Build the Google FlatBuffers library
FlatBuffers is another efficient cross-platform serialization library for C++ developed by Google for performance-critical applications. The TensorFlow Lite files are generated using FlatBuffers to serialize their TensorFlow Lite model data so Arm NN needs to use FlatBuffers to load and interpret the TensorFlow Lite files.
The FlatBuffers library is required for the Arm NN TensorFlow Lite parser. To build the FlatBuffers library, use the instructions found in the FlatBuffers Building Guide.
$ cd $BASEDIR/flatbuffers-1.12.0 $ CXXFLAGS="-fPIC" cmake .. $ -DFLATBUFFERS_BUILD_FLATC=1 $ -DCMAKE_INSTALL_PREFIX:PATH=$BASEDIR/flatbuffers