We investigate the phenomenon of vibrational resonance (VR) in neural populations, whereby weak low-frequency signals below the excitability threshold can be detected with the help of additional high-frequency driving. The considered dynamical elements consist of excitable FitzHugh–Nagumo neurons connected by electrical gap junctions and chemical synapses. The VR performance of these populations is studied in unweighted and weighted scale-free networks. We find that although the characteristic network features – coupling strength and average degree – do not dramatically affect the signal detection quality in unweighted electrically coupled neural populations, they have a strong influence on the required energy level of the high-frequency driving force. On the other hand, we observe that unweighted chemically coupled populations exhibit the opposite behavior, and the VR performance is significantly affected by these network features whereas the required energy remains on a comparable level. Furthermore, we show that the observed VR performance for unweighted networks can be either enhanced or worsened by degree-dependent coupling weights depending on the amount of heterogeneity.