给出相同预测的音乐分类器模型

Question

我试图用CNN模型做一个音乐分类器，但我总是得到同样的预测，这是错误的 我使用了GTZAN歌曲数据集

import librosa
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import os
import csv
# Preprocessing
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder, StandardScaler
#Keras
import keras
from keras import models
from keras import layers


# generating a dataset
header = 'filename chroma_stft rmse spectral_centroid spectral_bandwidth rolloff zero_crossing_rate'
for i in range(1, 21):
    header += f' mfcc{i}'
header += ' label'
header = header.split()

file = open('data.csv', 'w', newline='')
with file:
    writer = csv.writer(file)
    writer.writerow(header)
genres = 'blues classical country disco hiphop jazz metal pop reggae rock'.split()
for g in genres:
    for filename in os.listdir(f'C:/Users/USER/Desktop/sem8/AI/project/gtzan.keras/data/genres/{g}'):
        songname = f'C:/Users/USER/Desktop/sem8/AI/project/gtzan.keras/data/genres/{g}/{filename}'
        y, sr = librosa.load(songname, mono=True, duration=30)
        chroma_stft = librosa.feature.chroma_stft(y=y, sr=sr)
        rmse = librosa.feature.rms(y=y)
        spec_cent = librosa.feature.spectral_centroid(y=y, sr=sr)
        spec_bw = librosa.feature.spectral_bandwidth(y=y, sr=sr)
        rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr)
        zcr = librosa.feature.zero_crossing_rate(y)
        mfcc = librosa.feature.mfcc(y=y, sr=sr)
        to_append = f'{filename} {np.mean(chroma_stft)} {np.mean(rmse)} {np.mean(spec_cent)} {np.mean(spec_bw)} {np.mean(rolloff)} {np.mean(zcr)}'    
        for e in mfcc:
            to_append += f' {np.mean(e)}'
        to_append += f' {g}'

        file = open('datatest3.csv', 'a', newline='')
        with file:
            writer = csv.writer(file)
            writer.writerow(to_append.split())

# reading dataset from csv

data = pd.read_csv('datatest3.csv')
data.head()

# Dropping unneccesary columns
data = data.drop(['filename'],axis=1)
data.head()

genre_list = data.iloc[:, -1]
encoder = LabelEncoder()
y = encoder.fit_transform(genre_list)
print(y)

# normalizing
scaler = StandardScaler()
X = scaler.fit_transform(np.array(data.iloc[:, :-1], dtype = float))

# spliting of dataset into train and test dataset
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)


# creating a model
model = models.Sequential()
model.add(layers.Dense(256, activation='relu', input_shape=(X_train.shape[1],)))

model.add(layers.Dense(128, activation='relu'))

model.add(layers.Dense(64, activation='relu'))

model.add(layers.Dense(10, activation='softmax'))

model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

history = model.fit(X_train,
                    y_train,
                    epochs=20,
                    batch_size=128)

# calculate accuracy
test_loss, test_acc = model.evaluate(X_test,y_test)
print('test_acc: ',test_acc)

# predictions
predictions = model.predict(X_test)
np.argmax(predictions[0])

这就是我制作csv文件的方式

data = pd.read_csv('C:\\Users\\USER\\Desktop\\sem8\\AI\\project\\try\\datatest.csv')
data.head()

genre_list = data.iloc[:, -1]
encoder = LabelEncoder()
y = encoder.fit_transform(genre_list)
print(y)

# normalizing
scaler = StandardScaler()
X = scaler.fit_transform(np.array(data.iloc[:, :-1], dtype = float))

# spliting of dataset into train and test dataset
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)


# creating a model
model = models.Sequential()
model.add(layers.Dense(256, activation='relu', input_shape=(X_train.shape[1],)))

model.add(layers.Dense(128, activation='relu'))

model.add(layers.Dense(64, activation='relu'))

model.add(layers.Dense(10, activation='softmax'))

model.compile(optimizer='adam',
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

model.summary()              
history = model.fit(X_train,
                    y_train,
                    epochs=20,
                    batch_size=128)

# calculate accuracy
test_loss, test_acc = model.evaluate(X_test,y_test)
print('test_acc: ',test_acc)


# predictions
print(X_test[0])
predictions = model.predict(X_test)
print(np.argmax(predictions[0]))
# model.summary()
model.save("C:\\Users\\USER\\Desktop\\sem8\\AI\\project\\try\\mymodel.h5")
print("Saved model to disk")

然后我创建并保存了模型

genres = {
    'metal': 0, 'disco': 1, 'classical': 2, 'hiphop': 3, 'jazz': 4, 
    'country': 5, 'pop': 6, 'blues': 7, 'reggae': 8, 'rock': 9
}
def majority_voting(scores, dict_genres):
    preds = np.argmax(scores, axis = 1)
    values, counts = np.unique(preds, return_counts=True)
    counts = np.round(counts/np.sum(counts), 2)
    votes = {k:v for k, v in zip(values, counts)}
    votes = {k: v for k, v in sorted(votes.items(), key=lambda item: item[1], reverse=True)}
    return [(get_genres(x, dict_genres), prob) for x, prob in votes.items()]


def get_genres(key, dict_genres):
    # Transforming data to help on transformation
    labels = []
    tmp_genre = {v:k for k,v in dict_genres.items()}

    return tmp_genre[key]
def prepare(filename):
    # y, sr = librosa.load(filename, mono=True, duration=30)

    # return y
    # signal, sr = librosa.load(filename, sr=None)
    y, sr = librosa.load(filename, mono=True, duration=30)
    chroma_stft = librosa.feature.chroma_stft(y=y, sr=sr)
    rmse = librosa.feature.rms(y=y)
    spec_cent = librosa.feature.spectral_centroid(y=y, sr=sr)
    spec_bw = librosa.feature.spectral_bandwidth(y=y, sr=sr)
    rolloff = librosa.feature.spectral_rolloff(y=y, sr=sr)
    zcr = librosa.feature.zero_crossing_rate(y)
    mfcc = librosa.feature.mfcc(y=y, sr=sr)
    to_append = f'{np.mean(chroma_stft)} {np.mean(rmse)} {np.mean(spec_cent)} {np.mean(spec_bw)} {np.mean(rolloff)} {np.mean(zcr)}'    
    for e in mfcc:
        to_append += f' {np.mean(e)}'

    # Append the result to the data structure
    # features = get_features(signal, sr)
    song = pd.DataFrame([to_append.split()])
    return song

model=models.load_model("C:\\Users\\USER\\Desktop\\sem8\\AI\\project\\try\\mymodel.h5")
print(prepare("C:\\Users\\USER\\Desktop\\sem8\\AI\\project\\try\\song1.mp3"))
newsong=prepare("C:\\Users\\USER\\Desktop\\sem8\\AI\\project\\try\\song1.mp3")
song = (np.array(newsong)).reshape(26)
print(song.shape)
print(song)
prediction = model.predict(song)
votes = majority_voting(prediction, genres)
print("This song is a {} song".format(votes[0][0]))
print("most likely genres are: {}".format(votes[:3]))
print(prediction)

这就是我试图预测的方式
我总是得到同样的结果
我认为我所做的一切都是正确的

This song is a hiphop song
most likely genres are: [('hiphop', 1.0)]
[[0. 0. 0. 1. 0. 0. 0. 0. 0. 0.]]

Answer 1

问题在于对预测数据进行规范化，它的形状应该与使用standardscaler进行训练时使用的数据相同