A detailed Analysis of the hello_world_test.cc program
Line: 11 to 11
tensorflow/lite/micro/examples/hello_world/models/hello_world_float_model_data.h and
tensorflow/lite/micro/examples/hello_world/modules/hello_world_int8_model_data.h
Changed:
< <
are the models created with train.py and converted to a C++ array with xxd.
> >
are the models created with train.py and converted to a C++ array. While the book describes how to create the .cc file using the Unix xxd command ,in which case the corresponding .h file must be created by hand, the new version uses the python script generate_cc_arrays.py in tensorflow/lite/micro/tools to acconplish the conversion.
You can easily try this with the command:
The first parameter to the command is the name of the output directory and the second parameter is the filename of the model (should have the extension .tflite) The same python script can also be used to convert audio files (.wav), image files (.bmp) or excel sheets (.cvs).
tensorflow/lite/micro/micro_interpreter.h
The Tensorflow Lite Micro interpreter, which will run the model