【发布时间】：2022-08-03 19:35:12
【问题描述】：
我试图按照音频分类教程进行操作，但遇到了两个错误。我搜索了解决方案，但是当我尝试将 -sparse_categorical_crossentropy 代替 categorical_crossentropy 时，它们没有工作。我不太确定如何修复这些错误，因为其中一个是关于 num_labels 引用，我在代码中的其他任何地方都找不到。
**这是我得到的两个错误：**
```model.add(Dense(num_labels))```
-------------------------------------------------------------------------
NameError Traceback (most recent call last)
<ipython-input-67-ed8693eee280> in <module>()
----> 1 model.add(Dense(num_labels))
NameError: name \'num_labels\' is not defined
[1]: https://i.stack.imgur.com/LhWTq.png
```from tensorflow.keras.callbacks import ModelCheckpoint
from datetime import datetime
num_epochs = 200
num_batch_size = 32
checkpointer = ModelCheckpoint(filepath=\'saved_models/audio_classification.hdf5\',
verbose=1, save_best_only=True)
start = datetime.now()
model.fit(X_train, y_train, batch_size=num_batch_size, epochs=num_epochs, validation_data=(X_test, y_test), callbacks=[checkpointer], verbose=1)
duration = datetime.now() - start
print(\"Training completed in time: \", duration)```
---------------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-73-d89e0e6707f0> in <module>()
9 start = datetime.now()
10
---> 11 model.fit(X_train, y_train, batch_size=num_batch_size, epochs=num_epochs, validation_data=(X_test, y_test), callbacks=[checkpointer], verbose=1)
12
13
4 frames
/usr/local/lib/python3.7/dist-packages/keras/engine/training_utils_v1.py in check_loss_and_target_compatibility(targets, loss_fns, output_shapes)
811 raise ValueError(\'A target array with shape \' + str(y.shape) +
812 \' was passed for an output of shape \' + str(shape) +
--> 813 \' while using as loss `\' + loss_name + \'`. \'
814 \'This loss expects targets to have the same shape \'
815 \'as the output.\')
ValueError: A target array with shape (6985, 10) was passed for an output of shape (None, 100) while using as loss `categorical_crossentropy`. This loss expects targets to have the same shape as the output.
这是到目前为止的代码：
```#@title Preprocessing...
import librosa
import os
from sklearn.model_selection import train_test_split
import tensorflow as tf
#from keras.utils import to_categorical
from tensorflow.keras.utils import to_categorical
import numpy as np
from tqdm import tqdm
DATA_PATH = \"./data/\"
# Input: Folder Path
# Output: Tuple (Label, Indices of the labels, one-hot encoded labels)
def get_labels(path=DATA_PATH):
labels = os.listdir(path)
label_indices = np.arange(0, len(labels))
return labels, label_indices, to_categorical(label_indices)
# convert file to wav2mfcc
# Mel-frequency cepstral coefficients
def wav2mfcc(file_path, n_mfcc=20, max_len=11):
wave, sr = librosa.load(file_path, mono=True, sr=None)
wave = np.asfortranarray(wave[::3])
mfcc = librosa.feature.mfcc(wave, sr=16000, n_mfcc=n_mfcc)
# If maximum length exceeds mfcc lengths then pad the remaining ones
if (max_len > mfcc.shape[1]):
pad_width = max_len - mfcc.shape[1]
mfcc = np.pad(mfcc, pad_width=((0, 0), (0, pad_width)), mode=\'constant\')
# Else cutoff the remaining parts
else:
mfcc = mfcc[:, :max_len]
return mfcc
def save_data_to_array(path=DATA_PATH, max_len=11, n_mfcc=20):
labels, _, _ = get_labels(path)
for label in labels:
# Init mfcc vectors
mfcc_vectors = []
wavfiles = [path + label + \'/\' + wavfile for wavfile in os.listdir(path + \'/\' + label)]
for wavfile in tqdm(wavfiles, \"Saving vectors of label - \'{}\'\".format(label)):
mfcc = wav2mfcc(wavfile, max_len=max_len, n_mfcc=n_mfcc)
mfcc_vectors.append(mfcc)
np.save(label + \'.npy\', mfcc_vectors)
def get_train_test(split_ratio=0.6, random_state=42):
# Get available labels
labels, indices, _ = get_labels(DATA_PATH)
# Getting first arrays
X = np.load(labels[0] + \'.npy\')
y = np.zeros(X.shape[0])
# Append all of the dataset into one single array, same goes for y
for i, label in enumerate(labels[1:]):
x = np.load(label + \'.npy\')
X = np.vstack((X, x))
y = np.append(y, np.full(x.shape[0], fill_value= (i + 1)))
assert X.shape[0] == len(y)
return train_test_split(X, y, test_size= (1 - split_ratio), random_state=random_state, shuffle=True)
def prepare_dataset(path=DATA_PATH):
labels, _, _ = get_labels(path)
data = {}
for label in labels:
data[label] = {}
data[label][\'path\'] = [path + label + \'/\' + wavfile for wavfile in os.listdir(path + \'/\' + label)]
vectors = []
for wavfile in data[label][\'path\']:
wave, sr = librosa.load(wavfile, mono=True, sr=None)
# Downsampling
wave = wave[::3]
mfcc = librosa.feature.mfcc(wave, sr=16000)
vectors.append(mfcc)
data[label][\'mfcc\'] = vectors
return data
def load_dataset(path=DATA_PATH):
data = prepare_dataset(path)
dataset = []
for key in data:
for mfcc in data[key][\'mfcc\']:
dataset.append((key, mfcc))
return dataset[:100]
# print(prepare_dataset(DATA_PATH))
!pip install librosa
upload kaggle.json
import pandas as pd
import os
import librosa
import librosa.display
import numpy as np
import IPython.display as ipd
import matplotlib.pyplot as plt
%matplotlib inline
!pip install kaggle
os.environ[\'KAGGLE_CONFIG_DIR\'] = \"/content\"
!kaggle datasets download -d chrisfilo/urbansound8k
!unzip urbansound8k.zip
file_name=\'fold5/100263-2-0-121.wav\'
audio_data, sampling_rate = librosa.load(file_name)
librosa.display.waveplot(audio_data,sr=sampling_rate)
ipd.Audio(file_name)
audio_data
sampling_rate
audio_dataset_path=\'/content/\'
metadata=pd.read_csv(\'UrbanSound8K.csv\')
metadata.head()
metadata[\'class\'].value_counts()
mfccs = librosa.feature.mfcc(y=audio_data, sr=sampling_rate, n_mfcc=40)
mfccs
def features_extractor(file):
audio, sample_rate = librosa.load(file_name, res_type=\'kaiser_fast\')
mfccs_features = librosa.feature.mfcc(y=audio, sr=sample_rate, n_mfcc=40)
mfccs_scaled_features = np.mean(mfccs_features.T,axis=0)
return mfccs_scaled_features
extracted_features=[]
for index_num,row in tqdm(metadata.iterrows()):
file_name = os.path.join(os.path.abspath(audio_dataset_path),\'fold\'+str(row[\"fold\"])+\'/\',str(row[\"slice_file_name\"]))
final_class_labels=row[\"class\"]
data=features_extractor(file_name)
extracted_features.append([data,final_class_labels])
extracted_features_df=pd.DataFrame(extracted_features,columns=[\'feature\',\'class\'])
extracted_features_df.head(10)
X=np.array(extracted_features_df[\'feature\'].tolist())
y=np.array(extracted_features_df[\'class\'].tolist())
from tensorflow.keras.utils import to_categorical
from sklearn.preprocessing import LabelEncoder
labelencoder=LabelEncoder()
y=to_categorical(labelencoder.fit_transform(y))
from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.2,random_state=0)
Preprocessing is finished
import tensorflow as tf
print(tf.__version__)
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense,Dropout,Activation,Flatten
from tensorflow.keras.optimizers import Adam
from sklearn import metrics
model=Sequential()
###first layer
model.add(Dense(100,input_shape=(40,)))
model.add(Activation(\'relu\'))
model.add(Dropout(0.5))
###second layer
model.add(Dense(200))
model.add(Activation(\'relu\'))
model.add(Dropout(0.5))
###third layer
model.add(Dense(100))
model.add(Activation(\'relu\'))
model.add(Dropout(0.5))
###final layer
model.add(Activation(\'softmax\'))
model.add(Dense(num_labels))
model.compile(loss=\'categorical_crossentropy\',metrics=[\'accuracy\'],optimizer=\'adam\')
from tensorflow.keras.callbacks import ModelCheckpoint
from datetime import datetime
num_epochs = 200
num_batch_size = 32
checkpointer = ModelCheckpoint(filepath=\'saved_models/audio_classification.hdf5\',
verbose=1, save_best_only=True)
start = datetime.now()
model.fit(X_train, y_train, batch_size=num_batch_size, epochs=num_epochs, validation_data=(X_test, y_test), callbacks=[checkpointer], verbose=1)
duration = datetime.now() - start
print(\"Training completed in time: \", duration)```
标签： python tensorflow