Skip to content

Simple Tensorflow implementation of "Unsupervised Image to Image Translation Networks" (NIPS 2017 Spotlight)

License

Notifications You must be signed in to change notification settings

taki0112/UNIT-Tensorflow

Repository files navigation

UNIT-Tensorflow

Simple Tensorflow implementation of "Unsupervised Image to Image Translation Networks" (NIPS 2017 Spotlight)

Requirements

  • Tensorflow 1.4
  • Python 3.6

Usage

├── dataset
   └── YOUR_DATASET_NAME
       ├── trainA
           ├── xxx.jpg (name, format doesn't matter)
           ├── yyy.png
           └── ...
       ├── trainB
           ├── zzz.jpg
           ├── www.png
           └── ...
       ├── testA
           ├── aaa.jpg 
           ├── bbb.png
           └── ...
       └── testB
           ├── ccc.jpg 
           ├── ddd.png
           └── ...
> python main.py --phase train --dataset cat2tiger
  • See main.py for other arguments
  • If you want to multi_gpu_version, then use main_multi_gpu.py (batch_size = The batch_size per gpu)
  • If you want to faster_UNIT, then use DatasetAPI (code is more simple !)

Issue

Too much Slow !!!

  • The slower reason is that it stores checkpoints
  • If you want to speed up, do not save checkpoints per iteration

Arichitecture

architecture

Framework

framework

Model

compare

vae

gan

cycle

Training Objective

objective

Result

Success

success

Fail

fail

Related works

Reference

Author

Junho Kim

About

Simple Tensorflow implementation of "Unsupervised Image to Image Translation Networks" (NIPS 2017 Spotlight)

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages