qqsuhao / anogan-mvtec-grid-

pytorch

AnoGAN-MvTec(grid)

Introduction

I only try to reconstruction texture samples in MvTec by AnoGAN, but not very successfully.

Even so, I still push my relative code with hopeness to help someone.

• My reconstruction results
• 
• 
• 

You Need Following Packages

pytorch；torchvision； opencv4 ；tensorboard； tensorflow(it's necessary if want use tensorbard )

torchsummary； numpy； tqdm； argparse； PIL

If you do not want install tensorboard and tensorflow, you can annotate relative code.

Structure

• dataload
  • dataload.py get dataset
  • mvtec.py load images of mvtec
  • random_cut.py data argumentation by cutting sub-images randomly
• experiments save experiments results
  • size-64-epoch-100 pretrained weights when sub-images is 64*64 and epoches is 100
  • size-128-epoch-200 pretrained weights when sub-images is 128*128 and epoches is 200
  • anogan_train results during training
  • anogan_test results during testing(inference)
• models
  • CNN_anogan.py DCGAN with input size 128128; if you want to train the model with input size 6464, you need modify the model simply.
• runs save logs generated by tensorboard
• test infernece(testing) code
• train training code

anogan-mvtec-grid-'s Issues

模型训练问题

请教一下，如果要训练输入大小为 64×64 的模型，需修改模型的哪里呢