nyumaya_audio_recognition_lib

Warning this is still work in progress, especially the CMake file has a hardcoded target for Linux x86/64. Any other platforms will currently not work out of the box and require some tweaking

You don't have to build this library by yourself if you want to run the nyumaya_audio_recognition on the Raspberry Pi ,macOS or Linux x86-64. There are prebuilt libraries here.

Depedencies

-Cmake

How to build the library

• Before cloning this repository clone the official tensorflow repository

git clone https://github.com/tensorflow/tensorflow.git 
cd tensorflow
git checkout 49372a1c6e4e83f153c740996c88b6bf76ec1abf 

• Clone this repository within the top folder of the tensorflow repository

git clone https://github.com/nyumaya/nyumaya_audio_recognition_lib.git

• Modify the registered ops

cd nyumaya_audio_recognition_lib/
git apply --unsafe-paths --directory=../ tflite.patch

• Download dependencies

../tensorflow/lite/tools/make/download_dependencies.sh 

• Modifying CMakeLists.txt for your platform

Depending on your platform (like x86-64 desktop, arm on RPi...) you have to set the correct CXX_FLAGS and C_FLAGS. Examples for x86-64, RPI3 and RPI_0 are given. To use those architectures comment out the default Flags and comment in the flags for the chosen platform. The default build is a generic x86-64 version.

• Make

cmake ./
make 

Credits

• The library uses https://github.com/mborgerding/kissfft