This project is based on tensorflow/playground.
It aims to bring interactivity to the user interface with USB controllers by using WebMIDI.
It has been authored at CREATIS Laboratory (Lyon, France).
Available here.
- Go to releases
- Download the
coolearning-playground-v{version}.zipasset - Unzip the archive
- Open
index.htmlwith chrome
Automatic publishing and versioning are enabled on the master branch and handled by semantic-release.
Commit messages are expected to be formatted following the Angular Conventions.
git clone https://github.com/coolearning/playground.git
cd playground
yarn
yarn dev
# navigate to http://localhost:3000There are two main categories of devices you can attach.
This device usually provides a grid of colored backlit pads.
Use the grid to display the inputs, neurons and output of the current neural network.
- Short press action:
enable/disableinputs and output linksselect/unselectneuronsselect/unselectlayer
- Long press action:
enable/disableneurons
This device usually provides faders, potentiometers and buttons.
A controller has 3 main modes:
- Default mode:
- When nothing is selected, you can bind controls to parameters through the ** mappings UI**
- Selection mode:
- When neurons are selected, you can change their:
- Source weights
- Source biases
- Source meta (learning rate, activation, etc.)
- When neurons are selected, you can change their:
- Layer mode:
- When a layer is selected, you can change its:
- Neurons biases
- Neurons meta (learning rate, activation, etc.)
- When a layer is selected, you can change its:
chromeonly compatibilityselectorat least 64 padscontrollerat least 8 fadersd3version 3 dependency- No automatic configuration for unknown devices, you must provide a configuration file.
