def cmd1():
paser = argparse.ArgumentParser(description='Speech Emotion Recognition')
# python3 cmd.py -o p - a 'Test/fear.wav'
# paser.add_argument(
# '-o',
# '--option',
# type = str,
# dest = 'option',
# help = "Use 'p' to predict directly or use 't' to train a model.")
#
# paser.add_argument(
# '-a',
# '--audio',
# type = str,
# dest = 'audio',
# help = "The path of audio which you want to predict.")
#
# args = paser.parse_args()
# option = args.option.lower() # p / t
option = 'p'
model_name = "lstm"
model_type = "lstm"
feature = "l"
load = True
# audio = args.audio if args.audio else 'default.wav'
audio = "output.wav"
# predict
if option == 'p':
model = load_model(load_model_name=model_name, model_name=model_type)
emotion = Predict(model, file_path=audio)
return emotion
# train
# elif option == 't':
# Train(model_name = model_type, save_model_name = model_name, if_load = load, feature_method = feature)
# else:
# print("Wrong option. 'p' for predicting, 't' for training")
# return
# if __name__ == '__main__':
# cmd1()
def upload_file():
if request.method == 'POST' or request.method == 'GET':
print(request.method)
# check if the post request has the file part
if 'file' not in request.files:
print('No file part')
return
file = request.files['file']
print(file.filename)
# if user does not select file, browser also
# submit a empty part without filename
if file.filename == '':
print('No selected file')
return redirect(request.url)
if file: # and allowed_file(file.filename):
filename = secure_filename(file.filename)
file.save(os.path.join(app.config['UPLOAD_FOLDER'], filename))
inputloc = os.path.join(app.config['UPLOAD_FOLDER'], filename)
model = load_model(load_model_name="LSTM_LIBROSA",
model_name="lstm")
emotion, prob = Predict(model,
model_name="lstm",
file_path=inputloc,
feature_method='l')
keras.backend.clear_session()
result = {
"emotion": emotion,
"prob": {
"angry": prob[0].tolist(),
"fear": prob[1].tolist(),
"happy": prob[2].tolist(),
"neutral": prob[3].tolist(),
"sad": prob[4].tolist(),
"surprise": prob[5].tolist()
}
}
return jsonify(json.dumps(result))
def cmd():
paser = argparse.ArgumentParser(description='Speech Emotion Recognition')
paser.add_argument(
'-o',
'--option',
type=str,
dest='option',
help="Use 'p' to predict directly or use 't' to train a model.")
paser.add_argument('-mt',
'--model_type',
type=str,
dest='model_type',
help="The type of model.")
paser.add_argument('-mn',
'--model_name',
type=str,
dest='model_name',
help="The name of saved model file.")
paser.add_argument('-l',
'--load',
type=bool,
dest='load',
help="Whether to load exist features.")
paser.add_argument(
'-f',
'--feature',
type=str,
dest='feature',
help=
"The method for features extracting: use 'o' to use opensmile or use 'l' to use librosa."
)
paser.add_argument('-a',
'--audio',
type=str,
dest='audio',
help="The path of audio which you want to predict.")
args = paser.parse_args()
option = args.option.lower() # p / t
model_type = args.model_type if args.model_type else 'svm' # svm / mlp / lstm
model_name = args.model_name if args.model_name else 'default'
load = args.load if args.load else True # True / False
feature = args.feature if args.feature else 'o' # o / l
audio = args.audio if args.audio else 'default.wav'
# 预测
if option == 'p':
model = load_model(load_model_name=model_name, model_name=model_type)
Predict(model,
model_name=model_type,
file_path=audio,
feature_method=feature)
# 训练
elif option == 't':
Train(model_name=model_type,
save_model_name=model_name,
if_load=load,
feature_method=feature)
else:
print("Wrong option. 'p' for predicting, 't' for training")
return
def get_emotion(path):
print(Predict(model, model_name='svm', file_path=path, feature_method='l'))