Code examples for Machine Learning for Interaction Design
Andreas Refsgaard and Jen Sykes
We Will use Slack to add any updates and useful information. Link to Slack channel
In order to see the Arduino board's serial port you must have the correct driver installed on your computer first. The drivers are contained in the "Driver" folder in this repository. Make sure you select folder for your operating system (Windows / Mac osX).
For more details: See here and here
Please also make sure you have Processing and Wekinator downloaded.
You may need the following Processing libraries for certain examples. Please open Processing and go to Sketch/import library/ Add library and install the following...
- Serial
- Video
- Open CV
- oscP5
- Sound
Introduction to Jen, Andreas and creative machine learning.
Jen will look at Wekinator, including the use of Physical Computing tools with Machine Learning.
Introduction to Wekinator and machine learning concepts.
- Simple Classification - exploring the concept and different inputs such as web cameras and sound
- How can we connect an input to an output?
- Exploring image, video and sound as outputs.
Task: In small groups train a simple classifier on four different states.
- Simple - look at the provided examples we explored earlier and try to train your own inputs.
- Advanced - Adapt an input example and/or try change the output content.
Exploring continuous training models. How do we change our approach to this and how do the outputs change as a result?
- Simple Regression with mouse, webCam and face detection
- How do we connect an input to an output?
- Exploring continuous outputs such as graphical shapes and typography.
Task: In small groups.
- Simple - look at the provided examples we explored earlier and try to train a new response.
- Advanced - Adapt an output example perhaps create something generative?
Exploring Physical computing elements. What is taken into consideration when we move into the physical world?
Capacitive touch, accelerometers.
Task:
- Simple - explore the examples provided and train a classifier.
- Advanced - Try adapt an output example with multiple inputs (multiple- touch?). - Train a regression (continuous) sensor input.
Motors.
Task:
- Simple - explore any of the classifier examples from earlier in the day to make a motor move.
- Advanced - Try adapt the input examples from earlier to create a unique control of a motor.
Andreas will continue exploring varying methods covering P5 and ML5 examples.
Runway and group exploration.
- How to use Runway models.
- The key elements you can control.
- Different workflows for communication with other software.
Much of the code used in examples has been adapted, modified and expanded upon from original sources provided by...