LayoutGAN in Tensorflow
Official Tensorflow implementation of "LayoutGAN: Generating Graphic Layouts with Wireframe Discriminators" publishsed in ICLR 2019: https://openreview.net/forum?id=HJxB5sRcFQ.
Some codes are implemented from https://github.com/carpedm20/DCGAN-tensorflow.
This project is licensed under the terms of the “Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International” license.
Online Demo
Animation videos to demonstrate the movements of all the graphic elements.
- MNIST
- Tangram
Prerequisites
- Python 2.7
- Tensorflow 1.2.0
Usage
First, download the trasformed point layout representation of MNIST dataset from
https://drive.google.com/file/d/1R1iRZxADR_RcDsuR4gyStyLAo7i5LRAH/view?usp=sharing,
and put it under ./data directory.
To train a model with downloaded dataset:
$ bash ./experiments/scripts/train_mnist.sh
For bounding box layout experiments, you may refer to the commented lines in the code.
Results on MNIST
