Learning What and Where to Draw
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.

README.md

###Learning What and Where to Draw Scott Reed, Zeynep Akata, Santosh Mohan, Samuel Tenka, Bernt Schiele, Honglak Lee

This is the code for our NIPS 2016 paper on text- and location-controllable image synthesis using conditional GANs. Much of the code is adapted from reedscot/icml2016 and dcgan.torch.

####Setup Instructions

You will need to install Torch, CuDNN, stnbhwd and the display package.

####How to train a text to image model:

  1. Download the data including captions, location annotations and pretrained models.
  2. Download the birds and humans image data.
  3. Modify the CONFIG file to point to your data.
  4. Run one of the training scripts, e.g. ./scripts/train_cub_keypoints.sh

####How to generate samples:

  • ./scripts/run_all_demos.sh.
  • html files will be generated with results like the following:

Moving the bird's position via bounding box:

Moving the bird's position via keypoints:

Birds text to image with ground-truth keypoints:

Birds text to image with generated keypoints:

Humans text to image with ground-truth keypoints:

Humans text to image with generated keypoints:

####Citation

If you find this useful, please cite our work as follows:

@inproceedings{reed2016learning,
  title={Learning What and Where to Draw},
  author={Scott Reed and Zeynep Akata and Santosh Mohan and Samuel Tenka and Bernt Schiele and Honglak Lee},
  booktitle={Advances in Neural Information Processing Systems},
  year={2016}
}