Skip to content

README.md

Latest commit

 

History

History
28 lines (18 loc) · 1.16 KB

README.md

File metadata and controls

28 lines (18 loc) · 1.16 KB
Raw

CycleGAN Style Transfer API

This repository contains the backend for a CycleGAN-based style transfer web application that transforms user-uploaded images into artworks in the styles of famous painters — Vincent van Gogh, Edvard Munch, and Claude Monet.

All models were personally trained using artist-specific datasets and implemented with reference to research literature. This repo includes the Flask API that powers the web app, the trained models, and the Jupyter notebooks used for model development.

Features

  • RESTful API built with Flask
  • Supports style transfer using pretrained CycleGAN models
  • Models trained using custom curated datasets per artist
  • Supports styles: Van Gogh, Munch, and Monet
  • Integration-ready: communicates seamlessly with a Flutter frontend

Models

Each model is a CycleGAN trained for unpaired image-to-image translation using TensorFlow/Keras. Architectures were adapted from research papers and tuned specifically for stylistic consistency.

The following styles are available:

  • vanGoghModel.pth
  • munchModel.pth
  • monetModel.pth

Models are located in the /models/ directory.