/
cross_validate.py
175 lines (153 loc) · 7.03 KB
/
cross_validate.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
#!/usr/bin/env python
"""Script to run Cross-Validation"""
import logging
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.INFO)
import argparse
from time import time
import scipy
import sklearn
from sklearn.externals import joblib
from sklearn import cross_validation
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
import numpy as np
import feature_extraction
import machine_learning
import util
def report(actual, predicted, plot=False, save=False):
labels = np.unique(np.concatenate([actual, predicted]))
confusion = confusion_matrix(actual, predicted, labels)
confusion_string = machine_learning.confusion_str(confusion, labels)
scores = classification_report(actual, predicted, target_names=labels)
return '{}\n{}'.format(confusion_string, scores)
def sample_report(tracks, plot=False, save=False):
actual, predicted = [], []
for track in tracks:
predicted.extend(track['sample_predictions'])
actual.extend([track['label']] * len(track['sample_predictions']))
return report(actual, predicted, plot, save)
def track_report(tracks, plot=False, save=False):
actual, predicted = [], []
for track in tracks:
actual.append(track['label'])
predicted.append(track['prediction'])
return report(actual, predicted, plot, save)
def best_svm(tracks, feature_names, n_iter=200, save=False):
clf = machine_learning.Classifier('rbfsvm')
X, Y = machine_learning.shape_features(tracks, feature_names)
param_dist = {
'C': scipy.stats.expon(scale=1000),
'class_weight': ['auto'],
#'loss': ['squared_hinge'],
#'penalty': ['l2'],
#'dual': [False],
'tol': scipy.stats.expon(scale=0.1),
}
logging.info('Optimizing parameters: {}'.format(param_dist))
random_search = sklearn.grid_search.RandomizedSearchCV(
clf.clf,
param_distributions=param_dist,
n_iter=n_iter,
verbose=10,
)
random_search.fit(X, Y)
for score in random_search.grid_scores_:
print(score)
print('Best Score: {}'.format(random_search.best_score_))
print('Best Params: {}'.format(random_search.best_params_))
if save:
logging.info('Saving classifier to disk...')
joblib.dump(random_search.best_estimator_, save, compress=True)
return random_search.best_estimator_
def cross_val_score(tracks, feature_names, folds=5):
X, Y = machine_learning.shape_features(tracks, feature_names)
clf = sklearn.svm.LinearSVC(class_weight='auto')
scores = cross_validation.cross_val_score(
clf,
X,
Y,
cv=folds,
scoring='f1_weighted'
)
return scores
def kfold(tracks, feature_names, folds=5, shuffle=True, **kwargs):
labels = [track['label'] for track in tracks]
kf = cross_validation.StratifiedKFold(labels, n_folds=folds, shuffle=shuffle)
for train, test in kf:
train_tracks = [tracks[i] for i in train]
test_tracks = [tracks[i] for i in test]
clf = machine_learning.Classifier(**kwargs)
clf = machine_learning.train_tracks(clf, train_tracks, feature_names)
predicted_all = []
Y_test_all = []
for track in test_tracks:
X_test, Y_test = machine_learning.shape_features([track], feature_names)
predicted = machine_learning.predict(X_test, clf)
track['sample_predictions'] = predicted
track['prediction'], track['predictions'] = util.most_common(predicted)
predicted_all.extend(predicted)
Y_test_all.extend(Y_test)
yield test_tracks
def main(**kwargs):
start = time()
tracks, args = feature_extraction.load_tracks(**kwargs)
if kwargs['action'] == 'kfold':
folds = kfold(tracks, **kwargs)
for tracks in folds:
scores = sample_report(tracks, plot=True)
print(scores)
scores = track_report(tracks, plot=True)
print(scores)
elif kwargs['action'] == 'cross_val_score':
scores = cross_val_score(tracks, args['feature_names'], folds=args.folds)
print(scores)
elif kwargs['action'] == 'optimize':
clf = best_svm(tracks, args['feature_names'], save=kwargs['save_classifier'])
print(clf)
end = time()
logging.info('Elapsed time: {}'.format(end - start))
if __name__ == '__main__':
parser = argparse.ArgumentParser(
description="Extract instrument stems from medleydb")
parser.add_argument('action', type=str,
choices={'kfold', 'cross_val_score', 'optimize'},
help='Action to take')
parser.add_argument('label', type=str,
choices={'instrument', 'genre'},
help='Track label')
parser.add_argument('-s', '--save_features', type=str, default=None,
help='Location to save pickled features to')
parser.add_argument('-l', '--load_features', type=str, default=None,
help='Location to load pickled features from')
parser.add_argument('-m', '--min_sources', type=int, default=10,
help='Min sources required for instrument selection')
parser.add_argument('-i', '--instruments', nargs='*', default=None,
help='List of instruments to extract')
parser.add_argument('-g', '--genres', nargs='*', default=None,
help='List of genres to extract')
parser.add_argument('-c', '--count', type=int, default=None,
help='Max number of tracks for each label')
parser.add_argument('-r', '--rm_silence', action='store_true',
help='Remove silence from audio files')
parser.add_argument('-t', '--trim', type=int, default=None,
help='Trim audio files to this length (in seconds)')
parser.add_argument('-k', '--folds', type=int, default=5,
help='Number of folds in kfold cross validation')
parser.add_argument('-n', '--n_fft', type=int, default=2048,
help='FFT size of MFCCs')
parser.add_argument('--hop_length', type=int, default=1024,
help='Hop size of MFCCs')
parser.add_argument('-a', '--average', type=int, default=None,
help='Number of seconds to average features over')
parser.add_argument('--normalize', action='store_true',
help='Normalize MFCC feature vectors between 0 and 1')
parser.add_argument('-f', '--feature_names', nargs='+', default=None,
choices=['mfcc', 'mfcc_delta', 'mfcc_delta_delta'],
help='List of features names to use')
parser.add_argument('--save_classifier', type=str, default=None,
help='Location to save pickled classifier to')
parser.add_argument('--classifier', type=str, default='svm',
choices=['linearsvm', 'rbfsvm', 'adaboost'],
help='Type of classifier to use.')
args = parser.parse_args()
main(**vars(args))