FB-CLIP: Fine-Grained Zero-Shot Anomaly Detection with Foreground-Background Disentanglement

1. 1Utilize vision-language models like CLIP as a foundation
2. 2Implement foreground-background disentanglement to reduce feature entanglement
3. 3Employ multi-strategy textual representations for fine-grained anomaly detection
4. 4Evaluate and refine the FB-CLIP framework for specific use cases

AI engineers and researchers on a team can benefit from FB-CLIP for improving zero-shot anomaly detection, while data scientists can apply the framework to industrial and medical applications

💡 Foreground-background disentanglement improves zero-shot anomaly detection

💡 Enhance anomaly detection with FB-CLIP's foreground-background disentanglement