Soil moisture measurements are used to determine the effects of changes in soil moisture on hydrology, meteorology, agriculture, and watershed condition in semi-arid lands. Reliable measurements of soil moisture are needed for a wide variety of applications, including water balance and water flow calculations, entered precipitation runoff. Indeed, the objective of our study is to compare sensors at very low cost by a measurement campaign by adding liquid to a vase with earth inside, in an appropriate and programmed way, for different types of soil. (sandy, clayey, and silty). Analyze the state of the art, to understand that very low-cost sensors are capable of measuring soil moisture in the long term for farmers who do not have enough money to invest in innovative technologies for profitability of their agriculture. The sensors are placed in deep vases placed on a digital scale to collect data by reading the sensors via raspberry pi. Our result showed that during the calibration, there was less error after the least squares regression fit of the data for the different soil types. Indeed, the capacitive soil moisture sensor manufacturer DFROBOT and RASBERRY PI can be used by any type of soil in the long term. Then we develop backend applications and software (artificial intelligence) supporting farmers in their farming practices. The objective here is to automate the irrigation by means of a solenoid valve, for this we use deep learning algorithms because of the volume of data that we had collected with the soil moisture sensor, we had used the artificial neural network to predict whether the soil will be wet or dry, it’s the classification problem. This binary result will be preprocessed for the prediction. It is in this context that we created an algorithm to automatically operate the solenoid valve. With the prediction of the wet / dry binary system, we will couple with turn off / turn on of solenoid valve.