Inquiry about Training Data and Text Descriptions for Anomaly Generation

[yjtlab]

Hi Jie Hu,

First of all, thank you for sharing your work on this project. I have a couple of questions regarding the training process and the generation of anomalies using the trained model:

Training Data: Could you please provide details on the datasets you used for training the model? Specifically, I'm interested in knowing the sources and types of data that were utilized. Do you use all of the MVTec dataset to train the diffusion model?

Text Descriptions for Anomaly Generation: When generating anomalies using the trained model, how do you choose or construct the text descriptions? Are there specific guidelines or examples you follow to ensure the text descriptions are effective for generating the desired anomalies?

Thank you for your time and assistance. I look forward to your response.