Researching machine learning
Spending a little time today thinking about machine learning and how it can be useful for what I do. In the past I have used generative algorithms to create patterns and designs that have worked their way into the design. Decades ago I was using PovRay to do raytracing based on simple objects and transparency. Later on, I played with Action Script and Processing to create interactive art pieces. I am currently using C++ and MicroPython/CircuitPython to mess about with microcontrollers, servos, e-ink, and LEDs.
I have used various forms of machine learning in the past to create the base images. Sometimes it is just filling areas in intelligently, or stretching out dimensions without obvious distortion. Other times, it has been making subtle changes to faces that relate to re-lighting or re-posing them. I have also experimented with with causing basic changes to facial features.
At the moment, I am interested in training a system based on a certain genre of painting. I may need a bit more computing horsepower than I currently have. At least I can work out the software now.
https://runwayml.com/ Offers some cool video processing options. It also has a web based interface that seems related to image processing and training. The nice thing about this option is that there really is very little barrier to entry. At the moment, I just have to come up with 500 – 5,000 images that can be used to train the system.
The systems behind this work are called Generative Adversarial Networks (GAN) and here is a little list of places they are collected.
This is a free and open-source library for machine learning and artificial intelligence.
On a Mac
Working on a Mac laptop so there is specific code that you can run from Apple. Unfortunately it isn’t really working on my laptop.