Paper Link: https://openreview.net/forum?id=BJy0fcgRZ
Referenced Paper Link: https://arxiv.org/abs/1605.09782
In the paper writers used 2 GAN types which are DCGAN and BiGAN classify images. DCGAN wass used for creating happy, sad, male, female faces. BiGAN, which employed from Adversarial Feature Learning, was ran on Imagenet ILSVRC 2012 dataset to classify images.
- ILSVRC 2012 dataset should be downlaoded and extracted under data/imagenet/
- Python2
- theano
- pytorch
- scipy
- matplotlib
- numpy
- joblib
- lazy_python
- flask
- itertools
- OrderedDict
- data
- argparse
- rescale
To train the model from strach, run train_imagenet.sh
To evaluate the model, run eval_model.sh