Using this data a multi-layer perceptron is trained to directly estimate the number of present targets from the received signals in order to determine if resolution in the spatial domain is necessary.