Convert TensorFlow Graph to UFF using C++ API


After scavenging the internet and documentations regarding how to convert a Tensor Flow graph to UFF format, i have found plenty of resources for doing this using Python API, which is not available on TEGRA platforms

Therefore, can anyone point how to achieve the same using C++ API, on TEGRA(TX2) platforms ??

Hi wahaj_ahmad, since TensorFlow is challenging to install natively on arm64 architecture and Jetson, and the TensorRT Python API is not currently available, it is recommend to perform the TensorFlow graph conversion/export procedure on host PC (i.e. platform that may be doing the training).

The exported UFF model can then be copied to the Jetson and loaded at runtime using TensorRT’s NvUffImporter C++ API.