I am currently studying the various frameworks for deep learning, in particular for the training and deployment of convolutional neural networks. The requirements are that it can be trained on a regular PC using a graphics processor, but then the trained model must be deployed to three major mobile operating systems: Android, iOS and Windows Phone.
TensorFlow caught my eye because of its simplicity and great python interface. There is an example Android application ( https://jalammar.imtqy.com/Supercharging-android-apps-using-tensorflow/ ), but I'm not sure if it can also be deployed to iOS and Windows Phone? If not, can you recommend an alternative structure that meets these requirements? Ideally, using a simple scripting interface for rapid prototyping?
Thanks so much for your answers!
EDIT: I am currently testing Microsoft CNTK . Based on Windows and Linux from the source code, it works great, it can be expanded into the Lego blocks mode, and its own NDL (Network Description Language) is very easy to read and learn and provides enough freedom to create many different neural Network architectures. The execution mechanism is only a small part of the framework, and it can be read in the NN model defined by NDL, as well as in trained parameters. I will keep the message updated on how the process of transferring to ARM processors is going.
source share