The ability of algorithms to evolve or learn (compositional) communication protocols has been traditionally measured in the language evolution literature by emergent communication tasks. In this work, we scale up this research by using contemporary deep learning materials and train reinforcement learning neural-network agents on referential communication games. We extend previous work in this direction in which agents learn from symbolic environments to those learning from using raw pixel input data, a more challenging and realistic input representation. We find that the degree of structure found in the input data affects the nature of the emerged protocols, and thereby corroborate the hypothesis that structured compositional language is most likely to emerge when agents perceive the world as being structured.