Awesome

<p align="center"> DTLC-GAN </p>

Tensorflow implementation of DTLC-GAN (CVPR 2018): Generative Adversarial Image Synthesis with Decision Tree Latent Controller.

Usage

• Prerequisites

  • Tensorflow 1.9
  • Python 3.6

• Training

  • Important Arguments (See the others in train.py)
    • att: attribute to learn (default: '')
    • ks: # of outputs of each node of each layer (default: [2, 3, 3])
    • lambdas: loss weights of each layer (default: [1.0, 1.0, 1.0])
    • --n_d: # of d steps in each iteration (default: 1)
    • --n_g: # of g steps in each iteration (default: 1)
    • --loss_mode: gan loss (choices: [gan, lsgan, wgan, hinge], default: gan)
    • --gp_mode: type of gradient penalty (choices: [none, dragan, wgan-gp], default: none)
    • --norm: normalization (choices: [batch_norm, instance_norm, layer_norm, none], default: batch_norm)
    • --experiment_name: name for current experiment (default: default)
  • Example

    CUDA_VISIBLE_DEVICES=0 \
    python train.py \
    --att Eyeglasses \
    --ks 2 3 3 \
    --lambdas 1 1 1 \
    --n_d 1 \
    --n_g 1 \
    --loss_mode hinge \
    --gp_mode dragan \
    --norm layer_norm \
    --experiment_name att{Eyeglasses}_ks{2-3-3}_lambdas{1-1-1}_continuous_last{False}_loss{hinge}_gp{dragan}_norm{layer_norm}
    

Dataset

• Celeba dataset
  • Images should be placed in ./data/img_align_celeba/*.jpg
  • Attribute labels should be placed in ./data/list_attr_celeba.txt
  • the above links might be inaccessible, the alternatives are
    • img_align_celeba.zip
      • https://pan.baidu.com/s/1eSNpdRG#list/path=%2FCelebA%2FImg or
      • https://drive.google.com/drive/folders/0B7EVK8r0v71pTUZsaXdaSnZBZzg
    • list_attr_celeba.txt
      • https://pan.baidu.com/s/1eSNpdRG#list/path=%2FCelebA%2FAnno&parentPath=%2F or
      • https://drive.google.com/drive/folders/0B7EVK8r0v71pOC0wOVZlQnFfaGs

Exemplar Results

1. Eyeglasses, 3 layers

2. Bangs, 3 layers