Text-to-image synthesis with self-supervised learning

This repository contains the pytorch code for the paper: "Text-to-image synthesis with self-supervised learning" Yong Xuan Tan, Chin Poo Lee, Mai Neo, Kian Ming Lim

Environment

The code is tested on Windows 10 with Anaconda3 and following packages:

• python 3.7.11
• pytorch 1.9.0

Dataset

1. Download the preprocessed char-CNN-RNN text embeddings for flowers and birds and put them into ./data/oxford and ./data/cub folder.

2. Download the flowers and birds images and put them into ./data/oxford and ./data/cub folder.

3. The dataset organization can be viewed in data folder.

Experiments

To train on CUB:

python main.py --dataset birds --exp_num cub_exp1

To evaluate on CUB:

python main.py --dataset birds --exp_num cub_exp1 --is_test true

Pre-trained Models

Download the pretrained models. Save them to the saved_model folder.

Examples generated by SSTIS:

Citation

If you find this repo useful for your research, please consider citing the paper:

@article{TAN2022119,
  title = {Text-to-image synthesis with self-supervised learning},
  journal = {Pattern Recognition Letters},
  volume = {157},
  pages = {119-126},
  year = {2022},
  issn = {0167-8655},
  doi = {https://doi.org/10.1016/j.patrec.2022.04.010},
  author = {Yong Xuan Tan and Chin Poo Lee and Mai Neo and Kian Ming Lim},
}

Our follow-up work

Text-to-image synthesis with self-supervised bi-stage generative adversarial network [code]

Contacts

For any questions, please contact:

Yong Xuan Tan ([email protected])
Jit Yan Lim ([email protected])

Acknowlegements

• Text-to-Image Synthesis
• StackGAN
• HDGAN
• SS-GAN

License

This code is released under the MIT License (refer to the LICENSE file for details).