def predict_pipeline(audio_fpath, model):
import feature
import os
if type(audio_fpath) != str:
audio_fpath = os.path.join(*audio_fpath)
feats = feature.extract(audio_fpath, verbose=False)
predictions = np.argmax(model.predict(feats), axis=1)
return predictions
def predict_pipeline(audio_fpath, model, raw_prob=False):
import feature
import os
if type(audio_fpath) != str:
audio_fpath = os.path.join(*audio_fpath)
feats = feature.extract(audio_fpath)
predictions = model.predict(feats)
if not raw_prob:
predictions = np.argmax(predictions, axis=1)
return predictions
def preprocess(sample_size=10,
feature_path='data/feature_vectors',
label_path='data/labels'):
print('\n===== REMOVE OLD FILES =====\n')
output_dirs = ["./data/feature_vectors/", "./data/labels/", 'model/']
for path in output_dirs:
filelist = [
path + f for f in os.listdir(path)
if f.endswith(".npy") or f.endswith('.pkl')
]
for f in filelist:
os.remove(f)
print('Deleted: ' + f)
print('\n===== FEATURE EXTRACTION =====\n')
feature.extract(sample_size)
print('\n===== BINARIZE LABELS =====\n')
label.binarize(sample_size)
print('\n===== LOADING FEATURE VECTORS AND LABELS =====\n')
X = []
y = []
i = 0
for filename in os.listdir(feature_path):
if i == int(sample_size):
break
if os.path.isfile(label_path + '/' + filename):
feature_vector = np.load(feature_path + '/' + filename)
label_vector = np.load(label_path + '/' + filename)
X.append(feature_vector.tolist())
y.append(label_vector.tolist())
i = i + 1
X = np.matrix(X)
y = np.matrix(y)
return X, y
def getprob(filename):
"""
Find the probality that a song belongs to each genre.
"""
x = feature.extract(filename)
clf = cmpr
prob = clf.predict_proba(x)[0]
#prob = np.round(prob,decimals=-5)
#dd = dict(zip(feature.getlabels(),prob))
dd = dict(zip(['Classical','Hipop','Jass','Metal','Pop','Rock'],prob))
print(prob)
# max probablity
m = max(dd,key=dd.get)
print(m, dd[m])
sorted_genre = sorted(dd,key=dd.get,reverse=True)
has_features_of = []
for i in sorted_genre:
if (dd[i] > 0.15 or dd[i] >= dd[m]) and len(has_features_of) < 3:
has_features_of.append(i)
return dd, has_features_of
def getGenre(filename):
music_feature = feature.extract(os.path.abspath(os.path.dirname(__name__)) \
+'/django-jquery-file-upload/' +filename)
clf = cmpr
return clf.predict(music_feature)
for i in range(1, persons + 1):
for j in range(1, palms_per_person + 1):
file_name = '{}\\{:0>4d}_{:0>2d}.bmp'.format(data_path, i, j)
if os.path.exists(file_name):
file_names.append(file_name)
# generate filters
filters = gabor.generate()
# distance matrix
dist_matrix = np.zeros([len(file_names), len(file_names)])
for i in range(0, len(file_names) - 1):
im1 = Image.open(file_names[i])
f1 = feature.extract(im1, filters)
m1 = matcher.get_mask(im1)
# start = time.clock()
for j in range(i + 1, len(file_names)):
im2 = Image.open(file_names[j])
start = time.clock()
f2 = feature.extract(im2, filters)
m2 = matcher.get_mask(im2)
d = matcher.compute_distance(f1, f2, m1, m2, 4)
dist_matrix[i][j] = d
dist_matrix[j][i] = dist_matrix[i][j]
def main():
csv_file_object = csv.reader(open('csv/train.csv',
'rb')) # Load in the training csv file
header = csv_file_object.next() # Skip the fist line as it is a header
train_data = [] # Creat a variable called 'train_data'
for row in csv_file_object: # Skip through each row in the csv file
train_data.append(row) # adding each row to the data variable
train_data = np.array(train_data) # Then convert from a list to an array
train_data = feature.extract(train_data, "train")
# I need to do the same with the test data now so that the columns are in the same
# as the training data
test_file_object = csv.reader(open('csv/test.csv',
'rb')) # Load in the test csv file
header = test_file_object.next() # Skip the fist line as it is a header
test_data = [] # Creat a variable called 'test_data'
for row in test_file_object: # Skip through each row in the csv file
test_data.append(row) # adding each row to the data variable
test_data = np.array(test_data) # Then convert from a list to an array
test_data = feature.extract(test_data, "test")
# The data is now ready to go. So lets train then test!
print 'Training'
cv_data = np.array([row[1:] for row in train_data])
cv_target = np.array([row[0] for row in train_data])
forest = RandomForestClassifier(n_estimators=100)
scores = cross_validation.cross_val_score(forest, cv_data, cv_target, cv=5)
print "Cross Validation for random forest, scores: " + str(np.mean(scores))
best_score = 0
best_c = 0
best_gamma = 0
C_list = [0.1, 0.3, 1, 3, 10, 30, 100, 300, 1000]
gamma_list = [2**-15, 2**-13, 2**-11, 2**-9, 2**-7, 2**-5, 2**-3, 2**-1]
for c in C_list:
for gamma in gamma_list:
svm_clf = svm.SVC(C=c, gamma=gamma)
scores = cross_validation.cross_val_score(svm_clf,
cv_data,
cv_target,
cv=5)
score = np.mean(scores)
print "Cross Validation for SVM, scores with C = %f gamma = %f: %f" % (
c, gamma, score)
if score > best_score:
best_score = score
best_c = c
best_gamma = gamma
print "Best svm score is %f, with c=%f and gamma=%f" % (best_score, best_c,
best_gamma)
print 'Predicting'
clf = svm.SVC(C=best_c, gamma=best_gamma)
clf.fit(cv_data, cv_target)
output = clf.predict(test_data)
open_file_object = csv.writer(open("csv/submission.csv", "wb"))
test_file_object = csv.reader(open('csv/test.csv',
'rb')) # Load in the csv file
test_file_object.next()
open_file_object.writerow([
"survived", "pclass", "name", "sex", "age", "sibsp", "parch", "ticket",
"fare", "cabin", "embarked"
])
i = 0
for row in test_file_object:
row.insert(0, output[i].astype(np.uint8))
open_file_object.writerow(row)
i += 1
print "Done"
# --encoding:utf-8--
import svm
# 一、模型训练
# 1. 交叉验证
# svm.cross_validation(data_percentage=0.9)
# 2. 根据交叉验证修改代码参数后进行模型训练并输出
# svm.fit_dump_model(train_percentage=0.9, fold=100)
# # 4. 模型调用进行预测
import feature
# path = "../data/test/孙燕姿 - 我也很想他 - 怀旧.mp3"
path = "../data/test/Maize - I Like You-浪漫.mp3"
# path = "../data/test/Lasse Lindh - Run To You.mp3" # 清新
music_feature = feature.extract(path)
clf = svm.load_model()
label = svm.fetch_predict_label(clf, music_feature)
print("预测标签为:%s" % label)
def main():
csv_file_object = csv.reader(open('csv/train.csv', 'rb')) # Load in the training csv file
header = csv_file_object.next() # Skip the fist line as it is a header
train_data = [] # Creat a variable called 'train_data'
for row in csv_file_object: # Skip through each row in the csv file
train_data.append(row) # adding each row to the data variable
train_data = np.array(train_data) # Then convert from a list to an array
train_data = feature.extract(train_data, "train")
# I need to do the same with the test data now so that the columns are in the same
# as the training data
test_file_object = csv.reader(open('csv/test.csv', 'rb')) # Load in the test csv file
header = test_file_object.next() # Skip the fist line as it is a header
test_data = [] # Creat a variable called 'test_data'
for row in test_file_object: # Skip through each row in the csv file
test_data.append(row) # adding each row to the data variable
test_data = np.array(test_data) # Then convert from a list to an array
test_data = feature.extract(test_data, "test")
# The data is now ready to go. So lets train then test!
print 'Training'
cv_data = np.array([row[1:] for row in train_data])
cv_target = np.array([row[0] for row in train_data])
forest = RandomForestClassifier(n_estimators=100)
scores = cross_validation.cross_val_score(forest, cv_data, cv_target, cv=5)
print "Cross Validation for random forest, scores: " + str(np.mean(scores))
best_score = 0
best_c = 0
best_gamma = 0
C_list = [0.1, 0.3, 1, 3, 10, 30, 100, 300, 1000]
gamma_list = [2 ** -15, 2 ** -13, 2 ** -11, 2 ** -9, 2 ** -7, 2 ** -5, 2 ** -3, 2 ** -1]
for c in C_list:
for gamma in gamma_list:
svm_clf = svm.SVC(C=c, gamma=gamma)
scores = cross_validation.cross_val_score(svm_clf, cv_data, cv_target, cv=5)
score = np.mean(scores)
print "Cross Validation for SVM, scores with C = %f gamma = %f: %f" % (c, gamma, score)
if score > best_score:
best_score = score
best_c = c
best_gamma = gamma
print "Best svm score is %f, with c=%f and gamma=%f" % (best_score, best_c, best_gamma)
print 'Predicting'
clf = svm.SVC(C=best_c, gamma=best_gamma)
clf.fit(cv_data, cv_target)
output = clf.predict(test_data)
open_file_object = csv.writer(open("csv/submission.csv", "wb"))
test_file_object = csv.reader(open('csv/test.csv', 'rb')) # Load in the csv file
test_file_object.next()
open_file_object.writerow(["survived", "pclass", "name", "sex", "age", "sibsp", "parch", "ticket", "fare", "cabin", "embarked"])
i = 0
for row in test_file_object:
row.insert(0, output[i].astype(np.uint8))
open_file_object.writerow(row)
i += 1
print "Done"