Kaz Hishida (.tech)

Snake AI

After watching a video of some developers teaching bots to play a game of hide and seek using a genetic algorithm, I was inspired to create a similar algorithm of my own. I modified an open source snake game to work for my purpose, and created a neural network using TensorFlow and trained it with a genetic algorithm. It was a struggle at first, but it was incredibly rewarding to get to watch my AI learn in real-time. Check it out on Github below.

View Projects

Rock Paper Scissors Classifier

In an effort to learn more about deep learning and neural networks, I created an image classifier to detect whether a picture of a hand is holding rock, paper, or scissors. I represented the images using NumPy, and after some optimizations, put a large corpus of training images through a Convlutional Neural Network. Despite the training data being animated photos, the model worked with up to 80% accuracy on real images as well (and up to 94% on images from the initial dataset). Check it out on Github below.

View Project