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

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 4f2a9acaff04bb81684c2b49c955f296315473ac

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

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

• Modify two tensorflow files

cd nyumaya_audio_recognition_lib/
cp download_dependencies.sh ../tensorflow/lite/tools/make/
cp register.cc ../tensorflow/lite/kernels/

• Download dependencies

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

• Init kiss_fft

git submodule update --init kissfft/

• Make

cmake ./
make 

Credits

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