importKerasNetwork

Import a pretrained Keras network and weights

Syntax

net = importKerasNetwork(modelfile)

net = importKerasNetwork(modelfile,Name,Value)

Description

net = importKerasNetwork(modelfile) imports a pretrained TensorFlow™-Keras network and its weights from modelfile.

This function requires Deep Learning Toolbox™ Importer for TensorFlow-Keras Models support package. If this support package is not installed, the function provides a download link.

net = importKerasNetwork(modelfile,Name,Value) imports a pretrained TensorFlow-Keras network and its weights with additional options specified by one or more name-value pair arguments.

For example, importKerasNetwork(modelfile,'WeightFile',weights) imports the network from the model file modelfile and weights from the weight file weights. In this case, modelfile can be in HDF5 or JSON format, and the weight file must be in HDF5 format.

Examples

Download and Install Deep Learning Toolbox Importer for TensorFlow-Keras Models

Download and install the Deep Learning Toolbox Importer for TensorFlow-Keras Models support package.

Type importKerasNetwork at the command line.

importKerasNetwork

If the Deep Learning Toolbox Importer for TensorFlow-Keras Models support package is not installed, then the function provides a link to the required support package in the Add-On Explorer. To install the support package, click the link, and then click Install. Check that the installation is successful by importing the network from the model file 'digitsDAGnet.h5' at the command line. If the required support package is installed, then the function returns a DAGNetwork object.

modelfile = 'digitsDAGnet.h5';
net = importKerasNetwork(modelfile)

Warning: Saved Keras networks do not include classes. Classes
will be set to categorical(1:N), where N is the number of classes
in the classification output layer of the network.  To specify
classes, use the 'Classes' argument. 

net = 

  DAGNetwork with properties:

         Layers: [13×1 nnet.cnn.layer.Layer]
    Connections: [13×2 table]

Import and Plot Keras Network

Specify the file to import. The file digitsDAGnet.h5 contains a directed acyclic graph convolutional neural network that classifies images of digits.

modelfile = 'digitsDAGnet.h5';

Import the network.

net = importKerasNetwork(modelfile)

Warning: Saved Keras networks do not include classes. Classes will be set to categorical(1:N), where N is the number of classes in the classification output layer of the network.  To specify classes, use the 'Classes' argument.

net = 
  DAGNetwork with properties:

         Layers: [13×1 nnet.cnn.layer.Layer]
    Connections: [13×2 table]

Plot the network architecture.

figure
plot(net);
title('DAG Network Architecture')

Import Keras Network and Weights

Specify the network and the weight files to import.

modelfile = 'digitsDAGnet.json';
weights = 'digitsDAGnet.weights.h5';

This is a directed acyclic graph convolutional neural network trained on the digits data.

Import network architecture and import the weights from separate files. The .json file does not have an output layer or information on the cost function. Specify the output layer type when you import the files.

net = importKerasNetwork(modelfile,'WeightFile',weights, ...
      'OutputLayerType','classification')

Warning: Saved Keras networks do not include classes. Classes will be set to categorical(1:N), where N is the number of classes in the classification output layer of the network.  To specify classes, use the 'Classes' argument.

net = 
  DAGNetwork with properties:

         Layers: [13×1 nnet.cnn.layer.Layer]
    Connections: [13×2 table]

Import Pretrained Keras Network to Classify Image

Specify the model file.

modelfile = 'digitsDAGnet.h5';

Specify class names.

classNames = {'0','1','2','3','4','5','6','7','8','9'};

Import the Keras network with the class names.

net = importKerasNetwork(modelfile,'Classes',classNames);

Read the image to classify.

digitDatasetPath = fullfile(toolboxdir('nnet'),'nndemos','nndatasets', ...
    'DigitDataset');
I = imread(fullfile(digitDatasetPath,'5','image4009.png'));

Classify the image using the pretrained network.

label = classify(net,I);

Display the image and the classification result.

figure
imshow(I)
title(['Classification result: ' char(label)])

Input Arguments

modelfile — Name of Keras model file

character vector | string scalar

Name of the model file containing the network architecture, and possibly the weights, specified as a character vector or a string scalar. The file must be in the current folder, in a folder on the MATLAB^® path, or you must include a full or relative path to the file.

If modelfile includes

• The network architecture and weights, then it must be in HDF5 (.h5) format.

• Only the network architecture, then it can be in HDF5 or JSON (.json) format.

If modelfile includes only the network architecture, then you must supply the weights in an HDF5 file, using the 'WeightFile' name-value pair argument.

Example: 'digitsnet.h5'

Data Types: char | string

Name-Value Pair Arguments

Specify optional comma-separated pairs of Name,Value arguments. Name is the argument name and Value is the corresponding value. Name must appear inside quotes. You can specify several name and value pair arguments in any order as Name1,Value1,...,NameN,ValueN.

Example: importKerasNetwork(modelfile,'OutputLayerType','classification','Classes',classes) imports a network from the model file modelfile, adds an output layer for a classification problem at the end of the Keras layers, and specifies classes as the classes of the output layer.

'WeightFile' — Name of file containing weights

character vector | string scalar

Name of file containing weights, specified as a character vector or a string scalar. WeightFile must be in the current folder, in a folder on the MATLAB path, or you must include a full or relative path to the file.

Example: 'WeightFile','weights.h5'

'OutputLayerType' — Type of output layer

'classification' | 'regression' | 'pixelclassification'

Type of the output layer that the function appends to the end of the imported network architecture when modelfile does not specify a loss function, specified as 'classification', 'regression', or 'pixelclassification'. Appending a pixelClassificationLayer object requires Computer Vision Toolbox™.

If a network in modelfile has multiple outputs, then you cannot specify the output layer types using this argument. Use importKerasLayers instead. importKerasLayers inserts placeholder layers for the outputs. After importing, you can find and replace the placeholder layers by using findPlaceholderLayers and replaceLayer, respectively.

Example: 'OutputLayerType','regression'

'ImageInputSize' — Size of input images

vector of two or three numerical values

Size of the input images for the network, specified as a vector of two or three numerical values corresponding to [height,width] for grayscale images and [height,width,channels] for color images, respectively. The network uses this information when the modelfile does not specify the input size.

If a network in modelfile has multiple inputs, then you cannot specify the input sizes using this argument. Use importKerasLayers instead. importKerasLayers inserts placeholder layers for the inputs. After importing, you can find and replace the placeholder layers by using findPlaceholderLayers and replaceLayer, respectively.

Example: 'ImageInputSize',[28 28]

'Classes' — Classes of the output layer

'auto' (default) | categorical vector | string array | cell array of character vectors

Classes of the output layer, specified as a categorical vector, string array, cell array of character vectors, or 'auto'. If you specify a string array or cell array of character vectors str, then the software sets the classes of the output layer to categorical(str,str). If Classes is 'auto', then the function sets the classes to categorical(1:N), where N is the number of classes.

Data Types: char | categorical | string | cell

Output Arguments

net — Pretrained Keras network

SeriesNetwork object | DAGNetwork object

Pretrained Keras network, returned as one of the following:

• If the Keras network is of type Sequential, then net is a SeriesNetwork object.

• If the Keras network is of type Model, then net is a DAGNetwork object.

Tips

• importKerasNetwork can import a network with the following Keras layer types, with some limitations. If the network contains any other type of layer, then the software returns an error message. In this case, you can still use importKerasLayers to import the network architecture and weights.

  Supported Keras Layer	Corresponding Deep Learning Toolbox Layer
  Add	additionLayer
  Activation, with activation names: 'elu' 'relu' 'linear' 'softmax' 'sigmoid' 'tanh'	Layers: eluLayer reluLayer or clippedReluLayer None softmaxLayer nnet.keras.layer.SigmoidLayer tanhLayer
  Advanced activations: ELU Softmax ReLU LeakyReLU PReLu*	Layers: eluLayer softmaxLayer reluLayer, clippedReluLayer, or leakyReluLayer leakyReluLayer nnet.keras.layer.PreluLayer
  AveragePooling2D	averagePooling2dLayer
  BatchNormalization	batchNormalizationLayer
  Bidirectional(LSTM(__))	bilstmLayer
  Concatenate	depthConcatenationLayer
  Conv2D	convolution2dLayer
  Conv2DTranspose	transposedConv2dLayer
  CuDNNLSTM	lstmLayer
  Dense	fullyConnectedLayer
  DepthwiseConv2D	groupedConvolution2dLayer
  Dropout	dropoutLayer
  Embedding	wordEmbeddingLayer (Text Analytics Toolbox™)
  Flatten	nnet.keras.layer.FlattenCStyleLayer
  GlobalAveragePooling2D	globalAveragePooling2dLayer
  GlobalMaxPooling2D	globalMaxPooling2dLayer
  GRU	gruLayer
  Input	imageInputLayer
  LSTM	lstmLayer
  MaxPooling2D	maxPooling2dLayer
  SeparableConv2D	groupedConvolution2dLayer or convolution2dLayer
  ZeroPadding2D	nnet.keras.layer.ZeroPadding2DLayer

  *For a PReLU layer, importKerasNetwork replaces a vector-valued scaling parameter with the average of the vector elements. You can change the parameter back to a vector after import. For an example, see Import Keras PReLU Layer.

• importKerasNetwork supports the following Keras loss functions:

  • mean_squared_error

  • categorical_crossentropy

  • sparse_categorical_crossentropy

  • binary_crossentropy

• You can import a Keras network with multiple inputs and multiple outputs (MIMO). Use importKerasNetwork if the network includes input size information for the inputs and loss information for the outputs. Otherwise, use importKerasLayers. The importKerasLayers function inserts placeholder layers for the inputs and outputs. After importing, you can find and replace the placeholder layers by using findPlaceholderLayers and replaceLayer, respectively. The workflow for importing MIMO Keras networks is the same as the workflow for importing MIMO ONNX™ networks. For an example, see Import ONNX Network with Multiple Outputs. To learn about a deep learning network with multiple inputs and multiple outputs, see Multiple-Input and Multiple-Output Networks.

• To use a pretrained network for prediction or transfer learning on new images, you must preprocess your images in the same way as the images used to train the imported model were preprocessed. Resizing images, subtracting the average image, and converting the images from RGB to BGR format are the most common preprocessing operations.

  • To resize images, use imresize. For example, imresize(im,[227 227]).

  • To convert images from RGB to BGR format, use flip. For example, flip(im,3).

  For more information on preprocessing images for training and prediction, see Preprocess Images for Deep Learning.

Compatibility Considerations

'ClassNames' option will be removed

Not recommended starting in R2018b

'ClassNames' will be removed. Use 'Classes' instead. To update your code, replace all instances of 'ClassNames' with 'Classes'. There are some differences between the corresponding properties in classification output layers that require additional updates to your code.

The ClassNames property of a classification output layer is a cell array of character vectors. The Classes property is a categorical array. To use the value of Classes with functions that require cell array input, convert the classes using the cellstr function.

References

[1] Keras: The Python Deep Learning library. https://keras.io.