Generate MATLAB Code from Deep Network Designer

The Deep Network Designer app enables you to generate MATLAB^® code that recreates the building, editing, and training of a network in the app.

In the Designer tab, you can generate a live script to:

Recreate the layers in your network. Select Export > Generate Code.

Recreate the layers in your network, including any initial parameters. Select Export > Generate Code with Initial Parameters.

In the Training tab, you can generate a live script to:

Recreate the building and training of an image classification network you construct in Deep Network Designer. Select Export > Generate Code for Training.

Generate MATLAB Code to Recreate Network Layers

Generate MATLAB code for recreating the network constructed in Deep Network Designer. In the Designer tab, choose one of these options:

To recreate the layers in your network, select Export > Generate Code. This network does not contain initial parameters, such as pretrained weights.
To recreate the layers in your network, including any initial parameters, select Export > Generate Code with Initial Parameters. The app creates a live script and a MAT-file containing the initial parameters (weights and biases) from your network. Run the script to recreate the network layers, including the learnable parameters from the MAT-file. Use this option to preserve the weights if you want to perform transfer learning.

Running the generated script returns the network architecture as a variable in the workspace. Depending on the network architecture, the variable is a layer graph named lgraph or a layer array named layers.

You can supply the generated layer graph or layer array to the trainNetwork function. For example, assume that the layers produced by the generated code are named lgraph_1, the images are in an augmented image datastore called images and options contains the training options. To train the network, type:

trainedNet = trainNetwork(images,lgraph_1,options)

For an example of training a network exported from Deep Network Designer, see Create Simple Sequence Classification Network Using Deep Network Designer.

Generate MATLAB Code to Train Network

To recreate the construction and training of an image classification network in Deep Network Designer, generate MATLAB code. For an example of using Deep Network Designer to train a network, see Transfer Learning with Deep Network Designer.

Once training is complete, on the Training tab, select Export > Generate Code for Training. The app creates a live script and a MAT-file containing the initial parameters (weights and biases) from your network. If you import data as an image datastore then this is also contained in the generated MAT-file.

Running the generated script builds the network (including the learnable parameters from the MAT-file), imports the data, sets the training options, and trains the network. Examine the generated script to learn how to construct and train a network at the command line.

Use Network for Prediction

Suppose that the trained network is contained in the variable net. To use the trained image classification network for prediction, use the predict function. For example, use the network to predict the class of peppers.png.

img = imread("peppers.png");
img = imresize(img, net.Layers(1).InputSize(1:2));
label = predict(net, img);
imshow(img);
title(label);

References

[1] Kudo, Mineichi, Jun Toyama, and Masaru Shimbo. “Multidimensional Curve Classification Using Passing-through Regions.” Pattern Recognition Letters 20, no. 11–13 (November 1999): 1103–11. https://doi.org/10.1016/S0167-8655(99)00077-X.

[2] Kudo, Mineichi, Jun Toyama, and Masaru Shimbo. Japanese Vowels Data Set. Distributed by UCI Machine Learning Repository. https://archive.ics.uci.edu/ml/datasets/Japanese+Vowels.

Documentation