Inspiration

Our inspiration was to build something from our childhood memories and apply what we know now as college students, and bring those memories to life.

What it does

Given an image of either Turtwig, Chimchar, or Piplup, the machine learning program will return what it thinks the image is.

How I built it

Using the fast.ai library, we created a classifier that will take pixel inputs of a given image and return the probability of it being one of the starter pokemon. We created our training data set from Google images, and a validation set from 20% of those images. We from those inputs and random parameters, we created our neural network.

Challenges I ran into

Creating our training data from Google images meant that our training data was not going to be 100% accurate, so it would cause some error when training our neural network.

Accomplishments that I'm proud of

Putting together something from our childhoods as a team with our currently knowledge and skill set was impressive and nostalgic.

What I learned

We've strength our knowledge on how machine learning algorithms operate, and manipulate our data.

What's next for Machine Learning Pokemon

Possible creating more classes for different types of pokemon, and creating some type of live video feed as input for our data.

Built With

Share this project:

Updates