This is a nice work, but I have some confusion about the auxiliary dataset that I hope you can answer. It seems there is a bit of a contradiction here. I tried to reproduce the results. i found that using mvtec and colondb as auxiliary data is closer to the paper's results. I hope the author can clarify my doubts, thank you so much.
Thank you so much for your rigorous check! I have checked again and found that you are right, we use MVTec AD & Colondb for the default training. I have updated the repo for clarification.
This is a nice work, but I have some confusion about the auxiliary dataset that I hope you can answer. It seems there is a bit of a contradiction here. I tried to reproduce the results. i found that using mvtec and colondb as auxiliary data is closer to the paper's results. I hope the author can clarify my doubts, thank you so much.