Closed this issue 10 months ago · 2 comments
When I change the Batch Size during testing I get different results What is the reason? Thanks
In the paper's Eq. 9, we normalize over the batch to compute the final anomaly score. We fount it more stable than normalizing per image.
thanks