Style Transfer Website

A website that applies artistic styles to your photo. Based on the Artistic Style Transfer using Neural Networks. This project is an extension of Style transfer app

Live at bit.ly/stylizeapp

Hosted using AWS EC2, AWS S3, Gunicorn and Flask.

Architecture

Quick Start

Running locally

1. Clone the repository

   git clone https://github.com/puneet29/stylizeapp.git
   cd stylizeapp
   

2. Create new virtual environment

   • On windows:

     py -3 -m venv venv
     

   • On Linux:

     python3 -m venv venv
     

3. Activate virtual environment

   • On windows:

     venv\Scripts\activate
     

   • On Linux:

     . venv/bin/activate
     

4. Install all the dependencies

   pip3 install -r requirements
   

5. Change the secret key in __init__.py at #19. Replace os.urandom(24) with actual bytes value.

6. Run Web App

   gunicorn -w 4 -t 600 __init__:app
   

Running on server

Follow the same procedure, except 6th step. While running webapp we want the process to be daemonized, so that it runs in the background. For that:

• Run Web App

  gunicorn -b 0.0.0.0:80 -w 4 -t 600 __init__:app -D
  

We bind our application to all HTTP IPv4 addresses on the server using 0.0.0.0:80 and run 4 gunicorn processes that serve our application. The timeout for each gunicorn worker is set to 600 seconds, to account for 10 min delay during uploading process. -D option daemonizes the process.

Refer to this medium article on how I built this app. Also refer to this gist on how to setup an AWS EC2 instance.

Libraries used

• Pytorch
• Flask
• Scipy
• Gunicorn
• Boto3
• Beautiful Soup 4
• Requests

References

• A Neural Algorithm of Artistic Style
• Perceptual Losses for Real-Time Style Transfer and Super-Resolution
• Instance Normalization