Predicted mask masked everything eventhough I tested normal test dataset.

[papago2355]

Hi.
I am currently using Padim with classification tasks. I am using a custom dataset which is about a plastic belt.
The dataset is originally so big(4000x2000), so I resized it into 1000x500z
At Training results, I got 0.94 auc and 0.93 f1 score.
But when I tested normal test dataset which are almost identical to normal train dataset pred mask covers everything and always returns anomaly as ‘True’
I checked my inference code. I resized my test images properly. Is there anyone know how to solve this issue?

Here is a link who has a similar issue.
#367