Creating an interpreter in OpenCV C++ for the ONNX model built from pytorch-based jupyter-notebook code

I can ONNX export the pytorch model created with the notebook file at https://github.com/jeremycochoy/style-transfer. But the ONNX model created here only contains feature maps and what I want is to create an interpreter with dynamic input and output working with OpenCV in C++.
I prefer it to work with opencv dnn module. But onnxruntime can also use.