gudovskiy/cflow-ad

why taking dataloader and training times very slow?

ingbeeedd opened this issue · 2 comments

ingbeeedd commented

Hello, running the command below.

python3 main.py --gpu 0 --pro -inp 512 --dataset mvtec --class-name bottle

Can you tell me why it takes more time to train than the existing Anomaly algorithm and why it takes more time to bring dataloaders? Thank you for sharing code

gudovskiy commented

@ingbeeedd try to

  1. remove --pro estimation because it is slow CPU code
  2. decrease -inp input resolution to 256/128
gudovskiy commented
  1. use smaller feature extractor e.g. ResNet-18. For example:
    python3 main.py --gpu 0 -enc resnet18 -inp 256 --dataset mvtec --class-name bottle
  • 👍1