From Keras to C
This small demo project is about deploying deep learning models on embedded platforms. The techniques exposed here have been particularly useful to me in the deployment of deep learning models in industrial applications.
We start with a simple example model, trained with Tensorflow 1.x + Keras. In the end, we'll freeze the model and export a GraphDef that can be loaded and executed through the Tensorflow C API (without Python). The training and export are coded into the file train_and_export.py whereas the inference is coded into the file model_run.cpp. The Tensorflow binaries must be loaded into the folder lib, they can be downloaded from the following link.
Note Once the API for Tensorflow 2.0 will be stable the project will be updated to the newest version of the framework.