Generate Code and Deploy SqueezeNet Network to Raspberry Pi

Third-Party Prerequisites

This example is not supported for MATLAB Online.

The squeezenet_predict Entry-point Function

This example uses a dlnetwork network object to show image classification. The squeezenet_predict function uses the imagePretrainedNetwork to load the pretrained SqueezeNet network into a persistent dlnetwork object. On subsequent calls, the function reuses the object. A dlarray object is created within the entry-point function, input and output to the function are of single datatype. For more information, see Code Generation for dlarray (MATLAB Coder).

type squeezenet_predict

function out = squeezenet_predict(in) 
%#codegen
% Copyright 2018-2024 The MathWorks, Inc.

    dlIn = dlarray(in,'SSC');
    
    persistent dlnet;
    if isempty(dlnet)
           dlnet = imagePretrainedNetwork('squeezenet');
    end
    
    dlOut = predict(dlnet, dlIn);
        
    out = extractdata(dlOut);
end

Create Connection to Raspberry Pi

Use the MATLAB Support Package for Raspberry Pi Hardware function raspi to create a connection to the Raspberry Pi.

r = raspi;

If this is your first time connecting to a Raspberry Pi board or you want to connect to a different board, use this line of code and replace:

r = raspi('raspiname','username','password');

Configure Code Generation Hardware Parameters for Raspberry Pi

Create a coder.EmbeddedCodeConfig object to generate a static library. Set the VerificationMode property to 'PIL' to enable PIL-based execution. By default, the DeepLearningConfig property is a coder.DeepLearningCodeConfig (MATLAB Coder) object and target library is set to 'none'.

cfg = coder.config('lib','ecoder',true);
cfg.VerificationMode = 'PIL';

Create a coder.Hardware object and specify the target hardware as Raspberry Pi. Assign hardware object to the Hardware property of cfg.

hw = coder.hardware('Raspberry Pi');
cfg.Hardware = hw;

Specify a build folder on the Raspberry Pi.

buildDir = '/home/pi/squeezenet_predict';
cfg.Hardware.BuildDir = buildDir;

Generate PIL MEX Function

Run the codegen command by specifying an input size of 227-by-227-by-3. This value corresponds to the input layer size of the SqueezeNet network.

codegen -config cfg squeezenet_predict -args {ones(227, 227, 3, 'single')}

### Connectivity configuration for function 'squeezenet_predict': 'Raspberry Pi'
Location of the generated elf : /home/pi/squeezenet_predict/MATLAB_ws/R2024a/C/EM_Path/ExampleManager/yangzhu.Bdoc.24b/deeplearning_shared-ex98393496/codegen/lib/squeezenet_predict/pil
Code generation successful.

Perform Prediction on Test Image

Load an input image. Call squeezenet_predict_pil on the input image.

I = imread('coffeemug.png');
I = imresize(I,[227,227]);
output = squeezenet_predict_pil(single(I));

### Starting application: 'codegen\lib\squeezenet_predict\pil\squeezenet_predict.elf'
    To terminate execution: clear squeezenet_predict_pil
### Launching application squeezenet_predict.elf...

Multiply the output value by 100 to convert scores into percentage. Display the top five classification labels on the image.

[val,indx] = sort(output, 'descend');
scores = val(1:5)*100;
[~,classNames] = imagePretrainedNetwork('squeezenet');
top5labels = classNames(indx(1:5));

outputImage = zeros(227,400,3, 'uint8');
for k = 1:3
    outputImage(:,174:end,k) = I(:,:,k);
end

scol = 1;
srow = 20;

for k = 1:5
        outputImage = insertText(outputImage, [scol, srow], ...
            [top5labels{k},' ',num2str(scores(k), '%2.2f'),'%'], ...
            'TextColor', 'w','FontSize',15, 'BoxColor', 'black');
    srow = srow + 20;
end

imshow(imresize(outputImage,[448,800]));

Clear Generated PIL Executable

clear squeezenet_predict_pil

### Host application produced the following standard output (stdout) and standard error (stderr) messages: