Android Things TensorFlow image classifier sample

Опубликовано 22 декабря 2017, 22:12

This is a video of our Android Things sample sample-tensorflow-imageclassifier.

This sample demonstrates how to run TensorFlow inference on Android Things. When a GPIO button is pushed, the current image is captured from an attached camera. The captured image is then converted and piped into a TensorFlow model that identifies what is in the image. Up to three labels returned by the TensorFlow network is shown on logcat and on the screen, if there is an attached display. Also, the result is spoken out loud using text-to-speech and sent to an attached speaker, if any.

This project is based on the TensorFlow Android Camera Demo TF_Classify app. The TensorFlow training was done using Google inception model and the trained data set is used to run inference and generate classification labels via TensorFlow Android Inference APIs.

You can download the source code here: goo.gl/M5Cvp5

Android Things is provided by Google and enables developers to build their own production Internet of Things devices and maintain them at scale. Learn more about the Android Things SDK at developer.android.com/things and join our community at g.co/iotdev.

Watch more samples for Android Things here: goo.gl/tjYXHF

Subscribe to the Android Developers channel here: goo.gl/vLYDU