Luxi-Zhao/chinese-landscape-painting-stylization

A deep learning project for creating Chinese landscape paintings from human-drawn sketches.

Chinese Landscape Painting Stylization

This repo contains source code for training a Pix2Pix model that synthesizes Chinese landscape paintings from human-drawn sketches. Contents of synthesized paintings include mountains, rocks, trees, and houses. The model can also be used for applying style transfer to photos.

Example Training Data

Example Model Outputs

Sketch to Painting

Style Transfer

Overall Technical Workflow

Try it out on Colab

Get your own paintings:

Train model:

Acknowledgements

• This project is a continuation of a course project done in collaboration with .
• Training data: Chinese Landscape Painting Dataset.
• Procedure for obtaining sketch-like edge maps was inspired by Holistically-Nested Edge Detection and SketchyGAN.
• Pix2Pix model training: TensorFlow tutorial.