Bangla Text to Image Generation

Pytorch implementaion of Attentional Generative Adversial Network (AttnGAN) for Bengali Language

Attention maps of text-to-image synthesis by our Attentional GAN model.

examples of text-to-image synthesis by our Attentional GAN model.

Sample generated bengali text to image results

"রঙিন পাখি যার একটি নীল মাথা মুখ এবং পেট সাদা এবং সাদা নীল পালক"	"ছোট পাখির একটি ছোট বিল কালো উইংবার সাদা বুক এবং একটি সোনার গলা রয়েছে"	"একটি ছোট পাখি যার উজ্জ্বল লাল মুকুট এবং পেট বাদামী ডানা এবং বাদামী গালের প্যাচ"

Dependencies

python 3.6+

Pytorch 1.0+

In addition, please add the project folder to PYTHONPATH and pip install the following packages:

• python-dateutil
• easydict
• pandas
• torchfile
• nltk
• scikit-image

Copied from LICENSE file (MIT License) for visibility:

Copyright for most part of this project bengali T2I are held by Tao Xu, 2018 as part of project AttnGAN. All other copyright for project Bengali T2I are held by this project owner, 2021. All non-data files that have not been modified by owner include the copyright notice "Copyright (c) 2018 Tao Xu" at the top of the file.

Reference

• AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks [code]
• StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks [code]
• Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks [code]

TODO:

• preprocessed meta data and dataset
• Training
• pretrained model
• valiadation
• Deploy as a web app that makes it easy to control the specific image one wants to generate