Getting the software requirements– Creating Applications on the Edge

To run the code needed for the practical, if you are running under the Windows environment, we need to install several programs, as follows:

Python 3: Download it from https://www.python.org/ftp/python/3.9.9/python-3.9.9-amd64.exe. Make sure your python3 directory is already in the path environment; check this by searching environment variables in Windows search, clicking on Edit the system environment variables, and clicking on Environment Variables on the bottom-right window. After a new window appears, click on the System Variables window (bottom part), go down, and double-click on Path. You should find these two directories:

1. C:\Users\User Name\AppData\Local\Programs\Python\Python311
2. C:\Users\Guest User\AppData\Local\Programs\Python\Python311\Scripts

If it does not exist, then you must add NEW and type in those two directories to enable the Windows program to run Python and Python script files from other directories.

TensorFlow: Install it using pip from the relevant terminal application (for example, Command Prompt for Windows; Terminal for Mac):
python3 -m pip install tensorflow
# Verify install:
python3 -c “import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))”

pandas: Install it using pip from Command Prompt:
pip install pandas

You can check whether TensorFlow and pandas are already installed by typing the following:

pip list

You will find all the packages installed together with the version number.

With the package installed, we are ready to build our TensorFlow model.

Building the TensorFlow model

First, we need to import TensorFlow and layers as part of the TensorFlow Keras library. We also need to import pandas to run mathematical operations on our workflow:
import tensorflow as tf
from tensorflow.keras import layers as lyrs
import pandas as pd

We then need to separate the independent and dependent variables that are located within the dataset, and accordingly store them within the inputs and outputs variables.

You need to get the humidity_data.csv and temperature_data.csv files by retrieving them from your saved result history for a certain period from your sensor readings:
data = pd.read_csv(“temperature_data.csv”)
inputs = data.iloc[:, :-1].values
outputs = data.iloc[:, -1:].values

With Keras, we will then create an NN model that will be based on 2 layers of 16 neurons, both with the relu activation function. The model will then be trained using the fit function, with the inputs and outputs variables as the parameters to it. We then store it within the Model_humidity_predict file:
nn_model = tf.keras.Sequential([
    lyrs.Dense(16, activation=’relu’),
    lyrs.Dense(16, activation=’relu’),
    lyrs.Dense(1)])
nn_model.compile(optimizer=tf.keras.optimizers.RMSprop(), loss=tf.keras.losses.MeanSquaredError(), metrics=[tf.keras.metrics.MeanSquaredError()])
//Fitting model to have it understand and learn from the input data
nn_model.fit(inputs, outputs, epochs=1000, batch_size=16)
nn_model.save(‘Model_humidity_predict’)

With that, we have trained our NN model with TensorFlow Keras, and we can now look at converting the TensorFlow model to a TensorFlow Lite model.

---