By embedding active materials, 4D printing enables the fabrication of highly integrated structures capable of changing their shape and functionality through actuation by a stimulus. However, unleashing the full potential of the technology requires efficient computational design methods that can search vast design spaces. One problem formulation in this area is shape morphing, which can be defined as finding a distribution of active and passive materials such that actuation by a stimulus results in a desired shape change from an initial to a target shape. While existing approaches to solve such problems show promising results, they typically struggle to generate high-resolution results, e.g. when using stochastic optimization, or with discrete structural elements, e.g. as in the case of gradient-based optimization. Here, a multi-step approach is used to design shape-morphing structures. In a first step, an artificial neural network (ANN) is trained, which takes a coordinate of a point in the initial shape as input and returns the coordinate of the point after actuation as output, so that the ANN represents the deformation occurring upon actuation. Later, the mapping represented by the ANN is used to design each structural member of the shape morphing structure independently. As a result, this approach requires solving several small problems rather than a single large one, such that the computational cost is linear in the number of structural members. This enables the efficient design of detailed shape-morphing structures. Remarkably, when designing shape morphing structures with this approach, the response of the global structure to a stimulus never needs to be simulated. To validate that the resulting structures indeed show the desired shape-morphing behavior, an experimental investigation is conducted. Three sample shape morphing problems are considered, one of them using mechanical actuation and two actuated by the swelling of wood-filled filament. The multi-step approach is used to design shape morphing structures, which are then manufactured using multi-material 3D Printing. The experiments show that both the mechanically and swelling-actuated structures are capable of accurately matching the target shape. This result demonstrates that the multi-step design approach can efficiently design high-resolution shape-morphing structures and therefore leverage the possibilities offered by 4D printing to create smart multi-functional systems.