search

ORippler / gaussian-ad-mvtec

Code underlying our publication "Modeling the Distribution of Normal Data in Pre-Trained Deep Features for Anomaly Detection" at ICPR2020
GNU Affero General Public License v3.0
99 stars 18 forks source link

Question about mask #4

Closed bartproo closed 1 year ago

bartproo commented 3 years ago

Hi, if masks are not required in the gaussian model, why do we need it in the test_step? Can I set load_masks to false?

ORippler commented 3 years ago

Hi,

we also tested around with anomaly segmentation (provided e.g. by Autoencoder), and need the masks for evaluating these methods (results didn't make it into the paper and padim beat us to gaussian anomaly segmentation)

you can try setting it to false, but I don't know what will happen exactly, see #2 for a better workaround