Atelier I: Discovery 001
My project is a webpage that uses the ml5 library to guess what Pokemon type an inputted image would be, from either fire, water, grass, normal, fighting, flying, poison, electric, psychic, bug, and rock. To achieve this, I trained a classifier with a dataset consisting of Pokemon and their respective types. Then, when the user inputs an image, it will see what category the image falls under. The intention behind this was to investigate and compare real world design trends with Pokemon, such as water types being mostly blue or grass types being green, stout, quadrupeds.
See the Project Here
(errors are logged in the console, open it if your image isn't loading)
Examples and Experiments
Through the use of Coding Train tutorials, I was able to get a grasp on how to use ml5, as well as how to train, save, load and use a classifier. These experiments document my learning process, and were essential to the creation of the end product.
In the end, the projects works as intended, though it was unable to correlate design trends as I expected. If I had more time, I would like to give this project more polish. I would fix little things like the user image not displaying proportionally, as well as improve the design of the page through custom fonts. In addition, I would also like to make the project generate a list of Pokemon moves the image would have, almost creating a completely new Pokemon from an image.
(For a consistent, error-free experience, please only use images from imgur or wikipedia)