Train Once, Deploy Everywhere
Models can be trained for a variety of functions, such as classifying objects in video, interpreting gestures, or detecting patterns in financial transactions. Training requires large data sets and substantial computing resources. But once a model is trained and optimized, it may be widely deployed wherever it is needed, in consumer or industrial devices. This ability to deploy intelligent inference power everywhere will create an “Intelligent Internet of Things” and enable many solutions that improve our lives.