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
|
function trainCascadeObjectDetector(outputFolder,positiveInstances,negativeImages,varargin)
// It creates trained cascade XML file
//
// Calling Sequence
// trainCascadeObjectDetector(outputFolder,positiveInstances,"negativeImages")
//
// Parameters
// outputFolder: Folder name to store trained cascade (cascade.xml) and intermediate files
// positiveInstances: Array of structure with image file names (including path) and an M-by-4 matrix of bounding box.
// negativeImages: path to a negative images folder
// numPos: number of positive samples. Default-30
// numNeg: number of negative samples. Default- 20
// nsplits: number of splits. Default- 2
// numStages: number of cascade stages to be trained. Default- 30
// featureType: Type of features to be used, possible types are HAAR, LBP, HOG. Default-HAAR
// minHitRate: Minimal desired hit rate for each stage of the classifier and value in the range 0 and 1 inclusive. Default- 0.995
// maxFalseAlarmRate: Maximal desired false alarm rate for each stage of the classifier and value in the range 0 and 1 inclusive. Default- 0.5
// w: Width of training samples (in pixels). Default-25
// h: Height of training samples (in pixels). Default-25
//
// Description
// It produces dataset of positive samples in a file with .vec extension and negative samples are enumerated in a special text file in
// which each line contains an image filename of negative sample image. Negative images must not contain detected objects.
//
// By using these two files trainCascade will create cascade.xml file inside the outputFolder which is used to detect objects in an image.
//
// Examples
// positiveInstances(1)=struct("path",'image1.jpg',"bbox",[1 5 2 3]);
// positiveInstances(2)=struct("path",'image2.jpg',"bbox",[2 7 4 2]);
// trainCascadeObjectDetector("data",positiveInstances,"negativeImagesFolder");
// trainCascadeObjectDetector("data",positiveInstances,"negativeImagesFolder","numStages",35,"featureType","HOG");
// trainCascadeObjectDetector("data",positiveInstances,"negativeImagesFolder","minHitRate",0.9,"numPos",8,"numNeg",5);
[lhs rhs]=argn(0);
if rhs<3 then
error(msprintf(" Not enough input arguments"))
elseif rhs>21 then
error(msprintf(" Too many input arguments to the function"))
elseif modulo(rhs-3,2)
error(msprintf(" wrong number of input arguments,name-value pairs not macthed"))
end
//validating variables
if ~isdir(outputFolder) then
error(msprintf(" wrong input argument #1, existing directory expected"))
elseif ~isdir(negativeImages)
error(msprintf(" wrong input argument #3,existing directory expected"))
end
//default values
numPos=30;
numNeg=20;
numStages=30;
nsplits=2;
featureType="HAAR"
minHitRate=0.995
maxFalseAlarmRate=0.5
w=25
h=25
for i=1:2:rhs-3
if strcmpi(varargin(i),"numPos")==0 then
i=i+1;
numPos=varargin(i);
if numPos<0 then
error(msprintf(" numPos value must be positive"))
end
elseif strcmpi(varargin(i),'numNeg')==0 then
i=i+1;
numNeg=varargin(i);
if numNeg<0 then
error(msprintf(" numNeg value must be positive"))
end
elseif strcmpi(varargin(i),'numStages')==0 then
i=i+1;
numStages=varargin(i);
if numStages<0 then
error(msprintf(" numStages value must be positive"))
end
elseif strcmpi(varargin(i),'nslits')==0 then
i=i+1;
nsplits=varargin(i);
if nsplits<0 then
error(msprintf(" nsplits value must be positive"))
end
elseif strcmpi(varargin(i),'featureType')==0 then
i=i+1;
featureType=varargin(i);
if strcmpi(featureType,'haar') & strcmpi(featureType,'lbp') & strcmpi(featureType,'hog')
error(msprintf(" wrong input argument #%d,featureType not matched",i));
end
elseif strcmpi(varargin(i),'minHitRate')==0 then
i=i+1;
minHitRate=varargin(i);
if minHitRate<0 | minHitRate>1 then
error(msprintf(" minHitRate value must lie in between 0 and 1"))
end
elseif strcmpi(varargin(i),'maxFalseAlarmRate')==0 then
i=i+1;
maxFalseAlarmRate=varargin(i);
if maxFalseAlarmRate<0 | minFalseRate>1 then
error(msprintf(" maxFalseAlarmRate value must lie in between 0 and 1"))
end
elseif strcmpi(varargin(i),'w')==0 then
i=i+1;
w=varargin(i);
if h<0 then
error(msprintf(" w value must be positive"))
end
elseif strcmpi(varargin(i),'h')==0 then
i=i+1;
h=varargin(i);
if h<0 then
error(msprintf(" h value must be positive"))
end
else
error(msprintf(_(" Wrong value for input argument #%d",i)));
end
end
[nRows noOfPositiveInstances]=size(positiveInstances);
fields=fieldnames(positiveInstances);
fd = mopen('positive.txt','wt');
for i=1:noOfPositiveInstances
mfprintf(fd,'%s 1',getfield(fields(1),positiveInstances(i)));
boxVals=getfield(fields(2),positiveInstances(i));
for j=1:4
mfprintf(fd,' %d',boxVals(j));
end
mfprintf(fd,'\n');
end
mclose(fd);
disp("Creating positive samples:");
cmd=sprintf("opencv_createsamples -info positive.txt -num%d -vec positive.vec -w %d -h %d",numPos,w,h);
unix_w(cmd);
if isdir(negativeImages)
if getos()=="Linux"
temp=strcat(["ls ",negativeImages])
elseif getos()=="Windows"
temp=strcat(["dir ",negativeImages])
end
s=unix_g(temp);
[noOfFilesInFolder noOfCols]=size(s);
fd = mopen('negative.txt','wt');
for i=1:noOfFilesInFolder
[path,fname,extension]=fileparts(s(i))
if ~strcmp(extension,".jpg") | ~strcmp(extension,".jpeg") | ~strcmp(extension,".png") | ~strcmp(extension,".bmp") | ~strcmp(extension,".pgm") | ~strcmp(extension,".JPG") | ~strcmp(extension,".JPEG") | ~strcmp(extension,".PNG") | ~strcmp(extension,".BMP") | ~strcmp(extension,".PGM")
mfprintf(fd,'%s/%s\n',negativeImages,s(i));
end
end
end
disp("Training Cascade:");
cmd=sprintf("opencv_traincascade -data %s -vec positive.vec -bg negative.txt -numPos %d -numNeg %d -numStages %d -nsplits %d -featureType %s -minHitRate %d -maxFalseAlarmRate %d -w %d -h %d",outputFolder,numPos,numNeg,numStages,nsplits,featureType,minHitRate,maxFalseAlarmRate,w,h);
unix_w(cmd);
endfunction;
|
