CNN-based-Methods-for-Tactile-Object-Recognition/TactNet_6_halfRes.m at master · TaISLab/CNN-based-Methods-for-Tactile-Object-Recognition · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
%% TactNet-6 METHOD
%==========================================================================
% Juan M. Gandarias, Jesús M. Gómez-de-Gabriel and Alfonso J. García-Cerezo
% Telerobotics and Interactive Systems Lab
% System Engineering and Automation Department
% 20/03/2019
% University of Málaga, Spain
% -------------------------------------------------------------------------
% This code has been implemented using the Neural Network Toolbox.
% The script returns the confusion matrix as a result for the classification
% of the tactile dataset saved in the 'Experiment_IEEE/Images' folder,
% by using the TactNet-6 approach with 1/4 resolution tactile images.
% =========================================================================


clear
%% Tactile Network Architecture
TactNet = [
    %input
    imageInputLayer([14 25 1], 'Name', 'Input')

    %conv1
    convolution2dLayer([5 5],8,'Padding','same','Name','conv1')
    batchNormalizationLayer('Name','batch1')
    reluLayer('Name','ReLU1')
    maxPooling2dLayer(2,'Stride',1,'Padding','same','Name','pool1')

    %conv2
    convolution2dLayer([3 3],16,'Padding','same','Name','conv2')
    batchNormalizationLayer('Name','batch2')
    reluLayer('Name','ReLU2')
    maxPooling2dLayer(2,'Stride',2,'Padding','same','Name','pool2')

    %conv3
    convolution2dLayer([3 3],32,'Padding','same','Name','conv3')
    batchNormalizationLayer('Name','batch3')
    reluLayer('Name','ReLU3')
    maxPooling2dLayer(2,'Stride',2,'Padding','same','Name','pool3')

    %conv4
    convolution2dLayer([3 3],64,'Padding','same','Name','conv4')
    batchNormalizationLayer('Name','batch4')
    reluLayer('Name','ReLU4')
    maxPooling2dLayer(2,'Stride',2,'Padding','same','Name','pool4')

    %conv4
    convolution2dLayer([2 2],128,'Padding','same','Name','conv5')
    batchNormalizationLayer('Name','batch5')
    reluLayer('Name','ReLU5')
    maxPooling2dLayer(2,'Stride',2,'Padding','same','Name','pool5')

%     %fc5
%     fullyConnectedLayer(128,'Name','fc5')
%     reluLayer('Name','ReLU5')
%     dropoutLayer(0.6, 'Name','dropout5')

    %fc6
    fullyConnectedLayer(22,'Name','fc6')
    softmaxLayer('Name','softmax')
    classificationLayer('Name','output')
    ];


%% Download Tactile Data (training and test)
images_dir = fullfile('Experiment_IEEE/Images_halfRes');

images = imageDatastore(images_dir,'IncludeSubfolders',true,'LabelSource',...
    'foldernames','FileExtensions','.jpg');

tbl = countEachLabel(images);

% Notice that each set now has exactly the same number of images.
countEachLabel(images);

% Split the data in training, valiation and test sets
[trainingSet, testSet] = splitEachLabel(images, 0.8, 'randomize');
[trainingSet, validationSet] = splitEachLabel(trainingSet, 0.8, 'randomize');

%% Data Augmentation
dataAugmenter = imageDataAugmenter( ...
    'RandXReflection',1,...
    'RandYReflection',1,...
    'RandXTranslation',[-2 2], ...
    'RandYTranslation',[-2 2],...
    'RandRotation',[-3,3]);


augmented_trainingSet = augmentedImageSource([14 25], trainingSet,...
    'DataAugmentation', dataAugmenter);

%% Training options
options = trainingOptions('sgdm', ...
    'MaxEpochs',5000, ...
    'Momentum',0.8,...
    'InitialLearnRate', 0.1,...
    'LearnRateSchedule', 'piecewise', ...
    'LearnRateDropFactor', 0.7, ...
    'LearnRateDropPeriod', 100, ...
    'MiniBatchSize', 4096,...
    'ValidationData',validationSet, ...
    'ValidationFrequency',15,...
    'ValidationPatience',7,...
    'ExecutionEnvironment','cpu'); %, ...
%     'VerboseFrequency',1, ...
%     'Verbose',1,...
%     'Plots','training-progress',...
%     'ExecutionEnvironment','gpu');


%% Training the Network
tic;
TactNet_trained = trainNetwork(augmented_trainingSet,TactNet,options);
training_time = toc;

%% Evaluate the performance of the network
tic
predictedLabels = classify(TactNet_trained,testSet);
testLabels = testSet.Labels;
toc

test_accuracy = sum(predictedLabels==testLabels)/numel(predictedLabels)

% Tabulate the results using a confusion matrix.
confMat = confusionmat(testLabels, predictedLabels);

% Convert confusion matrix into percentage form
confMatrix = bsxfun(@rdivide,confMat,sum(confMat,2));

% Plot the confusion matrix
% Plot_ConfMatrix(confMatrix);

% Calculate the recognition rate
recognition_rate = mean(diag(confMatrix));

%% Save the workspace
% save('training_TactNet6_ws')