Hands-On Python Deep Learning for the Web

Loading data into the script runtime and setting the model

Once we have imported the necessary modules, we load the saved model JSON and weights, as in the following code snippet:

json_file = open('model.json','r')model_json = json_file.read()json_file.close()model = model_from_json(model_json)model.load_weights("weights.h5")model.compile(loss='categorical_crossentropy',optimizer='adam',metrics=['accuracy'])graph = tf.get_default_graph()

Note that we have also created a default graph item for the session ahead. This was implicitly created during the model training but is not carried over in the saved model and weights files, so we must explicitly create it here.