https://anomalygpt.github.io/

see the examples first.

1. AnomalyGPT eliminates the need for manual threshold adjustments.
   1. 通过合成异常数据，训练一个segmentation网络，岂不是更好。
   2. 可以更文本的描述anomaly的color, shape, and categories；
2. With only one normal shot, AnomalyGPT achieves the state-of-the-art performance with an accuracy of 86.1%, an image-level AUC of 94.1%, and a pixel-level AUC of 95.3% on the MVTec-AD dataset.
   1. this is interesting

Large Vision-Language Models (LVLMs) such as MiniGPT-4 and LLaVA

Industrial Anomaly Detection (IAD)

generate training data by simulating anomalous images and producing corresponding textual descriptions for each image.

AnomalyGPT: Detecting Industrial Anomalies using Large Vision-Language Models, 23