search

tientrandinh / Revisiting-Reverse-Distillation

(CVPR 2023) Revisiting Reverse Distillation for Anomaly Detection
MIT License
116 stars 25 forks source link

Why the target in loss_contrast is 1? #1

Closed hitlei closed 1 year ago

hitlei commented 1 year ago

image

to push away abnormal information from projected normal space, should target in this code be -1?

tientrandinh commented 1 year ago

Hi, You are correct that the 'target' should be -1; it was a typo during my code refactoring process. I have already updated it. The repository is gradually being cleaned and uploaded, as it has not been officially completed yet. Thank you so much for the prompt notice.