A purely data-driven approach using deep convolutional neural networks is discussed in the context of Large Eddy Simulation (LES) of turbulent premixed flames. The assessment of the method is conducted a priori using direct numerical simulation data. The network has been trained to perform deconvolution on the filtered density and the filtered densityprogress variable product, and by doing so obtain estimates of the un-filtered progress variable field. Any non-linear function of the progress variable can then be approximated on the LES coarse mesh and explicitly filtered to advance the LES solution in time. This new strategy for tackling turbulent combustion modelling is demonstrated with success, for both the sub-grid scale progress variable variance and the flamelet filtered reaction rate, two fundamental ingredients of premixed turbulent combustion modelling.