Data results with reproducibility

[lixinying0727]

I would like to ask that if I follow the code in the project to reproduce, the result shows that the data mvtec both image level and pixel level results are not as high as AnomalyCLIP's data, and the visa also only has an improvement in PRO metrics, but the others are still lower than AnomalyCLIP.

[YUL-git]

Hello, Thank you for your interest in my work. In my experimental setup, the results are reproduced as expected. The outcomes may vary slightly depending on the hyperparameters, so I recommend experimenting with different settings.

In my reproduction of other papers, the results also varied slightly from the original paper. Thank you for your understanding.

[lixinying0727]

Hello, Thank you for your interest in my work. In my experimental setup, the results are reproduced as expected. The outcomes may vary slightly depending on the hyperparameters, so I recommend experimenting with different settings.

In my reproduction of other papers, the results also varied slightly from the original paper. Thank you for your understanding.

Thank you very much for your reply, I will try to change the hyperparameters and continue to experiment to achieve satisfactory results!