This is the official repository to the WACV 2021 paper "Same Same But DifferNet: Semi-Supervised Defect Detection with Normalizing Flows" by Marco Rudolph, Bastian Wandt and Bodo Rosenhahn.
215
stars
67
forks
source link
Train loss might be negative during training, what does this situation mean? #44
During training, the loss might become negative. What does this situation mean?