i created the environment as per the requirements. But still i am getting an error

$ python2 mlmodel.py 1
Traceback (most recent call last):
File "mlmodel.py", line 9, in
from . import Model
ValueError: Attempted relative import in non-package

Hello,
can we use the current model without training?

TypeError: pad_width must be of integral type
getting this error when i run the code

[3 1 1 0 1 2 1 3 3 0 1 2 2 1 1 1 1 3 2 0 3 0 1 1 1 1 1 2 1 3 1 0 1 3 0 0 3
3 1 0 1 0 1 2 0 3 3 3 0 0 3 1 1 1 0 1 0 3 3 0 3 3 3 1 2 0 0 0]
(0, 0, 0, 0, 1, 0, 0, 3, 3, 0, 2, 0, 2, 0, 0, 0, 2, 3, 0, 0, 3, 0, 0, 0, 1, 0, 0, 0, 3, 3, 0, 0, 0, 3, 0, 0, 3, 3, 1, 0, 0, 0, 1, 0, 0, 0, 3, 3, 0, 0, 0, 1, 2, 1, 0, 1, 0, 3, 0, 0, 3, 0, 3, 0, 0, 0, 0, 0)
Accuracy:0.574

Confusion matrix: [[18 0 0 0]
[14 7 3 1]
[ 6 0 1 0]
[ 5 0 0 13]]
Traceback (most recent call last):
File "emotionRecognition_cnn.py", line 63, in
cnn_example()
File "emotionRecognition_cnn.py", line 33, in cnn_example
predicted = cnn.predict_one(feature)
File "/home/diego/Desktop/progetto/cnn_emotionrecognition/dnn.py", line 98, in predict_one
return np.argmax(self.model.predict(np.array([sample])))
File "/home/diego/.local/lib/python3.7/site-packages/keras/engine/training.py", line 1149, in predict
x, _, _ = self._standardize_user_data(x)
File "/home/diego/.local/lib/python3.7/site-packages/keras/engine/training.py", line 751, in _standardize_user_data
exception_prefix='input')
File "/home/diego/.local/lib/python3.7/site-packages/keras/engine/training_utils.py", line 128, in standardize_input_data
'with shape ' + str(data_shape))
ValueError: Error when checking input: expected conv2d_1_input to have 4 dimensions, but got array with shape (1, 198, 39)

The code used is:
python3 cnn_examples.py

The envinronment is created via "pip3 install -r requirements.txt" as you suggested

I don't know what this problem is, can you help me?
Thanks

博主，运行完examples/下的文件，并没有保存训练的模型，想知道怎么您给的models/下的模型给出调用方法了吗？可以直接调用吗？

any research paper for this or proper doc

Hello from South Korea! I'm a grad school student who wants to use your code to train a new dataset.
The thing is, I want to know what and how many GPUs you have used for parallel processing, or the cloud computing server you have used to train your model with your best weight.

That's all! Thank you so much for reading :)

recently i asked you regarding accuracy .
#3
i did multiclassification not binary classification
Is keras calculates accuracy by sample wise or labels wise ?
just consider 2 samples only.
suppose y_pred=[ [0.35, 0.2, 0.2, 0.25], [0.33, 0.22, 0.18, 0.27] ]
and my t_test= [ [1, 0, 0, 0] , [0, 0, 0, 1] ]
scikit accuracy is 50% because in scikit all labels must be same
what about keras .how it actually calculates.
can you please explain me with exmples. it will help to me a lot

Is there any report about applied models accuracy?

actually i run the code below code .

from sklearn.metrics import accuracy_score
import numpy as np
import sys
from keras import Sequential
from keras.layers import LSTM, Dense, Dropout, Conv2D, Flatten, \
    BatchNormalization, Activation, MaxPooling2D
from keras.utils import np_utils
from keras.layers import Bidirectional
from tqdm import tqdm

from utilities import get_data, class_labels
import pickle
import os
os.environ["CUDA_VISIBLE_DEVICES"]="0"

models = ["CNN", "LSTM"]


def get_model(model_name, input_shape):
    """
    Generate the required model and return it
    :return: Model created
    """
    # Models are inspired from
    # CNN - https://yashk2810.github.io/Applying-Convolutional-Neural-Network-on-the-MNIST-dataset/
    # LSTM - https://github.com/harry-7/Deep-Sentiment-Analysis/blob/master/code/generatePureLSTM.py
    model = Sequential()
    if model_name == 'CNN':
        model.add(Conv2D(8, (13, 13),
                         input_shape=(input_shape[0], input_shape[1], 1)))
        model.add(BatchNormalization(axis=-1))
        model.add(Activation('relu'))
        model.add(Conv2D(8, (13, 13)))
        model.add(BatchNormalization(axis=-1))
        model.add(Activation('relu'))
        model.add(MaxPooling2D(pool_size=(2, 1)))
        model.add(Conv2D(8, (13, 13)))
        model.add(BatchNormalization(axis=-1))
        model.add(Activation('relu'))
        model.add(Conv2D(8, (2, 2)))
        model.add(BatchNormalization(axis=-1))
        model.add(Activation('relu'))
        model.add(MaxPooling2D(pool_size=(2, 1)))
        model.add(Flatten())
        model.add(Dense(64))
        model.add(BatchNormalization())
        model.add(Activation('relu'))
        model.add(Dropout(0.2))
    elif model_name == 'LSTM':
        model.add(Bidirectional(LSTM(128), input_shape=(input_shape[0], input_shape[1])))
        model.add(Dropout(0.5))
        model.add(Dense(32, activation='relu'))
        model.add(Dense(16, activation='tanh'))
    model.add(Dense(len(class_labels), activation='softmax'))
    model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
    print(model.summary())
    return model


def evaluateModel(model):
    """
    Train the model and evaluate it
    :param model: model to be evaluted
    """
    # Train the epochs
    best_acc = 0
    global x_train, y_train, x_test, y_test
    for i in tqdm(range(50)):
        # Shuffle the data for each epoch in unison inspired from https://stackoverflow.com/a/4602224
        p = np.random.permutation(len(x_train))
        x_train = x_train[p]
        y_train = y_train[p]
        model.fit(x_train, y_train, batch_size=32, epochs=1)
        loss, acc = model.evaluate(x_test, y_test)
        if acc > best_acc:
            print ('Updated best accuracy', acc)
            best_acc = acc
            model.save_weights(best_model_path)
    model.load_weights(best_model_path)
    print ('keras_Accuracy = ', model.evaluate(x_test, y_test)[1])
    y_pred=model.predict(x_test)
    return y_pred


if __name__ == "__main__":

    if len(sys.argv) != 2:
        sys.stderr.write('Invalid arguments\n')
        sys.stderr.write('Usage python2 train_DNN.py <model_number>\n')
        sys.stderr.write('1 - CNN\n')
        sys.stderr.write('2 - LSTM\n')
        sys.exit(-1)

    n = int(sys.argv[1]) - 1
    print ('model given', models[n])

    # Read data
    global x_train, y_train, x_test, y_test
    x_train, x_test, y_train, y_test = get_data(flatten=False)
    y_train = np_utils.to_categorical(y_train)
    y_test = np_utils.to_categorical(y_test)

    if n == 0:
        # Model is CNN so have to reshape the data
        in_shape = x_train[0].shape
        x_train = x_train.reshape(x_train.shape[0], in_shape[0], in_shape[1], 1)
        x_test = x_test.reshape(x_test.shape[0], in_shape[0], in_shape[1], 1)
    elif n > len(models):
        sys.stderr.write('Model Not Implemented yet')
        sys.exit(-1)

    model = get_model(models[n], x_train[0].shape)

    global best_model_path
    best_model_path = '../model/best_model_' + models[n - 1] + '.h5'

    y_pred=evaluateModel(model)
    t=[]
    for i in y_test:
        t.append(np.argmax(i))
    p=[]
    for i in y_pred:
        p.append(np.argmax(i))
    scikit_accuracy=accuracy_score(t,p)*100
    print(scikit_accuracy)

I got 92 as keras_accyracy from model.evalute , but i got 67 as scikit accuracy . could you anyone solve my problem please......@

how to load the previous model and continue training on that checkpoint ?

can you share your best score for CNN and LSTM.

FileNotFoundError Traceback (most recent call last)
in
22
23 if name == "main":
---> 24 ml_example()

in ml_example()
17 filename = './dataset/Neutral/srg1.wav'
18 print('prediction', model.predict_one(
---> 19 get_feature_vector_from_mfcc(filename, flatten=to_flatten)),
20 'Actual 3')
21

~\SERproject\Code\speechemotionrecognition\utilities.py in get_feature_vector_from_mfcc(file_path, flatten, mfcc_len)
31 numpy.ndarray: feature vector of the wav file made from mfcc.
32 """
---> 33 fs, signal = wav.read(file_path)
34 s_len = len(signal)
35 # pad the signals to have same size if lesser than required

c:\users\srg\appdata\local\programs\python\python39\lib\site-packages\scipy\io\wavfile.py in read(filename, mmap)
637 mmap = False
638 else:
--> 639 fid = open(filename, 'rb')
640
641 try:

FileNotFoundError: [Errno 2] No such file or directory: './dataset/Neutral/srg1.wav'

​
why this project does not recognising unknown speech data? please help me.... i want to give my own data.... not emodb dataset....

I am facing a similar error in github. However I tried it in Google Colab and I got the following error:
NN has no attribute called 'train'. Could you please look into the issue?
This is the Colab notebook.
https://colab.research.google.com/drive/19zXD0GzR5-gXqoTEfZIW0PbHswdbc4ZD

Originally posted by @Nandy-Saran in #9 (comment)

Hi, thanks for the code. The documentation link is broken, could you please update it?

When running dnn.py
I'm getting an error as follows:
Using TensorFlow backend.
Traceback (most recent call last):
File "dnn.py", line 11, in
from . import Model
ImportError: cannot import name 'Model'

I know the length of voice signals vary from file to file or you can say that there may b some outliers in the data set. But padding adds zeros in the data. So why we are aimed to equalize the length of the audio signal with zeros? If we are adding zeros in the data, will it distort original data, if yes then we are padding the data?

My second question is that how voice data is normalized, did you normalize the data in current project?

Dear sir,

I am getting error while executing the lstm_example.
could you please check once and please let me know if any modifications has to be done.

Thank you.

Currently one can download and use the package but there is no way to install the package. Create a setup.py and make it installable.

if any audio file is detecting in anger class so i wanted to know the % like it is detecting 10% or 50% or more? how to find this?

I didnt understand your code.. I have a doubt..
Can i provide my own dataset and get the output if the audio is angry,happy,sad or neutral?

Is there a part of the code to get the best models in h5 format? I'm navigating through the package and did not find any. Thanks for the help.

Most of the code has incode documentation. But there are no standard docs for the code and the logic behind it. Add the necessary documentation

im getting this error while passing 3 sec of audio

how to train on ml model with my own custom dataset of call recording ?

Epoch 1/1
271/271 [==============================] - 3s 12ms/step - loss: 0.0187 - acc: 0.9963
68/68 [==============================] - 0s 2ms/step
[3 1 1 0 1 2 1 3 3 0 1 2 2 1 1 1 1 3 2 0 3 0 1 1 1 1 1 2 1 3 1 0 1 3 0 0 3
3 1 0 1 0 1 2 0 3 3 3 0 0 3 1 1 1 0 1 0 3 3 0 3 3 3 1 2 0 0 0]
(3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1)
Accuracy:0.412

Confusion matrix: [[ 0 18 0 0]
[ 0 25 0 0]
[ 0 7 0 0]
[ 0 15 0 3]]
Traceback (most recent call last):
File "cnn_example.py", line 33, in
cnn_example()
File "cnn_example.py", line 27, in cnn_example
get_feature_vector_from_mfcc(filename, flatten=to_flatten)),
File "/home/sww/workspace/speech-emotion-recognition/speechemotionrecognition/dnn.py", line 97, in predict_one
return np.argmax(self.model.predict(np.array([sample])))
File "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py", line 1149, in predict
x, _, _ = self._standardize_user_data(x)
File "/usr/local/lib/python3.6/dist-packages/keras/engine/training.py", line 751, in _standardize_user_data
exception_prefix='input')
File "/usr/local/lib/python3.6/dist-packages/keras/engine/training_utils.py", line 128, in standardize_input_data
'with shape ' + str(data_shape))
ValueError: Error when checking input: expected conv2d_1_input to have 4 dimensions, but got array with shape (1, 198, 39)
root@2d14eba3fb00:/home/sww/workspace/speech-emotion-recognition# Error when checking input: expected conv2d_1_input to have 4 dimensions, but got array with shape (1, 198, 39)

The training code is:
'''
#!/usr/bin/env bash
set -x
export PYTHONPATH=/path_to/speech-emotion-recognition/:$PYTHONPATH
export PATH=/path_to/speech-emotion-recognition/:$PATH
cd examples
python cnn_example.py
'''

There seems to be some problems with the input.
Could you pls help me with this?
Many Thanks.

I ran dnn.py
but I have an error.

Cannot import name 'Model'

(My tensorflow-gpu version is 2.0.0)

How do I solve this problem...?

what does this particular line stands for and how did you reached this particular value?

The python file dl_example ran correctly and we got an accuracy of 0.98 with the training set. We then tried to predict the emotion using predict() function by giving a wave file as an argument. We preloaded the model best_model_LSTM.h5 from the models' folder.

Here is the screenshot of the code of the file dl_example.py

The error and stack trace is attached.

Please look into the issue as soon as you can Harry. Thanks again :)