-
Notifications
You must be signed in to change notification settings - Fork 16
/
Run_DA.m
194 lines (182 loc) · 10.2 KB
/
Run_DA.m
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
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
function Run_DA()
% -> List of parameters
listDatasets = {'Office','Saenko','Testbed','CrossDataset','Sentiment'};
numSrcClusters = [31, 10, 40, -1, 2];
listFeatures = {'CNN-fc7', 'BoW'};
listClasses = { ...
{'AMAZON', 'DSLR', 'WEBCAM'}, ...
{'AMAZON', 'DSLR', 'WEBCAM', 'CALTECH'}, ...
{'CALTECH256', 'IMAGENET','SUN', 'BING'}, ...
{'caltech101', 'office', 'pascal07'}, ... % {'CALTECH256', 'IMAGENET','SUN', 'BING'}, ... % {'caltech101', 'imagenet', 'msrcorid', 'bing', 'pascal07', 'eth80', 'office'}, ...
{'books', 'dvd', 'elec', 'kitchen'}};
isClassSupervised = [false, true];
typeDA = {'', 'ATI'}; % ,'gfk','SA','CORAL','TCA'};
isDA = [false, true]; % , true, true, true, true, true];
typeCaffe = {'AlexNet', 'VGG'};
% SetUp of DA parameters
% standard, NN-1, NN-2]
da.numCorr = [99, 99, 99];
da.numLambda = [0.5, 0.5, 0.5];
da.delta = [0.66, 0.66, 0.66]; % Progression update [0..1]
da.numNN = [0, 1, 2];
da.PCA = [256, 512, 1000];
nameRun = 'test\';
for sup = 1 % 1:length(isClassSupervised)
for d = 1 % 1:length(listDatasets)
classes = listClasses{d};
for feat = 1 % 1:length(listFeatures)
for m = 1:length(typeDA)
if(m == 1)
numTechniques = 1;
else % DA
numTechniques = 3;
end
for src = 1:length(classes)
for tgt = 1:length(classes)
for it = 1 % 1:5
for t = 1 % 1:numTechniques % da.X iters
for c = 1 % :length(typeCaffe);
if(src == tgt)
continue;
end
% PARAMETERS
input = InputParameters;
% Change for closed (false) or open set DA (true)
input.isOpenset = false;
input.isWSVM = false;
input.isSaveTSNE = false;
% -> Dataset specific cases
% if(strcmpi(listDatasets{d},'CrossDataset'))
% % if(protocolSparse(classes{src}, classes{tgt}))
% continue;
% end
% end
if(strcmpi(listDatasets{d},'Sentiment'))
if(~(strcmpi(classes{src},'kitchen') && strcmpi(classes{tgt},'dvd') || ...
strcmpi(classes{src},'dvd') && strcmpi(classes{tgt},'books') || ...
strcmpi(classes{src},'books') && strcmpi(classes{tgt},'elec') || ...
strcmpi(classes{src},'elec') && strcmpi(classes{tgt},'kitchen') ))
continue;
end
end
% -> Classifier
input.PATH_DATA = 'Z:\PhD\Data\';
if(input.isWSVM)
input.methodSVM = 'libsvm-open';
input.multiClassApproach = 'OVA';
else
input.methodSVM = 'libsvm';
input.multiClassApproach = 'OVO';
end
input.C_LSVM = 0.001;
input.CV_LSVM = false;
% -> Save results
input.PATH_RESULTS = ['Z:\Results\' nameRun '\' num2str(it) '\'];
% -> Datasets
input.sourceDataset = eval(listDatasets{d});
input.sourceDataset.source = classes{src};
input.targetDataset = eval(listDatasets{d});
input.targetDataset.target = classes{tgt};
% -> Protocol
input.daAllSrc = true;
input.daAllTgt = true;
% -> Features
input.typeDescriptor = listFeatures{feat};
input.cnnName = typeCaffe{c};
input.cnnModel = typeCaffe{c};
input.isZScore = true;
% -> Domain Adaptation
input.isDA = isDA(m);
input.typeDA = typeDA{m};
input.numTgtClusters = 99999;
if(input.isOpenset)
input.numSrcClusters = 11; % numSrcClusters(d);
elseif(input.isWSVM)
input.numSrcClusters = 10;
else
input.numSrcClusters = numSrcClusters(d);
end
input.iterDA = 4;
input.numIterOpt = 50;
if(m > 2)
input.dimPCA = da.PCA(t);
else
input.dimPCA = 0.3; % Reduce a third to not lose performance and speed-up
end
input.tol_residual = 0.01; % [0.001 - 0.01]
input.deltaW = da.delta(t);
input.isClosestNN = false;
input.numNN = da.numNN(t);
input.numCorr = da.numCorr(t);
input.numLambda = da.numLambda(t);
input.daSpecial = '';
% -> Open Set
input.includeBgClass = true;
input.isWild = false;
input.typeWildSupervision = 'Office';
% -> Supervision
input.isClassSupervised = isClassSupervised(sup);
input.daOnlySupervised = false;
input.daNumSupervised = 3;
% -> RUN!
input.isMidResultsDA = true;
input.seedRand = it;
if(it == 6)
input.seedRand = 1;
end
main(input);
end
end
end
end
end
end
end
end
end
end
function isContinue = protocolSparse(src, tgt)
isContinue = true;
if(strcmpi(src,'caltech256') && strcmpi(tgt,'imagenet') || strcmpi(src,'caltech256') && strcmpi(tgt,'sun') || ...
strcmpi(src,'imagenet') && strcmpi(tgt,'caltech256') || strcmpi(src,'imagenet') && strcmpi(tgt,'sun') || ...
strcmpi(src,'sun') && strcmpi(tgt,'caltech256') || strcmpi(src,'sun') && strcmpi(tgt,'imagenet') || ...
strcmpi(src,'eth80') && strcmpi(tgt,'caltech256') || strcmpi(src,'eth80') && strcmpi(tgt,'pascal07') || ...
strcmpi(src,'office') && strcmpi(tgt,'caltech256') || strcmpi(src,'office') && strcmpi(tgt,'pascal07') || ...
strcmpi(src,'bing') && strcmpi(tgt,'caltech256') || strcmpi(src,'bing') && strcmpi(tgt,'pascal07'))
isContinue = false;
end
end
function isContinue = protocolSparse2(src, tgt)
isContinue = true;
if(strcmpi(src,'caltech101') && strcmpi(tgt,'imagenet') || strcmpi(src,'caltech101') && strcmpi(tgt,'sun') || ...
strcmpi(src,'imagenet') && strcmpi(tgt,'caltech101') || strcmpi(src,'imagenet') && strcmpi(tgt,'sun') || ...
strcmpi(src,'sun') && strcmpi(tgt,'caltech101') || strcmpi(src,'sun') && strcmpi(tgt,'imagenet') || ...
strcmpi(src,'eth80') && strcmpi(tgt,'caltech101') || strcmpi(src,'eth80') && strcmpi(tgt,'pascal07') || ...
strcmpi(src,'office') && strcmpi(tgt,'caltech101') || strcmpi(src,'office') && strcmpi(tgt,'pascal07') || ...
strcmpi(src,'pascal07') && strcmpi(tgt,'caltech101') || strcmpi(src,'caltech101') && strcmpi(tgt,'pascal07'))
isContinue = false;
end
end
function str = getOfficeData(src, tgt)
str = '';
switch(src)
case 1
str = 'A'; % Amazon
case 2
str = 'D'; % DSLR
case 3
str = 'W'; % Webcam
case 4
str = 'C'; % Caltech
end
switch(tgt)
case 1
str = [str 'A']; % Amazon
case 2
str = [str 'D']; % DSLR
case 3
str = [str 'W']; % Webcam
case 4
str = [str 'C']; % Caltech
end
end