RNTFLiteExample

This is a plug-and-play example using react-native-tflite

Getting stared

1. Clone this repository

git clone https://github.com/kainolophobia/RNTFLiteExample.git

2. Install and link dependencies

npm install
react-native link

3. Install TensorFlow-experimental

cd ios
pod install

4. Build TensorFlow Lite v1.5.1 for iOS

Clone Tensorflow

git clone https://github.com/tensorflow/tensorflow.git

Checkout v1.5.1

git fetch --all --tags --prune
git checkout tags/v1.5.1

Follow the documentation to build here, skipping the last section titled "Using in your own application"

5. Link TensorFlow Lite

Open RNTFLiteExample.xcworkspace instead of RNTFLiteExample.xcodeproj
In the Xcode project's "Libraries" folder find RNTFLite.xcodeproj, then in "Build Settings" modify TF_ROOT to fit your Tensorflow v1.5.1 path
Add the library at $(TF_ROOT)/tensorflow/contrib/lite/gen/lib/libtensorflow-lite.a to your linking build stage

6. Run App

react-native run-ios --device "[DEVICE NAME]"

Example screenshot:

Notes

The .tflite model and associated labels are hackiliy stored at the project's root and hardcoded in RNTFLiteView.mm within the RNTFLite.xcodeproj Library.

To add your own model/labels, right-click the RNTFLiteExample project and select "Add files to project..."

Make sure to update the associated model/label filepaths in the RNTFLiteView.mm file.

Additionally, this can break in the simulator. Please use an actual device to test.

Support

Many thanks to Github user kaka-lin for creating this proof-of-concept. This repository summarizes the steps needed to build and add Tensorflow Lite to a React Native project using the react-native-tflite package; the main changes involve highlighting a working version of Tensorflow and simplifying the linking steps.

adammichaelwilliams/RNTFLiteExample