dcgan-tensorflow

Version

• tensorflow 1.4.0
• python 3.5

Results

loss

generating images

the model converges rapidly, but it's easy to collapse

• iteration == 1000

  

• iteration == 25000

  

• iteration == 50000

  

• iteration == 96000

  

• iteration == 97000

  

• iteration == 98000

  

Refereces

1. Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks[J]. arXiv preprint arXiv:1511.06434, 2015.
2. https://github.com/carpedm20/DCGAN-tensorflow
3. Liu Z, Luo P, Wang X, et al. Deep learning face attributes in the wild[C]//Proceedings of the IEEE International Conference on Computer Vision. 2015: 3730-3738.