The Hello World example as provided by the tflite-micro repository

The TensorFlowLite-Micro repository provides the code
  • creating the hello world neural network model
  • creating input tensors with x and sin(x) values used to train the model
  • converting the model to tflite flatbuffer format and quantizing it to int8_t numbers
    The quantized model can be read from a file in case of the MicroPython version of the hello world example. The Arduino SDK does not use a file system and the quantized model is therefore converted into C++ code included into the source code of the Arduino hello world version.
  • testing the model with input tensor values 0..2π

Comments

 
Topic revision: r2 - 15 Dec 2025, UliRaich