SAN FRANCISCO, Nov. 15 morning news, Google today released TensorFlow lite, so that applications to developers can deploy artificial intelligence on mobile devices.This mobile Google open source artificial intelligence project was first in I / O developers Announced at the conference.
TensorFlow Lite is available for Android and iOS developers.
Since TensorFlow Lite was announced in May of this year, several competing products have appeared on mobile devices, including CoreML from Apple, Clarifai artificial intelligence cloud training service on mobile devices, and Kirin 970 used by Huawei Mate 10 Artificial intelligence processor.
TensorFlow Lite also ships with a limited set of pre-trained artificial intelligence models, including the MobileNet and Inception v3 object recognition computer models, and the Smart Replay natural language processing model. Custom models trained by developers with their own data sets can also be deployed on top.
TensorFlow team in the official blog said that more models and features will be gradually added according to user needs.
TensorFlow Lite uses the Android Neural Network API, which can be called directly without CPU acceleration to ensure compatibility with different devices.