LGAN

This is the project for the following technical report:

Guo-Jun Qi, Liheng Zhang, Hao Hu. Global versus Localized Generative Adversarial Nets. arXiv: 1711.06020 [pdf]

Requirements

• Python == 2.7 or 3.5+
• Pytorch == 0.4.1

Train LGAN

python train.py

Test LGAN

python test.py

Result

After 20 epochs training. We can adjust noise z to control the output.In each row,the images are generated in one direction of a local coordinate.

Acknowledge

Parts of codes are reused from Localized-GAN at https://github.com/z331565360/Localized-GAN