Create Plugin for TensorRT for Tensorflow application

I have one Tensorflow application in Python and it’s output of model inference is post-processed in CPU. This post-processing part is not a tensorflow layer and just c++ implementation (WRAPPED TO Python).

Plugin samples in TensorRT for Tensorflow are for Tensorflow layers which are not available in TensorRT. Then during UFF conversion, TF operation is mapped to TRT plugin.

Say if I want that post-processing C++ class into TensorRT, how can I get it in UFF conversion?