Post processing in C++ tensorrt

Hi, I know already a great piece of work thanks for that. Could you please guide how to post process the output segmentation image in C++ cuda api (tensorrt)? in python we use torch.argmax() and interpolation for post processing. 