-
Notifications
You must be signed in to change notification settings - Fork 0
/
svm_origin.py
131 lines (110 loc) · 5.35 KB
/
svm_origin.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
##
## Created by Liu Deyuan on 2021/01/15.
##
import cv2
import time
import numpy as np
from sklearn import svm
from sklearn.metrics import accuracy_score, f1_score
from hog_origin import feature_extraction
def enhance_data(X):
for i in range(X.shape[0]):
image = np.array(X[i], np.float32)
r = np.random.randint(0, 4)
rotate_flags = [-1, cv2.ROTATE_90_CLOCKWISE, cv2.ROTATE_180, cv2.ROTATE_90_COUNTERCLOCKWISE]
flag = rotate_flags[r]
if flag != -1:
image = cv2.rotate(image, flag)
image = cv2.resize(image, dsize=(400, 300), interpolation=cv2.INTER_LINEAR)
X[i] = np.array(image, np.float16)
return X
def load_data():
k1 = ['dataset/data_float16_v1.npz', 'dataset/data_float16_v3.npz', 'dataset/data_float16_v4.npz']
oppo = ['dataset/data_float16_v5.npz', 'dataset/data_float16_v6.npz', 'dataset/data_float16_v7.npz']
files = []
files.extend(oppo)
X_train_all, X_test_all, y_train_all, y_test_all = [], [], [], []
for file in files:
data = np.load(file)
X_train, X_test, y_train, y_test = data['X_train'], data['X_test'], data['y_train'], data['y_test']
X_train_all.append(X_train)
X_test_all.append(X_test)
y_train_all.append(y_train)
y_test_all.append(y_test)
X_train_all = np.concatenate(X_train_all)
X_test_all = np.concatenate(X_test_all)
y_train_all = np.concatenate(y_train_all)
y_test_all = np.concatenate(y_test_all)
X_train_all = enhance_data(X_train_all)
X_test_all = enhance_data(X_test_all)
return X_train_all, X_test_all, y_train_all, y_test_all
def analysis_result(label, predict, time):
print(f'{time:.6f}s for {label.shape[0]} samples predict')
print(f'{np.sum(label > 0)} positive classes')
print(f'accuracy_score: {accuracy_score(label, predict)}')
print(f'accuracy_number: {int(accuracy_score(label, predict, normalize=False))}/{len(label)}')
print(f'f1_score: {f1_score(label, predict, average=None)}')
print()
if __name__ == "__main__":
# (1)读取数据
X_train, X_test, y_train, y_test = load_data()
# (2)提取数据特征
print('Begin Feature Extraction')
start_time = time.time()
X_train_feature = feature_extraction(X_train)
end_time = time.time()
print(f'{end_time - start_time:.6f}s for {y_train.shape[0]} samples feature extraction')
print()
start_time = time.time()
X_test_feature = feature_extraction(X_test)
end_time = time.time()
print(f'{end_time - start_time:.6f}s for {y_test.shape[0]} samples feature extraction')
print()
# (3)创建SVM模型
print('Begin SVM')
clf = svm.SVC() # 默认kernel为rbf，测试代码仅支持kernel为rbf
# (4)训练SVM模型
clf.fit(X_train_feature, y_train)
# (5)保存SVM模型参数
LIBSVM_IMPL = ["c_svc", "nu_svc", "one_class", "epsilon_svr", "nu_svr"]
print('X_test_feature', X_test_feature.shape)
print('self.support_', clf.support_.shape, clf.support_)
print('self.support_vectors_', clf.support_vectors_.shape, np.array(clf.support_vectors_, dtype=np.float16))
print('self._n_support', clf._n_support)
print('self._dual_coef_', clf._dual_coef_.shape, np.array(clf._dual_coef_, dtype=np.float16))
print('self._intercept_', np.array(clf._intercept_, dtype=np.float16))
print('self._probA', clf._probA)
print('self._probB', clf._probB)
print('svm_type', LIBSVM_IMPL.index(clf._impl))
print('kernel', clf.kernel)
print('self.degree', clf.degree)
print('self.coef0', np.float16(clf.coef0))
print('self.gamma', np.float16(clf._gamma))
print('self.cache_size', clf.cache_size)
print()
np.savez('model/params.npz', support = clf.support_, SV = np.array(clf.support_vectors_, dtype=np.float16), nSV = clf._n_support,
sv_coef = np.array(clf._dual_coef_, dtype=np.float16), intercept = np.array(clf._intercept_, dtype=np.float16), svm_type = LIBSVM_IMPL.index(clf._impl),
kernel = clf.kernel, degree = clf.degree, gamma = np.float16(clf._gamma), coef0 = np.float16(clf.coef0))
print(clf.break_ties, clf.decision_function_shape, len(clf.classes_), clf._sparse, callable(clf.kernel))
print()
save_path = 'model/bin/'
clf.support_.tofile(save_path+'01_support_int32.bin')
clf.support_vectors_.astype(np.float32).tofile(save_path+'02_SV_float32.bin')
clf._n_support.tofile(save_path+'03_nSV_int32.bin')
clf._dual_coef_.astype(np.float32).tofile(save_path+'04_sv_coef_float32.bin')
clf._intercept_.astype(np.float32).tofile(save_path+'05_intercept_float32.bin')
np.array(LIBSVM_IMPL.index(clf._impl)).astype(np.float32).tofile(save_path+'06_svm_type_float32.bin')
np.array(clf.kernel).tofile(save_path+'07_kernel_int32_string.bin')
np.array(clf.degree).tofile(save_path+'08_degree_int32.bin')
np.array(clf._gamma).astype(np.float32).tofile(save_path+'09_gamma_float32.bin')
np.array(clf.coef0).astype(np.float32).tofile(save_path+'10_coef0_float32.bin')
# (6)测试模型精度
print(' -------- 原版的SVM模型测试 ---------- ')
start_time = time.time()
result = clf.predict(X_train_feature)
end_time = time.time()
analysis_result(y_train, result, end_time-start_time)
start_time = time.time()
result = clf.predict(X_test_feature)
end_time = time.time()
analysis_result(y_test, result, end_time-start_time)