Polymer gel plates may be programmed to morph into three-dimensional configurations upon swelling. An effective strategy to control such shape transformations consists in patterning the in-plane cross-linking density of the polymer network to realize non-homogeneous swelling. In general, one needs to solve an inverse problem to determine the shear modulus field that produces a given target shape. Here, we propose a computational framework for the solution of such an inverse problem, which we validate against two benchmark problems, i.e. making cones and saddles from gel disks.