Mood Swings

Response to Dr. Rebecca Fiebrink:

Machine learning expands the realm of creative possibilities greatly but it also means greater responsibility for the artists. Machine learning definitely makes art more accessible for a lot of people and stretches the definition of art itself to its limits.

I also found it interesting to not think of data as ground truth but rather as an active choice by the designer. Consequently, I don’t think machine learning will ever replace artists but rather expand the meaning of what is considered an artist. The relationship between people and machines can become a much more holistic, intuitive one. Machine learning also allows artists to explore their mistakes (which are often more interesting than the desired results) in way that traditional coding doesn’t.

Code Exercise:

I am very intrigued by the possibility of machines that are able to detect human emotions or lack thereof. In a different class I had to research how we can allow a computer to monitor someone’s mental state to prevent suicides / self harm. An idea for a real time machine learning model would be a model that given many many inputs can make predictions on the user’s mood. Those inputs can vary from music the user listens to, browsing history, patterns in keyboard typing, taking into account patterns out of the user’s control like time and weather conditions.

For this week’s assignment, I tried something very small. I played around with pose detection. I trained my model to classify different poses or postures that correlate to mood and control a generative art sketch to show that mood. The colors of the shapes correlate to how sad / happy the model can guess from someone’s webcam. To keep it simple I trained my model on 3 different inputs and 3 different outputs.

A hunched down posture → Gray scale colors

Hands up → Vibrant colors

The colors change as I change my posture.