Chunk / ImageR

Blame view

test/test_model.py 4.78 KB
Edit Raw Normal View History
be12257b   Chunk   data-feat-model f... 
1
2
 
__author__ = 'chunk'
9371f8fa   Chunk   SVM param engenee... 
3
 
from sklearn import cross_validation
f4fb4381   Chunk   staged. 
4
2bf33465   Chunk   staged. 
5
 
from ..common import *
84648488   Chunk   reverted. 
6
 
from ..mdata import CV, ILSVRC, ILSVRC_S
61e78eb3   Chunk   staged. 
7
 
from ..mmodel.svm import SVM
84648488   Chunk   reverted. 
8
 
from ..mmodel.theano import  THEANO
2bf33465   Chunk   staged. 
9
5a469df5   Chunk   staged. 
10
11
12
 
timer = Timer()
84648488   Chunk   reverted. 
13
 
def test_SVM_CV():
5a469df5   Chunk   staged. 
14
15
 
    timer.mark()
    dcv = CV.DataCV()
be12257b   Chunk   data-feat-model f... 
16
 
    X, Y = dcv.load_data(mode='local')  # 90.468586s ->  5.392520s
d2603183   Chunk   staged. 
17
 
    # X, Y = dcv.load_data(mode='hbase') # 21.682754s
84648488   Chunk   reverted. 
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
 
    # X, Y = dcv.load_data(mode='spark') # 29.549597s
    timer.report()

    timer.mark()
    # msvm = SVM.ModelSVM(toolset='sklearn') # 3.030380s
    # msvm = SVM.ModelSVM(toolset='opencv') # 8.939880s
    # msvm = SVM.ModelSVM(toolset='libsvm') # 185.524023s
    msvm = SVM.ModelSVM(toolset='spark')

    msvm.train(X, Y)
    timer.report()

    timer.mark()
    for path, subdirs, files in os.walk('data/467/'):
        for name in files:
            imgpath = os.path.join(path, name)
            feat = dcv.get_feat(imgpath, 'hog')
            print name, msvm.predict(feat)
    timer.report()

    timer.mark()
    print msvm.test(X, Y)  # 0.948892561983 for svm_cv, 0.989024793388 for svm_sk, 0.9900826446280992 for svm_lib
    timer.report()  # 27.421949s for svm_lib


def test_SVM_ILSVRC():
    timer.mark()
    dil = ILSVRC.DataILSVRC(base_dir='/data/hadoop/ImageNet/ILSVRC/ILSVRC2013_DET_val', category='Test.0.2')
    X, Y = dil.load_data(mode='local')  #
    # X, Y = dil.load_data(mode='hbase') #
d0be60e7   Chunk   jpeg update. 
48
49
 
    # X, Y = dil.load_data(mode='spark') #
    X_train, X_test, Y_train, Y_test = cross_validation.train_test_split(X, Y, test_size=0.4, random_state=0)
84648488   Chunk   reverted. 
50
 
    print np.array(Y).shape, np.array(X).shape
d0be60e7   Chunk   jpeg update. 
51
 
    print np.array(X_train).shape, np.array(Y_train).shape
02528074   Chunk   staged. 
52
53
 
    print np.array(X_test).shape, np.array(Y_test).shape
84648488   Chunk   reverted. 
54
 
    timer.report()
6d6d75b8   Chunk   spider LOG system. 
55
02528074   Chunk   staged. 
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
 
    timer.mark()
    msvm = SVM.ModelSVM(toolset='sklearn')  # 4.884247s 0.777853030816
    # msvm = SVM.ModelSVM(toolset='opencv') #
    # msvm = SVM.ModelSVM(toolset='libsvm') #
    # msvm = SVM.ModelSVM(toolset='spark')
    msvm.train(X_train, Y_train)
    timer.report()

    timer.mark()
    print msvm.test(X_test, Y_test)  #
    timer.report()  #

    # timer.mark()
    # print 'or like this:'
    # scores = cross_validation.cross_val_score(msvm.model, X, Y)
    # print scores
    # timer.report()


def test_SVM_ILSVRC_HBASE():
    timer.mark()

    # dil = ILSVRC.DataILSVRC(base_dir='ILSVRC2013_DET_val', category='Train_3')
    # X, Y = dil.load_data(mode='hbase') # pass

    dils = ILSVRC_S.DataILSVRC_S(base='ILSVRC2013_DET_val', category='Test_1')
    X, Y = dils.load_data(mode='hbase')  # pass

    dil = ILSVRC.DataILSVRC(base_dir='/data/hadoop/ImageNet/ILSVRC/ILSVRC2013_DET_val', category='Test_1')
    X1, Y1 = dil.load_data(mode='local')
4f36b116   Chunk   staged. 
86
02528074   Chunk   staged. 
87
 
    X_train, X_test, Y_train, Y_test = cross_validation.train_test_split(X, Y, test_size=0.4, random_state=0)
9371f8fa   Chunk   SVM param engenee... 
88
 
    print Y,np.sum(np.array(Y)==0),np.sum(np.array(Y)==1)
51708346   Chunk   final experiments... 
89
90
 
    print np.array(Y).shape, np.array(X).shape
    print np.array(X_train).shape, np.array(Y_train).shape
2bd3da3e   Chunk   staged. 
91
 
    print np.array(X_test).shape, np.array(Y_test).shape
84648488   Chunk   reverted. 
92
d2603183   Chunk   staged. 
93
 
    timer.report()
02528074   Chunk   staged. 
94
9371f8fa   Chunk   SVM param engenee... 
95
96
97
 
    timer.mark()
    msvm = SVM.ModelSVM(toolset='sklearn')  # 4.884247s 0.777853030816
    # msvm = SVM.ModelSVM(toolset='opencv') #
d0be60e7   Chunk   jpeg update. 
98
99
100
 
    # msvm = SVM.ModelSVM(toolset='libsvm') #
    # msvm = SVM.ModelSVM(toolset='spark')
    msvm.train(X_train, Y_train)
4f36b116   Chunk   staged. 
101
 
    timer.report()
51708346   Chunk   final experiments... 
102
9371f8fa   Chunk   SVM param engenee... 
103
 
    timer.mark()
4f36b116   Chunk   staged. 
104
 
    print msvm.test(X_test, Y_test)  #
9371f8fa   Chunk   SVM param engenee... 
105
 
    timer.report()  #
d0be60e7   Chunk   jpeg update. 
106
107
108
 

    timer.mark()
    print msvm.test(X1, Y1)  #
9371f8fa   Chunk   SVM param engenee... 
109
 
    timer.report()  #
d0be60e7   Chunk   jpeg update. 
110
 
    # timer.mark()
6d6d75b8   Chunk   spider LOG system. 
111
 
    # print 'or like this:'
51708346   Chunk   final experiments... 
112
113
114
115
 
    # scores = cross_validation.cross_val_score(msvm.model, X, Y)
    # print scores
    # timer.report()
e3e7e73a   Chunk   spider standalone... 
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
 

def test_SVM_ILSVRC_SPARK():
    timer.mark()
    dils = ILSVRC_S.DataILSVRC_S(base='ILSVRC2013_DET_val', category='Test_1')
    rdd_dataset = dils.load_data(mode='spark')  # pass

    timer.report()

    timer.mark()
    # msvm = SVM.ModelSVM(toolset='sklearn')  #
    # msvm = SVM.ModelSVM(toolset='opencv') #
    # msvm = SVM.ModelSVM(toolset='libsvm') #
    msvm = SVM.ModelSVM(toolset='spark', sc=dils.sparker)
    msvm.train(rdd_dataset)
    timer.report()

    dataset = rdd_dataset.collect()
    length = len(dataset)
2bd3da3e   Chunk   staged. 
135
 
    X_test, Y_test = [dataset[i].features for i in range(length)], [dataset[i].label for i in range(length)]
02528074   Chunk   staged. 
136
137
138
139
140
141
142
143
144
 

    timer.mark()
    print msvm.test(dils.sparker.sc.parallelize(X_test), Y_test)  #
    timer.report()  #


def test_SVM_ILSVRC_S():
    test_SVM_ILSVRC_HBASE()
    # test_SVM_ILSVRC_SPARK()
f4fb4381   Chunk   staged. 
145
146
147
148
 



def test_THEANO_crop():
02528074   Chunk   staged. 
149
150
 

    timer.mark()
9371f8fa   Chunk   SVM param engenee... 
151
 
    dilc = ILSVRC.DataILSVRC(base_dir='/data/hadoop/ImageNet/ILSVRC/ILSVRC2013_DET_val', category='Test_crop_pil')
02528074   Chunk   staged. 
152
153
154
155
156
 
    X, Y = dilc.load_data(mode='local', feattype='coef')
    timer.report()

    timer.mark()
    mtheano = THEANO.ModelTHEANO(toolset='cnn')
9371f8fa   Chunk   SVM param engenee... 
157
158
 
    mtheano.train(X,Y)
    timer.report()
02528074   Chunk   staged. 
159
160
161
 


if __name__ == '__main__':
84648488   Chunk   reverted. 
162
 
    # test_SVM_CV()
02528074   Chunk   staged. 
163
164
 
    test_SVM_ILSVRC()
    print 'helllo'
51708346   Chunk   final experiments... 












02528074
 
 Chunk
 
staged.
















9371f8fa
 
 Chunk
 
SVM param engenee...
















84648488
 
 Chunk
 
reverted.
















be12257b
 
 Chunk
 
data-feat-model f...
















d0be60e7
 
 Chunk
 
jpeg update.
















be12257b
 
 Chunk
 
data-feat-model f...
















84648488
 
 Chunk
 
reverted.