Gogumee / DRPAN

DRPAN: Discriminative Region Proposal Adversarial Networks for High-Quality Image-to-Image Translation

This is our implementation of DRPAN.

Prerequisites

• Linux or OSX.
• Python 2 or Python 3.
• CPU or NVIDIA GPU + CUDA CuDNN.

Getting Started

Installation

• Install PyTorch and dependencies from http://pytorch.org/

• Clone this repo:

git clone https://github.com/godisboy/DRPAN.git
cd DRPAN

• Prepare your paired image2image translation dataset:

• Modify the config file: change the dataPath to your data set path

DRPAN train

• Train a model:

python main.py --config configs/facades.yaml --cuda --gpu_ids 0

StackGan-like model train

python train_stack_pix2pix.py --config configs/facades.yaml --cuda --gpu_ids 0

Paper

Chao Wang, Haiyong Zheng, Zhibin Yu, Ziqiang Zheng, Zhaorui Gu, Bing Zheng. "Discriminative Region Proposal Adversarial Networks for High-Quality Image-to-Image Translation", ECCV 2018 (to appear)

Acknowledgments

Code is inspired by pytorch-CycleGAN-and-pix2pix and StackGAN-v2.