UPG-GAN

This is the code base for our paper Unpaired Pose-Guided Human Image Generation.We propose a new network architecture to generate human images from body part models, with unpaired training dataset.

Here you can find the necessary training and testing code, and the datasets and pre-trained models for shirt and tshirt (upper body) and suit and dress (full body).

Prerequisites

• Linux or macOS
• Python 2 or 3
• NVIDIA GPU + CUDA CuDNN

Getting Started

Installation

• Install PyTorch and dependencies from http://pytorch.org
• Install Torch vision from the source.

git clone https://github.com/pytorch/vision
cd vision
python setup.py install

• Install python libraries visdom and dominate.

pip install visdom
pip install dominate

• Clone this repo:

git clone git@github.com:cx921003/UPG-GAN.git
cd UPG-GAN

Data Preparation

• Download a dataset from our Google Drive.
• Unzip the dataset under ./datasets/ folder.

Pre-trained Models

• Download a pre-trained model from our Google Drive.
• Unzip the model under ./checkpoints/ folder.

Testing:

• Configure the following arguments in ./testing.sh:
  • dataroot: the path to the dataset
  • name: the name of the model, make sure the model exists under ./checkpoint/
  • how_many: number of input images to test
  • n_samples: number of samples per input image
• Test the model: ./testing.sh

Training

• Configure the following arguments in ./training.sh:
  • dataroot: the path to the dataset
  • name: the name of the model
• Train a model:./training.sh
• To view training results and loss plots, run python -m visdom.server and click the URL http://localhost:8097. To see more intermediate results, check out ./checkpoints/suit_and_dress/web/index.html

The test results will be saved to a html file here: ./results/suit_and_dress/latest_test/index.html.

Citation

If you find this repository useful for your research, please cite our paper.

to be added

Acknowledgments

Code is heavily based on pytorch-CycleGAN-and-pix2pix written by Jun-Yan Zhu and Taesung Park.