ResidualGAN

Tensorflow implementation of Learning Residual Images for Face Attribute Manipulation

Prerequisites

• Python 3.6
• Tensorflow 1.6

Results for Facial Attribute Transfer on CelebA

The images are generated by ResidualGAN trained on the CelebA dataset.

Usage

1. Clone the repository

$ git clone https://github.com/ioanachelu/ResidualGAN.git
$ cd ResidualGAN/

2. Download the dataset into ../data. It should have the following substructure:

    .data
        celebA
        list_attr_celebs.txt

3. Train ResidualGAN with CelebA

$ python train.py --train=True --logdir=./logdir --config=default --dataset_path=../data

4. Resume ResidualGAN training with CelebA

$ python train.py --train=True --resume=True --load_from=./logdir/0-std_gan --config=std_gan --dataset_path=../data