DNN for image classification

An example of how to use the SqueezeNet model to classify an image.

The SqueezeNet model was trained for Image Classification.

For details pertaining to the usage of the model, have a look at this repository. You can infact train your own object detection models with the loss function which is implemented (in Caffe).

The size of the model being 4.9MB, just takes a time of 0.136401 seconds to classify an image.

Use class labels file synset_words.txt to map the prediction class index to class name.

References:

Sources:

load the network along with its trained weights

[net, blobOpts] = SqueezeNet();
assert(~net.empty(), 'Could not load model');

load test image

im = fullfile(mexopencv.root(), 'test', 'space_shuttle.jpg');
img = cv.imread(im, 'Color',true, 'FlipChannels',false);

convert the image into blob, and set the input blob

blob = cv.Net.blobFromImages(img, blobOpts{:});
net.setInput(blob);

get the output of the "predictions" layer probs is a 4D tensor of shape 1x1000x1x1, the output of softmax activation

tic
probs = net.forward('predictions');
toc

Elapsed time is 0.364908 seconds.

max prediction

probs = probs(:);  % 1000x1
[max_prob, class_idx] = max(probs);
fprintf('Best class Index: %d\n', class_idx);
fprintf('Probability: %f\n', max_prob*100);

Best class Index: 813
Probability: 16.019295

Pretrained models

function dname = get_dnn_dir(dname)
    %GET_DNN_DIR  Path to model files, and show where to get them if missing

    dname = fullfile(mexopencv.root(), 'test', 'dnn', dname);
    b = isdir(dname);
    if ~b
        % display help of calling function
        % (assumed to be a local function in current file)
        st = dbstack(1);
        help([mfilename() filemarker() st(1).name])
    end
    assert(b, 'Missing model: %s', dname);
end

function [net, blobOpts] = SqueezeNet()
    %SQUEEZENET  SqueezeNet model trained on ImageNet 2012 Dataset [Caffe]
    %
    % homepage = https://github.com/kvmanohar22/caffe
    % homepage = https://kvmanohar22.github.io/GSoC/
    % homepage = https://github.com/opencv/opencv_3rdparty/tree/dnn_objdetect_20170827
    % homepage = https://github.com/opencv/opencv_contrib/tree/3.4.1/modules/dnn_objdetect/samples/data
    %
    % ## Model
    %
    % file = test/dnn/SqueezeDet/SqueezeNet_deploy.prototxt
    % url  = https://github.com/opencv/opencv_contrib/raw/3.4.1/modules/dnn_objdetect/samples/data/SqueezeNet_deploy.prototxt
    % hash = b518ef2a11a72aea6ef1d0170ac7cc239d8a4694
    %
    % ## Weights
    %
    % file = test/dnn/SqueezeDet/SqueezeNet.caffemodel
    % url  = https://github.com/opencv/opencv_3rdparty/raw/dnn_objdetect_20170827/SqueezeNet.caffemodel
    % hash = 507d7975c86591361a30efffb1b31270cf8fa650
    % size = 4.77 MB
    %

    dname = get_dnn_dir('SqueezeDet');
    net = cv.Net('Caffe', ...
        fullfile(dname, 'SqueezeNet_deploy.prototxt'), ...
        fullfile(dname, 'SqueezeNet.caffemodel'));
    blobOpts = {'SwapRB',false, 'Crop',false, 'Size',[416 416], 'Mean',[104 117 123]};
end