The resolution of conventional imaging systems is inherently restricted by the diffraction limit. To surpass this diffraction barrier, a scheme using an external aperture modulation subsystem (EAMS) and related deep learning network (DLN) is presented in this paper. The EAMS facilitates the realization of various image acquisition strategies and related DLN architectures. In the specific scenario of 3-aperture modulation strategy, the capabilities of this approach are validated both in numerical simulations and experiments. The results show that both the resolution enhancement ability and the image fidelity can be improved by just adding one label data. This framework proposed here provides a more general way to further explore the ability of DLN-based method to surpass the diffraction limit, and permits a rapid data acquisition that enables new opportunities for the training data collection and further super resolution imaging of label-free moving objects, such as living cells.