Retrieving information from remotely sensed data is an important task. In the present work, data of L band microwave radiometer has been used to collect the brightness temperature over bare and vegetated fields in two polarizations at different moisture levels. Artificial neural network (ANN) trained with Levenberg-Marquardt algorithm has been used to determine soil moisture from brightness temperatures measured by microwave radiometry. ANN are trained to evaluate the moisture content in the range 0-36% from different sets of data of bare and vegetated fields. Properly trained feed-forward neural network with Levenberg-Marquardt algorithm predicted soil moisture content with less mean absolute error.