/Fila-sGAN-pytorch

My implementation of 《Synthesizing Filamentary Structured Images with GANs》

Primary LanguagePython

Fila-sGAN-pytorch

Author: Yuchao Gu

E-mail: 2015014178@buct.edu.cn

Date: 2018-05-27

Description: The code is an pytorch implementation of 《Synthesizing Filamentary Structured Images with GANs》

Overview

Data and Preprocess

DRIVE: Digital Retinal Images for Vessel Extraction you can download the train data here. For generating more data for data augmentation, you can random crop some labels from other dataset, and put one image for style transfer in eyedata/train/style.

Model

Training

python train.py

Results

left is the style image and others are my results.

How to use

Dependencies

This code depends on the following libraries:

  • Python 3.6
  • Pytorch
  • PIL

structure

 filps-gan
│
├── eyedata  # drive data
│ 
├── gycutils # my utils for data augmentation
│ 
├── Vgg.py # vgg net to style transfer
│ 
├── readmeDisplay  # some images for readme
│ 
├── datasets.py # dataset for dataloader
│ 
├── gan.py # generative adversial network for synthesize retina image
│ 
├── train.py # train code
│
├── transform.py 
│
└── readme.md # introduce to this project