Detection of gravitational waves (GW) from compact binary mergers provide a new window into multimessenger astrophysics. The standard technique to determine the merger parameters is matched filtering, consisting in comparing the signal to a template bank. This approach can be time consuming and computationally expensive due to the large amount of experimental data which needs to be analyzed. In the attempt to find more efficient data analysis methods we develop a new frequency domain convolutional neural network (FCNN) to predict the merger masses from the spectrogram of the detector signal, and compare it to previous time domain neural networks (TCNN). Since FCNNs are trained using spectrograms, the dimension of the input is reduced as compared to TCNNs, implying a substantially lower number of model parameters, consequently allowing faster processing, and a better out of sample performance. We find that in general FCNN have better performance on validation data and lower over-fit, providing a new promising approach to the analysis of GW detectors data.