Project #2 Teachable Machine

Description

Play a song and the machine will tell you whether or not it is a High Quality Video Game Rip.

Design Process

My inspiration for this project was the YouTube channel SiIvaGunner, who is dedicated to uploading meme edits or mashups of video game music and passing them off as the real thing, referring to them as "High Quality Video Game Rips." The channel reuses memes as well as video game music, so I wanted to see if the Teachable Machine could tell if a music sample was a high quality rip or an original song. In order to train my model, I created two categories: rip or original song. I originally included over 2000 samples for each category (I included different memes for the rips category), and the model was unable to be trained even after leaving my computer on overnight. After this, I decided to cut the number of models by half and the model was trained. Before, my p5 program could only say "yes" or "no" in response to sound being played, but after a suggestion, I decided to include different confidence levels to give it more variety. I also got a suggestion to include recommendations to high quality rips underneath responses that weren't "yes." For this, I created an array of links to rips. Then I created squares that act as buttons: if you click on one of them, you are redirected to one of the links in the array (the link chosen is random). I programmed it so that you could only click on the squares. If I had more time I would customize the CSS more and change the background colors based on the model's response. I would have also liked to add a score counter in order to keep track of the number of times the model was correct.

Credits

This project was created using Teachable Machine.