Professional Documents
Culture Documents
Machine Learning For Musicians and Artists
Machine Learning For Musicians and Artists
Machine Learning For Musicians and Artists
07/21/2016
Machine Learning Pipeline
Input sensors: game controller, webcam, GPS, Twitter data, MPC
Processing/Decision-making (Model) computer decides what action
to take based on input
o Without ML, youd spend a lot of time tweaking this
processing
o Sometimes code is impossible to write if input is too complex,
noisy, etc
Output sound synthesis code, game code, animation, etc
Supervised learning algorithms tool using algorithm to build middle
component (model)
Input
Training
Data
Learning
Algorithm
Model
Output
Web-Controlled Drum Machine
Input - Processing: extracts data from webcame
Model Wekinator sends info to Chuck
Output Chuck: drum machine
Training data set give Wekinator training examples of standing in
front of webcam, not standing in front of webcam, etc, tell it what
output to give for each example
Connecting inputs and outputs with Open Sound Control
Open sound control connects inputs and outputs to the model
Use OSC libraries for whatever platform/software I want to use
OSC Communication
Application 1 (sender) sends OSC message and data to Wekinator
(model) which sends OSC message to output (receiver)
When sending an OSC message, need to tell operating system
where to send osc message
o What port do you expect the other program to be listening for
an input
o Port temporary virtual location on computer like a
temporary mailbox