This is a basic example of train and use a basic Keras neural network model (XOR) on iPhone using Apple’s coremltools on iOS11. Note that showing the integration starting from a Keras model to having it running in the iOS app is the main point and not the particular choice of model, in principle a similar approach could be used for any kind of Deep Learning model, e.g. generator part of Generative Adversarial Networks, a Recurrent Neural Network (or LSTM) or a Convolutional Neural Network.
For easy portability I chose to run the Keras part inside docker (i.e. could e.g. usenvidia-docker for a larger model that would need a GPU to train e.g. in the cloud or on a desktop or a powerful laptop). The current choice of Keras backend was TensorFlow, but believe it should also work for other backends (e.g. CNTK, Theano or MXNet). The code for this blog post is available at github.com/atveit/keras2ios
Best regards,
1. Building and training Keras Model for XOR problem – PYTHON
1.1 Training data for XOR
Python
0 xor 0 = 0 # 0 xor 1 = 1 # 1 xor 0 = 1 # 1 xor 1 = 0 X = np.array([[0,0],[0,1],[1,0],[1,1]]) y = np.array([[0],[1],[1],[0]])
| 123456 | # 0 xor 0 = 0# 0 xor 1 = 1# 1 xor 0 = 1# 1 xor 1 = 0X = np.array([[0,0],[0,1],[1,0],[1,1]])y = np.array([[0],[1],[1],[0]]) |
|---|
1.2 Keras XOR Neural Network Model
Python
model = Sequential() model.add(Dense(8, input_dim=2)) # fully connected layer with 8 hidden nodes and input dimension of 2 model.add(Activation('tanh')) model.add(Dense(1)) model.add(Activation('sigmoid')) # 1 output, either 0 or 1
| 12345 | model = Sequential()model.add(Dense(8, input_dim=2)) # fully connected layer with 8 hidden nodes and input dimension of 2model.add(Activation('tanh'))model.add(Dense(1))model.add(Activation('sigmoid')) # 1 output, either 0 or 1 |
|---|
1.3 Train the Keras model with Stochastic Gradient Descent (SGD)
Python
sgd = SGD(lr=0.1) model.compile(loss='binary_crossentropy', optimizer=sgd) model.fit(X, y, nb_epoch=1000, batch_size=1)
| 123 | sgd = SGD(lr=0.1)model.compile(loss='binary_crossentropy', optimizer=sgd)model.fit(X, y, nb_epoch=1000, batch_size=1) |
|---|
1.4 Use Apple’s coreml tool to convert the Keras model to coreml model
Python
import coremltools coreml_model = coremltools.converters.keras.convert(model) coreml_model.save("keras_model.mlmodel")
| 123 | import coremltoolscoreml_model = coremltools.converters.keras.convert(model)coreml_model.save("keras_model.mlmodel") |
|---|
2. Using the converted Keras model on iPhone – SWIFT
2.1 Create new Xcode Swift project and add keras_model.mlmodel
2.2 Inspect keras_model.mlmodel by clicking on it in xcode
2.3 Update ViewController.swift with prediction function
Swift
func xor_with_keras(x:NSNumber,y:NSNumber) -> String{ // input data is a float32 array with 2 elements guard let input_data = try? MLMultiArray(shape:[2], dataType:MLMultiArrayDataType.float32) else { fatalError("Unexpected runtime error. MLMultiArray") } input_data[0] = x input_data[1] = y let i = keras_modelInput(input1: input_data) // actual prediction guard let xor_prediction = try? model.prediction(input: i) else { fatalError("Unexpected runtime error. model.prediction") } let result = String(describing: xor_prediction.output1[0]) print(input_data[0].stringValue + " XOR " + input_data[1].stringValue + " = " + result) return result }
| 123456789101112131415161718192021 | func xor_with_keras(x:NSNumber,y:NSNumber) -> String{ // input data is a float32 array with 2 elements guard let input_data = try? MLMultiArray(shape:[2], dataType:MLMultiArrayDataType.float32) else { fatalError("Unexpected runtime error. MLMultiArray") } input_data[0] = x input_data[1] = y let i = keras_modelInput(input1: input_data) // actual prediction guard let xor_prediction = try? model.prediction(input: i) else { fatalError("Unexpected runtime error. model.prediction") } let result = String(describing: xor_prediction.output1[0]) print(input_data[0].stringValue + " XOR " + input_data[1].stringValue + " = " + result) return result } |
|---|
2.4 Run app with Keras model on iPhone and look at debug output
0 xor 0 = 1 xor 1 = 0 (if rounding down), and 1 xor 0 = 0 xor 1 = 1 (if rounding up)
LGTM!
Sign up for Deep Learning newsletter!
Email address