The programm uses machine learning (tensorflow in python) to take any image, scale it and return a prediced pokemon type.
How to use: 0. Make sure you have installed the python libraries imageio, tensorflow and numpy.
- Extract the zip file PokeSprites.zip
- Inside the new folder called PokeSprites, you will find another folder called PokeSprites. Copy this second folder into the directory of the other files.
- Run the MLpokemon.py file to create a model. For how long you want to train can be adjustet by changing the epochs variable at the bottom of the programm (default is 100, more should not be nessecary)
- Place your image in the folder and call it usedImg.png.
- Run the loadAndUseModel.py file to use the model.
- Repeat 4-5 for any images you want to analyse (no need to make a new model)
Overview of files:
loadAndUseModel.py - run this code with a relevant image called usedImg.png to try the model
MLpokemon.py - run this code to create a model. You can determine the number of epochs (for how long the model learns) at the bottom of the file
PokeSprites.zip - A zip folder of pokemon sprites to be used for training the model
pokemon.csv - A file of what type each pokemon is