1 # TF Lite Android App Example 2 3 A simple Android example that demonstrates image classification and object 4 detection using the camera, as well as speech recognition using the microphone. 5 6 ## Building in Android Studio with TensorFlow Lite AAR from JCenter. 7 The build.gradle is configured to use TensorFlow Lite's nightly build. 8 9 If you see a build error related to compatibility with Tensorflow Lite's Java 10 API (example: method X is undefined for type Interpreter), there has likely been 11 a backwards compatible change to the API. You will need to pull new app code 12 that's compatible with the nightly build and may need to first wait a few days 13 for our external and internal code to merge. 14 15 ## Building from Source with Bazel 16 17 1. Follow the [Bazel steps for the TF Demo App](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/examples/android#bazel): 18 19 1. [Install Bazel and Android Prerequisites](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/examples/android#install-bazel-and-android-prerequisites). 20 It's easiest with Android Studio. 21 22 - You'll need at least SDK version 23. 23 - Make sure to install the latest version of Bazel. Some distributions 24 ship with Bazel 0.5.4, which is too old. 25 - Bazel requires Android Build Tools `26.0.1` or higher. 26 - You also need to install the Android Support Repository, available 27 through Android Studio under `Android SDK Manager -> SDK Tools -> 28 Android Support Repository`. 29 30 2. [Edit your `WORKSPACE`](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/examples/android#edit-workspace) 31 to add SDK and NDK targets. 32 33 NOTE: As long as you have the SDK and NDK installed, the `./configure` 34 script will create these rules for you. Answer "Yes" when the script asks 35 to automatically configure the `./WORKSPACE`. 36 37 - Make sure the `api_level` in `WORKSPACE` is set to an SDK version that 38 you have installed. 39 - By default, Android Studio will install the SDK to `~/Android/Sdk` and 40 the NDK to `~/Android/Sdk/ndk-bundle`. 41 42 2. Build this demo app with Bazel. The demo needs C++11. We configure the fat_apk_cpu flag to package support for 4 hardware variants. You may replace it with --config=android_arm64 on a 64-bit device and --config=android_arm for 32-bit device: 43 44 ```shell 45 bazel build -c opt --cxxopt='--std=c++11' --fat_apk_cpu=x86,x86_64,arm64-v8a,armeabi-v7a \ 46 //tensorflow/lite/examples/android:tflite_demo 47 ``` 48 49 3. Install the demo on a 50 [debug-enabled device](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/examples/android#install): 51 52 ```shell 53 adb install bazel-bin/tensorflow/lite/examples/android/tflite_demo.apk 54 ``` 55
