search

gudovskiy / cflow-ad

Official PyTorch code for WACV 2022 paper "CFLOW-AD: Real-Time Unsupervised Anomaly Detection with Localization via Conditional Normalizing Flows"
https://arxiv.org/abs/2107.12571
BSD 3-Clause "New" or "Revised" License
242 stars 61 forks source link

How many GPU hours are required? #16

Open caoyunkang opened 2 years ago

caoyunkang commented 2 years ago

Thanks for sharing the implementations!

I wonder how many GPU hours to train for a subset in MVTec? I've tried your code on a single RTX2060ti, but I find it really slow...

gudovskiy commented 2 years ago

@caoyunkang try to use MobileNetV3 feature extractor. It should be much faster.

gaowq2017 commented 2 years ago

I also have same issue especially in 'Compute loss and scores on test set'. It have easily unknow error when run to caculating the segmentation AUROC. BTW i used MobileNet_V3_large.

caoyunkang commented 2 years ago

Thanks. I'll try it. @gudovskiy