hCNN/example_MicroDoppler.m at master · pzhg/hCNN · 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
clc;
clear all;
close all;
dbstop if error

mpiprofile on -historysize 150000000;

%% Load Data (MicroDoppler)
load('MicroDoppler.mat');
% The MicroDoppler.mat file can be downloaded from https://1drv.ms/u/s!Akr-loZjbPYVufFK8J6pAMtHi1fEyA?e=T6Vgss

%% Training Options
to.epochs = 5;              % Epoch number
to.batch = 78;              % Batch number
to.batch_size = 64;         % Batch size
to.alpha = 0.1;             % Learning raten
to.momentum = 0.9;          % Momentum
to.mom = 0.5;               % Initial momentum
to.momIncrease = 20;        % Momemtum change iteration count
to.lambda = 0.0001;         % Weight decay parameter (a.k.a. L2 regularization parameter)
to.useGPU = 1;              % Use GPU

if to.useGPU == 1
    reset(gpuDevice(1));    % Initialize GPU
end

%% Initialize CNN
cnn = cnnInit(to);
feature('SetPrecision', 24);

%% Configure Layers
cnn = cnnAddInputLayer(cnn, [400, 144], 1);

    % subnet 1: CNN
    cnn1 = cnnInit(to);
    cnn1 = cnnAddInputLayer(cnn1, [400, 144], 1);
    cnn1 = cnnAddTransformLayer(cnn1, 'PCA', 1:8);
    cnn1 = cnnAddConvLayer(cnn1, [5, 5], 8, 'r');
    cnn1 = cnnAddBNLayer(cnn1);
    cnn1 = cnnAddActivationLayer(cnn1, 'relu');
    cnn1 = cnnAddPoolLayer(cnn1, 'max', [2, 2]);
    cnn1 = cnnAddConvLayer(cnn1, [3, 3], 8, 'r');
    cnn1 = cnnAddBNLayer(cnn1);
    cnn1 = cnnAddActivationLayer(cnn1, 'relu');
    cnn1 = cnnAddPoolLayer(cnn1, 'max', [2, 2]);
    cnn1 = cnnAddConvLayer(cnn1, [3, 3], 8, 'r');
    cnn1 = cnnAddBNLayer(cnn1);
    cnn1 = cnnAddActivationLayer(cnn1, 'relu');
    cnn1 = cnnAddPoolLayer(cnn1, 'max', [2, 2]);
    cnn1 = cnnAddConvLayer(cnn1, [3, 3], 8, 'r');
    cnn1 = cnnAddBNLayer(cnn1);
    cnn1 = cnnAddActivationLayer(cnn1, 'relu');
    cnn1 = cnnAddPoolLayer(cnn1, 'max', [2, 2]);
    cnn1 = cnnAddReshapeLayer(cnn1);
    cnn1 = cnnAddFCLayer(cnn1, 64, 'r');
    cnn1 = cnnAddActivationLayer(cnn1, 'relu');
    cnn1 = cnnAddOutputLayer(cnn1, 'end_BLOB');
    cnn1 = cnnInitVelocity(cnn1);

    % subnet 2: PCA
    cnn2 = cnnInit(to);
    cnn2 = cnnAddInputLayer(cnn2, [400, 144], 1);
    cnn2 = cnnAddTransformLayer(cnn2, 'PCA', 8:25);
    cnn2 = cnnAddConvLayer(cnn2, [5, 5], 8, 'r');
    cnn2 = cnnAddBNLayer(cnn2);
    cnn2 = cnnAddActivationLayer(cnn2, 'relu');
    cnn2 = cnnAddPoolLayer(cnn2, 'max', [2, 2]);
    cnn2 = cnnAddConvLayer(cnn2, [3, 3], 8, 'r');
    cnn2 = cnnAddBNLayer(cnn2);
    cnn2 = cnnAddActivationLayer(cnn2, 'relu');
    cnn2 = cnnAddPoolLayer(cnn2, 'max', [2, 2]);
    cnn2 = cnnAddConvLayer(cnn2, [3, 3], 8, 'r');
    cnn2 = cnnAddBNLayer(cnn2);
    cnn2 = cnnAddActivationLayer(cnn2, 'relu');
    cnn2 = cnnAddPoolLayer(cnn2, 'max', [2, 2]);
    cnn2 = cnnAddConvLayer(cnn2, [3, 3], 8, 'r');
    cnn2 = cnnAddBNLayer(cnn2);
    cnn2 = cnnAddActivationLayer(cnn2, 'relu');
    cnn2 = cnnAddPoolLayer(cnn2, 'max', [2, 2]);
    cnn2 = cnnAddReshapeLayer(cnn2);
    cnn2 = cnnAddFCLayer(cnn2, 64, 'r');
    cnn2 = cnnAddActivationLayer(cnn2, 'ReLu');
    cnn2 = cnnAddOutputLayer(cnn2, 'end_BLOB');
    cnn2 = cnnInitVelocity(cnn2);

% Combine
nets = {cnn1, cnn2};
cnn = cnnAddBLOBLayer(cnn, nets, 128, 2);
cnn = cnnAddFCLayer(cnn, 128, 'r');
cnn = cnnAddBNLayer(cnn);
cnn = cnnAddActivationLayer(cnn, 'relu');
cnn = cnnAddFCLayer(cnn, 64, 'r');
cnn = cnnAddBNLayer(cnn);
cnn = cnnAddActivationLayer(cnn, 'relu');
cnn = cnnAddFCLayer(cnn, 2, 'r');
cnn = cnnAddOutputLayer(cnn, 'softmax');
cnn = cnnInitVelocity(cnn);

%% Train CNN
[ERR, cnn] = cnnTrainBP(cnn, trainDataNoCarRe, trainLabelNoCarRe);
figure;
plot(ERR(1, :));
figure;
plot(ERR(2, :));

%% Test CNN
[acc, e] = cnnTestData(cnn, testDataNoCarRe, testLabelNoCarRe, 75);
fprintf('Validation accuracy is: %f\n', acc);
mpiprofile viewer;