性能不如WinCLIP ann CLIP-AD。

只看看question即可。

A toy experiment for using GPT-4V. This leads to uncontrollable and imprecise outputs.

1. region division: Mi;

2. Prompt Designing: Ti, this i is different with the i in Mi.

3)Text2Segmentation

answer To to pixel-level anomaly segmentation result Ao.

明明是region level呀；

其实啥也没写。

Exploring Grounding Potential of VQA-oriented GPT-4V for Zero-shot Anomaly Detection, 23