-
Notifications
You must be signed in to change notification settings - Fork 10
/
Copy pathcomput_motion_statistics_fast.py
221 lines (150 loc) · 6.46 KB
/
comput_motion_statistics_fast.py
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
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
import numpy as np
bin_size = 8
angle_unit = 360 / bin_size
def cell_gradient(cell_magnitude, cell_angle):
orientation_centers = [0] * bin_size
for k in range(cell_magnitude.shape[0]):
for l in range(cell_magnitude.shape[1]):
gradient_strength = cell_magnitude[k][l]
gradient_angle = cell_angle[k][l]
min_angle = int(gradient_angle / angle_unit) % 8
max_angle = (min_angle + 1) % bin_size
mod = gradient_angle % angle_unit
orientation_centers[min_angle] += (gradient_strength * (1 - (mod / angle_unit)))
orientation_centers[max_angle] += (gradient_strength * (mod / angle_unit))
return orientation_centers
def pattern_1(mag,ang):
max_sum = 0
max_idx = []
for i in range(4):
for j in range(4):
x_start = i * 28
x_end = x_start + 28
y_start = j * 28
y_end = y_start + 28
tmp_block = mag[x_start:x_end, y_start:y_end]
# print(tmp_block.shape)
block_sum = np.sum(tmp_block)
if block_sum > max_sum:
max_sum = block_sum
max_block = 4 * i + j + 1
j = (max_block - 1) % 4
i = int((max_block - 1 - j) / 4)
x_start = i * 28
x_end = x_start + 28
y_start = j * 28
y_end = y_start + 28
max_mag_block = mag[x_start:x_end, y_start:y_end]
max_ang_block = ang[x_start:x_end, y_start:y_end]
orientation_center = cell_gradient(max_mag_block, max_ang_block)
max_idx = np.argmax(orientation_center) + 1
return max_block, max_idx
def pattern_2(mag,ang):
total_sum = []
block1 = mag[42:70, 42:70] # 28 x 28
sum_1 = np.sum(block1)
block1_sum = sum_1 / (28 * 28)
total_sum.append(block1_sum)
block2 = mag[28:84, 28:84] # 56 * 56
sum_2 = np.sum(block2)
block2_sum = (sum_2 - sum_1) / (56 * 56 - 28 * 28)
total_sum.append(block2_sum)
block3 = mag[14:98, 14:98] # 84 * 84
sum_3 = np.sum(block3)
block3_sum = (sum_3 - sum_2) / (84* 84- 56 * 56)
total_sum.append(block3_sum)
block4 = mag[0:112, 0:112] # 112 x 112
sum_4 = np.sum(block4)
block4_sum = (sum_4 - sum_3) / (112 * 112 - 84 * 84)
total_sum.append(block4_sum)
max_idx = total_sum.index(max(total_sum))
if max_idx == 0: # block 1 28 x 28
max_mag_block = mag[42:70, 42:70]
max_ang_block = ang[42:70, 42:70]
elif max_idx == 1: # block 2 56 x 56
tmp_mag = np.zeros_like(block2)
tmp_mag[14:42, 14:42] = mag[42:70, 42:70]
max_mag_block = mag[28:84, 28:84] - tmp_mag
tmp_ang = np.zeros_like(block2)
tmp_ang[14:42, 14:42] = ang[42:70, 42:70]
max_ang_block = ang[28:84, 28:84] - tmp_ang
elif max_idx == 2: # block 3 84 x 84
tmp_mag = np.zeros_like(block3)
tmp_mag[14:70, 14:70] = mag[28:84, 28:84]
max_mag_block = mag[14:98, 14:98] - tmp_mag
tmp_ang = np.zeros_like(block3)
tmp_ang[14:70, 14:70] = ang[28:84, 28:84]
max_ang_block = ang[14:98, 14:98] - tmp_ang
elif max_idx == 3: # block 4 112 x 112
tmp_mag = np.zeros_like(block4)
tmp_mag[14:98, 14:98] = mag[14:98, 14:98]
max_mag_block = mag[0:112, 0:112] - tmp_mag
tmp_ang = np.zeros_like(block4)
tmp_ang[14:98, 14:98] = ang[14:98, 14:98]
max_ang_block = ang[0:112, 0:112] - tmp_ang
orientation_center = cell_gradient(max_mag_block, max_ang_block)
max_ang = np.argmax(orientation_center)
return (max_idx+1), (max_ang+1)
def pattern_3(mag,ang):
mag_block_all = []
ang_block_all = []
block_one = mag[0:56,0:56]
mag_block_1 = block_one[np.tril_indices(56)]
mag_block_2 = block_one[np.triu_indices(56)]
ang_block_one = ang[0:56, 0:56]
ang_block_1 = ang_block_one[np.tril_indices(56)]
ang_block_2 = ang_block_one[np.triu_indices(56)]
############################################################
block_two = mag[0:56, 56:112]
block_two = np.flip(block_two,1)
mag_block_3 = block_two[np.triu_indices(56)]
mag_block_4 = block_two[np.tril_indices(56)]
ang_block_two = ang[0:56, 56:112]
ang_block_two = np.flip(ang_block_two, 1)
ang_block_3 = ang_block_two[np.triu_indices(56)]
ang_block_4 = ang_block_two[np.tril_indices(56)]
################################################
block_three = mag[56:112, 0:56]
block_three = np.flip(block_three,1)
mag_block_5 = block_three[np.triu_indices(56)]
mag_block_6 = block_three[np.tril_indices(56)]
ang_block_three = ang[56:112, 0:56]
ang_block_three = np.flip(ang_block_three, 1)
ang_block_5 = ang_block_three[np.triu_indices(56)]
ang_block_6 = ang_block_three[np.tril_indices(56)]
############################################
block_four = mag[56:112, 56:112]
mag_block_7 = block_four[np.tril_indices(56)]
mag_block_8 = block_four[np.triu_indices(56)]
ang_block_four = ang[56:112, 56:112]
ang_block_7 = ang_block_four[np.tril_indices(56)]
ang_block_8 = ang_block_four[np.triu_indices(56)]
############################################
mag_block_all.append(mag_block_1)
mag_block_all.append(mag_block_2)
mag_block_all.append(mag_block_3)
mag_block_all.append(mag_block_4)
mag_block_all.append(mag_block_5)
mag_block_all.append(mag_block_6)
mag_block_all.append(mag_block_7)
mag_block_all.append(mag_block_8)
mag_block_all = np.array(mag_block_all)
sum_all = np.sum(mag_block_all,1)
ang_block_all.append(ang_block_1)
ang_block_all.append(ang_block_2)
ang_block_all.append(ang_block_3)
ang_block_all.append(ang_block_4)
ang_block_all.append(ang_block_5)
ang_block_all.append(ang_block_6)
ang_block_all.append(ang_block_7)
ang_block_all.append(ang_block_8)
ang_block_all = np.array(ang_block_all)
#############################################
max_mag_idx = np.argmax(sum_all)
max_mag_block = mag_block_all[max_mag_idx]
max_ang_block = ang_block_all[max_mag_idx]
max_mag_block = np.reshape(max_mag_block, [12, 133])
max_ang_block = np.reshape(max_ang_block, [12, 133])
orientation_center = cell_gradient(max_mag_block, max_ang_block)
max_ang_idx = np.argmax(orientation_center)
return max_mag_idx+1, max_ang_idx+1