About this document
AnomalyControl: Few-Shot Image Generation by Kamilla is a document available to read on EtoBox.
The paper presents AnomalyControl, a framework that utilizes ControlNet inpainting with stable diffusion to generate realistic defective images for few-shot anomaly detection in industrial quality inspection. By addressing the scarcity of defective images, the proposed method demonstrates significant improvements in anomaly classification and detection on the MVTec-AD dataset. The approach combines advanced generative techniques and a filtering method to enhance the quality of generated anomalies, setting a
- Author
- Kamilla
- Language
- EN