Question about use own tensorflow model in the movidius stick2

I trained my tensorflow model and transferred to the IR (.xml+.bin)successfully, then I want to use the movidius stick2

to inference my model, just wondering how to do it? I found some examples in the original installation package, including face_recognition, road_barrier_detection and so on, each package provide a C++ file about using the Inference Engine API. Need I write a C++ file to use movidius stick2? or have other method to use it? I found another document showed that I need to transfer tensorflow model to .graph, but it need mentioned how to use this graph model.