-
Notifications
You must be signed in to change notification settings - Fork 3
/
CIFAR10PlugAndPlayCollaboration.py
152 lines (125 loc) · 5.96 KB
/
CIFAR10PlugAndPlayCollaboration.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
import numpy as np
from keras.datasets import cifar10
import cv2
from tmu.models.classification.vanilla_classifier import TMClassifier
from tmu.preprocessing.standard_binarizer.binarizer import StandardBinarizer
device = "GPU"
max_included_literals = 32
resolution = 8
factor = 1
(X_train_org, Y_train), (X_test_org, Y_test) = cifar10.load_data()
Y_train=Y_train.reshape(Y_train.shape[0])
Y_test=Y_test.reshape(Y_test.shape[0])
##################################
##### Histogram of Gradients #####
##################################
imageSize = 32 #The size of the original image - in pixels - assuming this is a square image
channels = 3 #The number of channels of the image. A RBG color image, has 3 channels
classes = 10 #The number of classes available for this dataset
winSize = imageSize
blockSize = 12
blockStride = 4
cellSize = 4
nbins = 18
derivAperture = 1
winSigma = -1.
histogramNormType = 0
L2HysThreshold = 0.2
gammaCorrection = True
nlevels = 64
signedGradient = True
hog = cv2.HOGDescriptor((winSize,winSize),(blockSize, blockSize),(blockStride,blockStride),(cellSize,cellSize),nbins,derivAperture, winSigma,histogramNormType,L2HysThreshold,gammaCorrection,nlevels,signedGradient)
fd = hog.compute(X_train_org[0])
X_train_hog = np.empty((X_train_org.shape[0], fd.shape[0]), dtype=np.uint32)
for i in range(X_train_org.shape[0]):
fd = hog.compute(X_train_org[i])
X_train_hog[i] = fd >= 0.1
fd = hog.compute(X_test_org[0])
X_test_hog = np.empty((X_test_org.shape[0], fd.shape[0]), dtype=np.uint32)
for i in range(X_test_org.shape[0]):
fd = hog.compute(X_test_org[i])
X_test_hog[i] = fd >= 0.1
tm_hog = TMClassifier(
number_of_clauses=2000*factor,
T=50*factor,
s=10.0,
max_included_literals=max_included_literals,
platform=device,
weighted_clauses=False
)
#################################
##### Adaptive Thresholding #####
#################################
X_train_threshold = np.copy(X_train_org)
X_test_threshold = np.copy(X_test_org)
for i in range(X_train_threshold.shape[0]):
for j in range(X_train_threshold.shape[3]):
X_train_threshold[i,:,:,j] = cv2.adaptiveThreshold(X_train_org[i,:,:,j], 1, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2) #cv2.adaptiveThreshold(X_train[i,:,:,j], 1, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 5, 5)
for i in range(X_test_threshold.shape[0]):
for j in range(X_test_threshold.shape[3]):
X_test_threshold[i,:,:,j] = cv2.adaptiveThreshold(X_test_org[i,:,:,j], 1, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2)#cv2.adaptiveThreshold(X_test[i,:,:,j], 1, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 5, 5)
tm_threshold = TMClassifier(
number_of_clauses=2000*factor,
T=500*factor,
s=10.0,
max_included_literals=max_included_literals,
platform=device,
weighted_clauses=True,
patch_dim=(10, 10)
)
##############################
##### Color Thermometers #####
##############################
X_train_thermometer = np.empty((X_train_org.shape[0], X_train_org.shape[1], X_train_org.shape[2], X_train_org.shape[3], resolution), dtype=np.uint8)
for z in range(resolution):
X_train_thermometer[:,:,:,:,z] = X_train_org[:,:,:,:] >= (z+1)*255/(resolution+1)
X_test_thermometer = np.empty((X_test_org.shape[0], X_test_org.shape[1], X_test_org.shape[2], X_test_org.shape[3], resolution), dtype=np.uint8)
for z in range(resolution):
X_test_thermometer[:,:,:,:,z] = X_test_org[:,:,:,:] >= (z+1)*255/(resolution+1)
X_train_thermometer = X_train_thermometer.reshape((X_train_org.shape[0], X_train_org.shape[1], X_train_org.shape[2], 3*resolution))
X_test_thermometer = X_test_thermometer.reshape((X_test_org.shape[0], X_test_org.shape[1], X_test_org.shape[2], 3*resolution))
tm_thermometer_3 = TMClassifier(
number_of_clauses=2000*factor,
T=1500*factor,
s=2.5,
max_included_literals=max_included_literals,
platform=device,
weighted_clauses=True,
patch_dim=(3, 3),
)
tm_thermometer_4 = TMClassifier(
number_of_clauses=2000*factor,
T=1500*factor,
s=2.5,
max_included_literals=max_included_literals,
platform=device,
weighted_clauses=True,
patch_dim=(4, 4),
)
############################
##### Training of Team #####
############################
for epoch in range(100):
print("#%d" % (epoch+1), end=' ')
tm_hog.fit(X_train_hog, Y_train)
Y_test_hog, Y_test_scores_hog = tm_hog.predict(X_test_hog, return_class_sums=True)
print("HoG: %.1f%%" % (100*(Y_test_hog == Y_test).mean()), end=' ')
tm_threshold.fit(X_train_threshold, Y_train)
Y_test_threshold, Y_test_scores_threshold = tm_threshold.predict(X_test_threshold, return_class_sums=True)
print("Adaptive Thresholding: %.1f%%" % (100*(Y_test_threshold == Y_test).mean()), end=' ')
tm_thermometer_3.fit(X_train_thermometer, Y_train)
Y_test_thermometer_3, Y_test_scores_thermometer_3 = tm_thermometer_3.predict(X_test_thermometer, return_class_sums=True)
print("3x3 Color Thermometers: %.1f%%" % (100*(Y_test_thermometer_3 == Y_test).mean()), end=' ')
tm_thermometer_4.fit(X_train_thermometer, Y_train)
Y_test_thermometer_4, Y_test_scores_thermometer_4 = tm_thermometer_4.predict(X_test_thermometer, return_class_sums=True)
print("4x4 Color Thermometers: %.1f%%" % (100*(Y_test_thermometer_4 == Y_test).mean()), end=' ')
##### Team Decision #####
votes = np.zeros(Y_test_scores_hog.shape, dtype=np.float32)
for i in range(Y_test.shape[0]):
votes[i] += 1.0*Y_test_scores_threshold[i]/(np.max(Y_test_scores_threshold) - np.min(Y_test_scores_threshold))
votes[i] += 1.0*Y_test_scores_thermometer_3[i]/(np.max(Y_test_scores_thermometer_3) - np.min(Y_test_scores_thermometer_3))
votes[i] += 1.0*Y_test_scores_thermometer_4[i]/(np.max(Y_test_scores_thermometer_4) - np.min(Y_test_scores_thermometer_4))
votes[i] += 1.0*Y_test_scores_hog[i]/(np.max(Y_test_scores_hog) - np.min(Y_test_scores_hog))
Y_test_team = votes.argmax(axis=1)
print("Team: %.1f%%" % (100*(Y_test_team == Y_test).mean()))
print()