jarmitage / RapidLibBela

RapidLib is a lightweight library for interactive machine learning. Bela is a platform for interactive sensor and audio processing.

RapidLibBela

Context

RapidLib

main

dev

RapidLib is a lightweight library for interactive machine learning. It currently features classification (using kNN), regression (multilayer perceptron), and series classification (using dynamic time warping).

More at: https://mzed.github.io/RapidLib/

Bela

Bela is a platform for ultra-low latency audio and sensor processing.

http://bela.io

Install

See Releases.

Build

• Requires "experimental" Bela image with recent CMake: https://github.com/BelaPlatform/bela-image-builder/releases/tag/v0.5.0alpha2.
• Highly recommended to install distcc (guide).
• git clone RapidLib onto Bela (connecting to network on Bela)
• Comment out DTW multi-threading test (see mzed#20).
• Build commands from inside RapidLib dir:

mkdir build && cd build
CC="distcc-clang" CXX="distcc-clang++" cmake -DCMAKE_CXX_FLAGS="-Werror" ..
CC="distcc-clang" CXX="distcc-clang++" cmake
ctest

• Run ldconfig /root/RapidLib/build to add the binary path to the linker cache.

Usage

include "rapidLib.h"

• make parameters for RapidLib projects (see Bela Blog):

LDFLAGS=-L/root/RapidLib/build;LDLIBS=-lrapidLib;CPPFLAGS=-I/root/RapidLib/src;

Examples

Currently very simple, based on https://github.com/mzed/ofxRapidLib

Todo

• Benchmark RapidLib performance on Bela
• Add C++ examples using AuxiliaryTask
• Add C++ examples using sensor inputs and audio outputs
• Add C++ examples with Bela GUIs
• Add install to CMake to simplify installation
• Scope out SuperCollider / Pure Data support and examples