Given the recent detection of gravitational waves from individual sources, it is almost a certainty that some form of background of gravitational waves will be detected in future. The most promising candidate for such a detection is backgrounds made up of incoherent superposition of the signal of unresolved astrophysical, or backgrounds sourced by earlier cosmological events. Such backgrounds will also contain anisotropies about an average value. The information contained in the background level and any anisotropies will be extremely valuable as an astrophysical and cosmological probe. As such, the ability to reconstruct sky maps of the signal will become important as the sensitivity increases. We build and test a pixel-based, maximum-likelihood gravitational wave background (GWB) map-maker that uses the cross-correlation of sets of generalized baselines as input. The resulting maps are a representation of the GWB power, or strain ‘intensity’ on the sky. We test the algorithm by reconstructing known input maps with different baseline configurations. We also apply the map-maker to a subset of the Advanced Laser Interferometer Gravitational Wave observatory data.