transform_all_halfsize.m
6.15 KB
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
clear;
path_name = '/media/rfj/EEA4441FA443E923/nturgb_skeletones/';
fileID = fopen('/home/rfj/바탕화면/actionGAN/skeletone_INDEX/good_stand_2.txt','r');
formatSpec = '%s';
sizeA = [20 Inf];
perfect_list = fscanf(fileID,formatSpec,sizeA);
perfect_list = perfect_list.';
fclose(fileID);
L = length(perfect_list);
for K = 1:L
file_name = char(perfect_list(K,:));
name = strcat(path_name,file_name(1:20),'.skeleton');
[token,remainder] = strtok(file_name,'A');
class = str2num(remainder(2:4));
bodyinfo = read_skeleton_file(name);
frame_num = size(bodyinfo,2);
try
%initialize
cur_subject_x = zeros(frame_num, 25);
cur_subject_y = zeros(frame_num, 25);
cur_subject_z = zeros(frame_num, 25);
tot_x = zeros(frame_num,25);
tot_y = zeros(frame_num,25);
tot_z = zeros(frame_num,25);
joint_5 = zeros(1,3);
joint_9 = zeros(1,3);
joint_1 = zeros(1,3);
joint_3 = zeros(1,3);
%get total joints information
for FN = 1:frame_num
cur_body = bodyinfo(FN).bodies(1);
joints = cur_body.joints;
for JN = 1:25
tot_x(FN,JN) = joints(JN).x;
tot_y(FN,JN) = joints(JN).y;
tot_z(FN,JN) = joints(JN).z;
end
end
%Orientation normalization 1 : in space
%get median values
M_x = median(tot_x);
M_y = median(tot_y);
M_z = median(tot_z);
%set 3 points for make plane
joint_5 = [M_x(5) M_y(5) M_z(5)];
joint_9 = [M_x(9) M_y(9) M_z(9)];
joint_1 = [M_x(1) M_y(1) M_z(1)];
joint_3 = [M_x(3) M_y(3) M_z(3)];
%find RIGID TRNASFORMATION matrix
d1 = joint_1 - joint_5;
d2 = joint_1 - joint_9;
n1 = cross(d1,d2); % because we will parallel transform, don't need to find belly
u1 = n1/norm(n1);
u2 = [0 0 1];
cs1 = dot(u1,u2)/norm(u1)*norm(u2);
ss1 = sqrt(1-cs1.^2);
v1 = cross(u1,u2)/norm(cross(u1,u2));
R1 = [v1(1)*v1(1)*(1-cs1)+cs1 v1(1)*v1(2)*(1-cs1)-v1(3)*ss1 v1(1)*v1(3)*(1-cs1)+v1(2)*ss1];
R1(2,:) = [v1(1)*v1(2)*(1-cs1)+v1(3)*ss1 v1(2)*v1(2)*(1-cs1)+cs1 v1(2)*v1(3)*(1-cs1)-v1(1)*ss1];
R1(3,:) = [v1(1)*v1(3)*(1-cs1)-v1(2)*ss1 v1(2)*v1(3)*(1-cs1)+v1(1)*ss1 v1(3)*v1(3)*(1-cs1)+cs1];
%1-3 number tolls to parallel x axis. Rigid transformation on plane surface
%Z axis coords oyler angle transform
t = joint_3 - joint_1;
d3 = R1(1,:) * t.';
d3(1,2) = R1(2,:) * t.';
d3(1,3) = R1(3,:) * t.';
u3 = d3(1:2)/norm(d3(1:2));
v3 = [u3(1) -u3(2)];
v3(2,:) = [u3(2) u3(1)];
u4 = [1 0].';
csss = v3\u4;
cs2 = csss(1);
ss2 = csss(2);
R2 = [cs2 -ss2 0];
R2(2,:) = [ss2 cs2 0];
R2(3,:) = [0 0 1];
%apply rigid transformation
for FN = 1:frame_num
cur_body = bodyinfo(FN).bodies(1);
joints = cur_body.joints;
for JN = 1:25
a = R1(1,:) * [joints(JN).x joints(JN).y joints(JN).z].';
b = R1(2,:) * [joints(JN).x joints(JN).y joints(JN).z].';
c = R1(3,:) * [joints(JN).x joints(JN).y joints(JN).z].';
cur_subject_x(FN,JN) = R2(1,:) * [a b c].';
cur_subject_y(FN,JN) = R2(2,:) * [a b c].';
cur_subject_z(FN,JN) = R2(3,:) * [a b c].';
end
end
%orientation normalize 2 in plane surface
if cur_subject_x(1,4) < cur_subject_x(1,1)
cur_subject_x = 0 - cur_subject_x;
end
if cur_subject_y(1,9) > cur_subject_y(1,5)
cur_subject_y = 0 - cur_subject_y;
end
% for save origin subjects before data augment
clear_subject_x = cur_subject_x;
clear_subject_y = cur_subject_y;
clear_subject_z = cur_subject_z;
% Left <-> Right Change : 2option
for LR = 1:2
if LR == 1
augment_y = clear_subject_y;
else
augment_y = 0 - clear_subject_y;
end
%Height change : 3option
for HE = 1:3
if HE == 1
augment_x = clear_subject_x.* 1.2;
elseif HE==2
augment_x = clear_subject_x.* 1.0;
else
augment_x = clear_subject_x.* 0.8;
end
%Give Gaussian Random Variable : 0.01 - 6times
for RV = 1:6
%3. Gaussian Random filter 0.1
cur_subject_x = augment_x + 0.01.*randn(frame_num,25);
cur_subject_y = augment_y + 0.01.*randn(frame_num,25);
cur_subject_z = clear_subject_z + 0.01.*randn(frame_num,25);
% NORMALIZATION
cur_subject_x = cur_subject_x - min(cur_subject_x(:));
max_tall = max(cur_subject_x(:)) .*2;
cur_subject_x = cur_subject_x ./ max_tall;
cur_subject_y = cur_subject_y - min(cur_subject_y(:));
cur_subject_y = cur_subject_y ./ max_tall;
cur_subject_z = cur_subject_z - min(cur_subject_z(:));
cur_subject_z = cur_subject_z ./ max_tall;
%Write image
motionpatch = cur_subject_x;
motionpatch(:,:,2) = cur_subject_y;
motionpatch(:,:,3) = cur_subject_z;
new_file_name = strcat('/home/rfj/바탕화면/actionGAN/DCGAN/new_motionpatch_halfsize/',file_name(1:20),'_',num2str(LR),num2str(HE),num2str(RV),'.png');
imwrite(motionpatch,new_file_name);
end
end
end
catch
name
end
end