unoffical code of yout local gan based ylg(https://github.com/zhushaoquan/ylg) and sagan
just for UESTC-IVIPC Zhao Jinagwei experiment if you want to use it or find error in it ,you could tell that to me . thanks. different:
- add mask don't have ESA
- chang model to light for cifar10
self-attention-GAN-pytorch This is an almost exact replica in PyTorch of the Tensorflow version of SAGAN released by Google Brain [repo] in 8月2018日. Code structure is inspired from this repo, but follows the details of Google Brain's repo.
$ python train.py --data_path 'a/b/c' --save_path 'o/p/q' --batch_size 64 --name sagan
Model training will be recorded in a new folder inside --save_path
with the name <timestamp>_<name>_<basename of data_path>
. By default, model weights are saved in a subfolder called weights
, and train & validation samples during training in a subfolder called samples
(can be changed in parameters.py
). ## Testing/Evaluating Check test.py
.
@article{Zhang2018SelfAttentionGA, title={Self-Attention Generative Adversarial Networks}, author={Han Zhang and Ian J. Goodfellow and Dimitris N. Metaxas and Augustus Odena}, journal={CoRR}, year={2018}, volume={abs/1805.08318}
}