Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Nov 27, 2018
As a mobile app developer, you probably have noticed Google's active efforts on bringing their machine learning expertise to mobile development. From TensorFlow's earliest mobile app demo to TensorFlow Lite and Android Neural Networks API, we are witnessing how latest research (e.g. MobileNet) drastically reduced machine learning model size and CPU / memory consumption on mobile devices. This year, with the beta release of ML Kit, we now have another powerful toolbox to leverage machine learning in the mobile application development. This talk features a side project Magritte, an educational application that helps people learn foreign languages. It's sort of like Duolingo but with eyes, the application recognizes daily objects using custom machine learning models embedded on device and speaks the vocabulary out loud with the chosen language. By presenting the Magritte app, the talk will reveal how the models for TensorFlow mobile were initially trained and optimized, how the performance was improved with TensorFlow Lite and MobileNets models, and eventually the ongoing migration towards ML Kit.