PokeType

Have you ever wondered about what if you were a pokemon? Well, I did. However, to decide is even harder if you have so many options, like which type of element you are! Well, machine learning got your back, so no need to worry.

The goal of this project is to predict the pokemon type for a given image. Therefore, each user will be able to load an image, such as her/his selfie etc., click on the predict button, and then discover your type.

Usage

You can check it out immediately here.

To use it locally, edit etc., firstly clone this repo.

git clone https://github.com/rabiaedayilmaz/PokeType

src folder contains web application using HTML, CSS and JS. train folder contains machine learning model and dataset using TensorFlow framework and Python. After training the model, convert it into tensorflow-js format to use it in JS. To convert it, use following structure, then download the model. Later add model into corresponding file:

pip install tensorflowjs
tensorflowjs_converter --input_format keras \
                        train/model/PokemonTrainerLastBestModel.h5 \
                       src/model-js

For webpage interface, there are some commands to download packages etc.

npm install express
npm install jquery
npm install nodemon

Then, to activate the project on your local server, use these commands respectively:

npm init
node src/index.js

Now, you are all set up. You will be able to see and test on http://localhost:8080

Special thanks to Emir Öncü for designing the PokeType logo for me, you rock!