Cartoonizer

Convert images and videos into a cartoon!

The webapp is deployed here - https://cartoonize-lkqov62dia-de.a.run.app

You can find a writeup on this webapp's architecture here!

Prerequisites for Google Cloud and Algorithmia

These are important steps if you want to leverage Google buckets, signed URLs and Algorithmia's platform.

Cloud Run authentication

To use any functionalities pertaining to Google Cloud, you'll need a global authentication file (JSON). You can obtain this JSON by following the steps given here - Getting started with authentication

After you get the JSON file, rename it to token.json (so that it's compatible with the codebase).

Set the environment variable in your terminal -

export GOOGLE_APPLICATION_CREDENTIALS="/path/to/token.json"

Notes:

• You can set it permanently by adding this line to ~/.bashrc.
• Dockerfile already includes the setting of this particular environment variable. :)

Algorithmia

We used the Serveless AI Layer product of Algorithmia for inference on videos. To learn more on how to deploy your model in Algorithmia, check here - https://algorithmia.com/developers

Installation

Application tested on:

• python 3.7
• tensorflow 2.1.0
• tf_slim 1.1.0
• Cuda version 10.1
• OS: Linux (Ubuntu 18.04)

Using Docker

The easiest way to get the webapp running is by using the Dockerfile:

1. cd into the root directory and build the image

docker build -t cartoonize .

Note: Set the appropriate values in config.yaml before building the image.

2. Run the container by exposing the appropriate ports

docker run -p 8080:8080 cartoonize

Using virtualenv

1. Make a virtual environment using virutalenv and activate it

virtualenv -p python3 cartoonize
source cartoonize/bin/activate

2. Install python dependencies

pip install -r requirements.txt

3. Run the webapp. Be sure to set the appropriate values in config.yaml file before running the application.

python app.py

Using Google Colab

1. Clone the repository using either of the below mentioned way:

   • Using Command:

     • Create a new Notebook in Colab and in the cell execute the below command.

      ! git clone https://github.com/experience-ml/cartoonize.git
     

     Note: Don't forget to add ! at the beginning of the command

   • From Colab User Interface

       Open Colab
           └── File
                └── Open Notebook
                         └── Github
                               └── paste the Url of the repository

Note : Change the runtime to GPU before running the application

           Runtime
              └── Change runtime type
                          └── Select GPU

2. After cloning the repository navigate to the /cartoonize using below command in the notebook cell:

%cd cartoonize

3. Run the below code in cell to install the ngrok for creating the url to access the application:

!pip install flask-ngrok

4. In /app.py Add the following code:

   Import the ngrok:

   from flask_ngrok import run_with_ngrok
   

   After app variable add the following command at line 31:

   run_with_ngrok(app)   #starts ngrok when the app is run
   

   Change the app.run() to the below format:

   app.run() #remove the passed parameter 
   

5. Run the below commands in the notebook cell :

   Install the requirements

   !pip install -r requirements.txt
   

   Launch the flask app on ngrok

   !python app.py
   

Sample Image and Video

Emma Watson Cartoonized

Youtube Video of Avenger's Bar Scene Cartoonized

License

1. Copyright © Cartoonizer (Demo webapp)

   • Authors: Niraj Pandkar and Tejas Mahajan.

   • Licensed under the CC BY-NC-SA 4.0

   • Commercial application is prohibited by license

2. Copyright (C) Xinrui Wang, Jinze Yu. (White box cartoonization)

   • All rights reserved.
   • Licensed under the CC BY-NC-SA 4.0
   • Also, Commercial application is prohibited license (https://creativecommons.org/licenses/by-nc-sa/4.0/legalcode).