Summary

In this chapter, we first described how a drawing classification model works, then covered how to train such a model using the high-level TensorFlow Estimator API. We looked at how to write Python code to make predictions with a trained model, then discussed in great detail how to find the right input and output node names and how to freeze and transform the model in the right way so mobile apps can use it. We also offered a new method to build a new TensorFlow custom iOS library, and a step-by-step tutorial on building a TensorFlow custom library for Android, to fix runtime errors when using the model. Finally, we showed the iOS and Android code that captures and shows user drawings, converts them to the data expected by the model, and processes and presents the classification results returned by the model. Hopefully, you have learned as much as you have had fun over...