This module takes a trained Keras neural network and compiles it to fit a MCU. The compiling is done in two parts: first the model is compiled into a Tensorflow Lite model and then it is converted into a C-array, that can be used by Arduino Nano

source

plot_size

 plot_size (model_path)

Plots the size difference before and after quantization Args: model_path: Path to model files Returns: pandas dataframe: Pandas dataframe containing information

source

convert_model_to_cc

 convert_model_to_cc (model_path:str)

Creates model.cc from model.tflite in folder model_path

source

convert_to_c_array

 convert_to_c_array (bytes)

C array conversion

source

convert_model

 convert_model (model_path:str, output_path:str, dataset_path:str,
                model_params:dict, model_name:str)

Model conversion into TFLite model Args: