Skip to content

Latest commit

 

History

History
91 lines (59 loc) · 1.71 KB

README.md

File metadata and controls

91 lines (59 loc) · 1.71 KB

Conditional-DRAGAN

Conditional DRAGAN(cGAN & WGAN based) code(& model) repo.

  • Generator 4-layer-512-ReluMLP(last layer is tanh)
  • Discriminator 4-layer-512-ReluMLP.

NOTE: Not include any normalization.

Question is always welcome, Pls add some issue.

conditional method

Its cGAN. Concat Noise z and Label c(onehot), then linear layer(Fully-Coneccted Layer) handle it.

Train

python cdragan -fmnist --gpu 0 --epoch 100 --out result


QuickVisualize with Docker

clone repo and run as follows:

nvidia-docker build -t rizaudo/cdragan-vis .
nvidia-docker run -it -v (pwd):/dic rizaudo/cdragan-vis /usr/bin/python3 /dic/visualizer.py -o /dic/visu --gendir /dic/models/gen_iter_18000.npz --allnum --fmnist
ls ./visu/preview/[GENERATED NEW IMAGE]

Label

Label Description
0 T-shirt/top
1 Trouser
2 Pullover
3 Dress
4 Coat
5 Sandal
6 Shirt
7 Sneaker
8 Bag
9 Ankle boot

Generated Images

(model is gen_iter_18000.npz)

  • T-Shirt/top

T-shirt/top(label:0)

  • Trouser

Trouser(label:1)

  • pullover

pullover(label:2)

  • Dress

Dress(label:3)

  • Coat

Coat(label:4)

  • Sandal

Sandal(label:5)

  • Shirt

Shirt(label:6)

  • Sneaker

Sneaker(label:7)

  • Bag

Bag(label:8)

  • Ankle boot

Ankle boot(label:9)

LICENSE

This materials provided as MIT License(Please see LICENSE file).