Scene Stylization

This project implements an image style transfer application using a transformer model. Users can upload their content and style images, and generate a stylized output image using the implemented model. The project is written in Python 3, using PyTorch, and Streamlit libraries.

Project Structure

• demo/: contains sample content and style images for demonstration purposes
• networks/: contains pre-trained models for the image style transfer task
• transfer_it.py: style transfer inference utilities
• app.py: implementation of the Streamlit web application
• function.py: implementation of auxiliary functions used by the style transfer model
• models/: network implementation files
• output/: directory for storing preprocessed and output images
• util/: directory for storing network utility functions

Prerequisites

• Python 3
• PyTorch
• Streamlit
• Pillow
• Matplotlib
• NumPy
• OpenCV

Usage

streamlit run app.py

This will launch a local server and open the web application in your default browser.

To use the application, follow these steps:

• Upload a content image and a style image using the respective file pickers.
• Click the "Transfer Style" button to generate a stylized output image.
• Wait for the output images to be generated.
• The output images will be displayed on the web page.
• You can also clear your inputs by clicking the "Clear Inputs" button.

Acknowledgments

The style transfer model implemented in this project is based on the works of Y. Deng et al., “StyTr$^2$: Image Style Transfer with Transformers,” arXiv.org, 2021. https://arxiv.org/abs/2105.14576.