<img src="./draggan.png" width="750" alt="Architecture of DragGAN"/>

DragGAN

Implementation of DragGAN: Interactive Point-based Manipulation on the Generative Image Manifold.

# gui
pip install dearpygui
# run demo
python gui.py

<img src="./UI.png" width="600" alt="Demo UI"/>

<img src="./sample/start.png" width="200" alt="Demo Start"/><img src="./sample/end.png" width="200" alt="Demo End"/><img src="./sample/multi-point.png" width="200" alt="Multi Point Control"/>

TODO

• [x] GUI
• [x] drag it
• [ ] load real image
• [ ] mask

StyleGAN2 Pre-Trained Model

Rosinality's pre-trained model(256px) on FFHQ 550k iterations [Link].

References

• https://github.com/rosinality/stylegan2-pytorch