In Chapter 2, Classifying Images with Transfer Learning, we described how to add TensorFlow to your own Android app, but without any UI. Let's create a new Android app to use the fast-style transfer models we trained earlier and used in iOS.
Because this Android app offers a good opportunity to use the minimal TensorFlow-related code and Android UI and threaded code to run a complete TensorFlow model-powered app, we'll go through each line of the code added from scratch to help you further understand what it takes to develop an Android TensorFlow app from scratch:
-
In Android Studio, select File | New | New Project... and enter FastNeuralTransfer as the Application Name; accept all the defaults before clicking Finish.
-
Create a new assets folder, as shown in Figure 2.13, and drag the fast neural transfer models you...