AttnGAN is a work I found here. https://github.com/taoxugit/AttnGAN
It can convert text descriptions into quality images when trained on a dataset.
Only the GUI section is coded by me. All other parts were converted from original codebase in Python2 to Python3 using 2to3. All the models were obtained from the official repo. Some parts has been edited to get compatible with PyTorch cpu only.
I used anaconda for development in a windows 10 machine. Python version was 3.8.2. PyTorch CPU is used instead of PyTorch GPU.The conda package list provided in the requirements.txt.
Result from the GUI:
Save the text encoder into DAMSMencoders from https://drive.google.com/open?id=1GNUKjVeyWYBJ8hEU-yrfYQpDOkxEyP3V
Save the model into models from https://drive.google.com/open?id=1lqNG75suOuR_8gjoEPYNp8VyT_ufPPig
Install packages specified in requirements.txt
Move into the directory code.
Run
python gui.py