Running the keras neural network model on android mobile device

dependencies.md

• dependencies for android.
  • add the following lines to build.gradle

   compile 'org.deeplearning4j:deeplearning4j-core:0.7.2'
   compile 'org.nd4j:nd4j-native:0.7.2'
   compile 'org.nd4j:nd4j-native:0.7.2:android-x86'
   compile 'org.nd4j:nd4j-native:0.7.2:android-arm'
   

• For more look into the following
  • https://github.com/gauravshelangia/KerasModel_Android/blob/master/src/main/java/ImportKerasModel.java
  • https://github.com/gauravshelangia/MtargetClient/blob/master/app/src/main/java/com/example/gaurav/mtargetclient/ImportKerasModel.java

KerasOnAndroid.md

To run the keras model follow the below instructions

• Save the keras model as json file and model weight in h5 file

  model.save('mtarget_model_full1.h5') 
  # save model in json file
  # serialize model to JSON
  model_json = model.to_json()
  with open("Model_json", "w") as json_file:
      json_file.write(model_json)
  
  # Save model weights
  model.save_weights('Model_weights')

• import the saved model into dl4j

      String modelJsonFilename = "path_to_Model_json";
      String weightsHdf5Filename = "path_to_model_weights";
      
      // load model from two different file one : json flie having json config and another: weights file
       MultiLayerNetwork model =     KerasModelImport.importKerasSequentialModelAndWeights(modelJsonFilename,weightsHdf5Filename);
   
       //Save the model
       File locationToSave = new File("MyMultiLayerNetwork.zip"); //Where to save the model as zip file      
       boolean saveUpdater = true
       ModelSerializer.writeModel(model, locationToSave, saveUpdater);

• Last load the model (zip file) on android device

       File dir = new File(Environment.getExternalStorageDirectory(),"/Model");
       File modelzip = new File(dir,"MyMultiLayerNetwork.zip");
       //Load the model
       MultiLayerNetwork restored = ModelSerializer.restoreMultiLayerNetwork(modelzip);