Assignment 5

Reading Response:

How can machine learning support people’s existing creative practices? Expand people’s creative capabilities?

Machine Learning helps people to become more expressive and support their current/existing creative practices. The ability to find patterns, make predictions, and generate new content based on the training offers the opportunity to create numerous creative outputs. In the example she gave on her Wekinator music generator, it was already creative when she was able to generate beats just by training the model when she is off the screen vs in the screen. It’s such a simple way to create music without needing instruments or even skills; it detects your movements just by the movement of the pixels. It shows how anyone can produce music and create their own output. Beyond music, machine learning can help people produce content based on the data they’ve inputted that they never imagined, which would expand their creativity.

Although this isn’t directly connected to the neural networks concept, I used PoseNet() last year in my Designing Live Performances class to produce a musical piece just by placing my hands over buttons placed on different parts of the screen. Using video clips of my childhood, I created a narrative of my childhood where the video clips would appear on the screen when I hover my hands over the buttons that trigger each clip. Through mere hand movements, I created music just through the sound from each video clip and background music. It’s fascinating how someone without a musical background can create a creative musical, digital piece just through machine learning.

As for the assignment, I added onto the posenet examples we went over last class with the nose identification. I wanted to try the eyes and the nose, so I tried editing the code so that the model identifies your eyes and can be trained on it. Since ellipses are not interesting, I replaced them with illustrations of eyes and noses. While I was successful in mapping the eyes, I wasn’t as successful in training the model. The console log would just say “done” after I add the examples rather than getting trained. Ideally, I wanted to only have ellipses on the eyes and noses before the model is trained and have the ellipses replaced by the eye and nose images when the model is successfully trained and losses show. I will keep trying to make this work!

Here is the editor: https://editor.p5js.org/kimhy1031/sketches/2Rp71G5hi