Code for Differential-Critic GAN: Generating What You Want by a Cue of Preferences (TNNLS 2022)
- python3
- pytorch
- NumPy, SciPy, Matplotlib
- Pretrain mnist classifier:
- Under ./cls_mnist dir: run
python cls_mnist.py --save-model
- Expected Output
- mnist_cnn.pt
- Expected Output
- Pretrain WGAN:
- Run
python gan_mnist.py
- Expected Output
results/pretrain
- Expected Output
- Train DiCGAN:
- Run
python dicgan_mnist.py
- Expected Output
results/$RUN_NAME/0/positive.txt
records the percentage of desired samples at each epochresults/$RUN_NAME/0/samples
contain sampled outputs from generator- Default
$RUN_NAME
isdicgan_small_digits
- Expected Output
@article{yao2022differential,
title={Differential-Critic GAN: Generating What You Want by a Cue of Preferences},
author={Yao, Yinghua and Pan, Yuangang and Tsang, Ivor W and Yao, Xin},
journal={IEEE Transactions on Neural Networks and Learning Systems},
year={2022},
publisher={IEEE}
}