def main(argv):
# for SVC plotting: http://scikit-learn.org/stable/auto_examples/svm/plot_iris.html
genres = [
'classical',
'jazz',
# 'opera',
# 'reggae',
'rock',
'house',
'hip hop'
]
features = [
'Centroid_AVG',
'Centroid_SD',
# 'RollOff_AVG',
# 'RollOff_SD',
# 'Flux_AVG',
# 'Flux_SD',
# 'BPM',
# 'Noise',
# 'NoiseTwo'
]
print "retrieving"
db_service = DatabaseService()
training_set = db_service.get_training_set(genres, features, 100)
print "training"
learner = LearnerSVC()
learner.fit(training_set)
print "plotting"
# plot_compare(genres, features, training_set)
plot_compare_SVC(genres, features, training_set, learner)
def main(argv):
"""
:param argv:
[0] filepath to the song to be identified
[1] filepath to the pickled SVM
:return: outputs the genre
"""
SVM_filepath = "/home/damian/Music-Genre-Classification/Classifiers/SVM_Latest.pickled"
if len(argv) > 2 or len(argv) == 0:
print "Invalid parameters. Use example:"
print "\tpython SVM_classify.py <.wav filepath> (optional: <filepath to pickled SVM>)"
return
elif len(argv) == 2:
SVM_filepath = argv[1]
WAV_filepath = argv[0]
MP3_filepath = WAV_filepath + ".mp3"
# normalize the file
p = Popen(['ffmpeg', '-i', WAV_filepath, MP3_filepath, '-y', '-v', '0'])
p.wait()
try:
p = Popen(['mp3gain', '-q', '-r', '-k', '-f', MP3_filepath], stdout=PIPE)
p.wait()
except:
pass
p = Popen(['ffmpeg', '-i', MP3_filepath, WAV_filepath, '-y', '-v', '0'])
p.wait()
os.remove(MP3_filepath)
learner = pickle.load(open(SVM_filepath, "rb"))
dbs = DatabaseService()
evaluated_features = dbs.get_features_for_song(WAV_filepath, learner.required_features)
result = learner.predict(np.array(evaluated_features).reshape(1, -1))
# mapp house to electronic
if result == 'house':
result = "electronic"
print result
