PromptAD Robustness Evaluation

Overview

This repository extends the PromptAD method, which leverages prompt tuning to adapt vision-language models (VLMs) for few-shot anomaly detection.

In this work, we evaluate the robustness of PromptAD by testing its performance under various image corruptions. Our analysis provides insights into the method’s stability and reliability in real-world scenarios where data quality may be degraded.

References

• PromptAD GitHub Repository: https://github.com/FuNz-0/PromptAD