Skip to content

Latest commit

 

History

History
7 lines (6 loc) · 740 Bytes

README.md

File metadata and controls

7 lines (6 loc) · 740 Bytes
Raw

BGAN

To run this codes, you should do follow things:

  1. extract resnet feature and then run create_S.py to construct similarity matrix.
    You can download cifar-10.mat (here) training data and cifar_KNN.npz from (here)
  2. download vgg19 pretrained model on ImageNet based tensorflow here .
  3. run this command 'python BGAN.py 32' to train network and then generate 32 bit codes
  4. after training done, you can run evalutate.py to calculate MAP. HIT: you need to change some paths in evalutate.py.