example_basics_frozen_graph

Basic example

This is a very basic example wich demonstrates how to load and infer a Frozen Graph created using TensorFlow. Frozen GraphDef is the standard of storing a graph for versions of TensorFlow < 1.15. We wanted to support this feature for legacy reasons.

TensorFlow2

We will create a dummy graph for demonstration purposes. To create the model simply run the following command:

python main.py

openFrameworks

By default ofxTF2::Model will use the SavedModel format. To use the frozen GraphDef format you may either add the type to the constructor or call setModelType() afterwards. As the default names differ from the names in the SavedModel format make sure to overwrite names of the ins and outs by calling setup().

// set model type and i/o names
model.setModelType(cppflow::model::TYPE::FROZEN_GRAPH);
model.setup({{"x:0"}}, {{"Identity:0"}});

Afterwards you can load the pb file using the load() function.

// load the model, bail out on error
if(!model.load("model.pb")) {
	std::exit(EXIT_FAILURE);
}

Everything else should work the same.

Please understand that we wont be able to invest a lot of time in supporting this feature in the future.