genre_classification.py

import json
import numpy as np
from sklearn.model_selection import train_test_split
from tensorflow import keras
import matplotlib.pyplot as plt

DATASET_PATH = "data.json"


# load data
# split the data into train and test
# build the network architecture
# compile network
# train network

# load data
def load_data(dataset_path):
    with open(dataset_path, "r") as fp:
        data = json.load(fp)

    # convert lists into numpy arrays
    inputs = np.array(data["mfcc"])
    targets = np.array(data["labels"])

    return inputs, targets

def plot_history(history):

    fig,axs=plt.subplots(2)

    #create accuracy subplot
    axs[0].plot(history.history["accuracy"],label="train accuracy")
    axs[0].plot(history.history["val_accuracy"], label="test accuracy")
    axs[0].set_ylabel("Accuracy")
    axs[0].legend(loc="lower right")
    axs[0].set_title("Accuracy eval")

    #create error subplot
    axs[1].plot(history.history["loss"],label="train error")
    axs[1].plot(history.history["val_loss"], label="test error")
    axs[1].set_ylabel("Error")
    axs[1].set_xlabel("Epoch")
    axs[1].legend(loc="upper right")
    axs[1].set_title("Error eval")

    plt.show()

if __name__ == "__main__":
    # load data
    inputs, targets = load_data(
        DATASET_PATH)  # actually inputs has three dimension when we build model we dont use zero dimension

    # split dataset into train and test
    inputs_train, inputs_test, targets_train, targets_test = train_test_split(inputs,
                                                                              targets,
                                                                              test_size=0.3)

    # build the network architecture(model)
    model = keras.Sequential([
        # input layer
        keras.layers.Flatten(input_shape=(inputs.shape[1], inputs.shape[2])),
        # Flatten pass directly multi-dimension arrays

        # 1st hidden layer
        keras.layers.Dense(512, activation="relu",kernel_regularizer=keras.regularizers.l2(0.001)),
        keras.layers.Dropout(0.3),

        # 2nd hidden layer
        keras.layers.Dense(256, activation="relu",kernel_regularizer=keras.regularizers.l2(0.001)),
        keras.layers.Dropout(0.3),

        # 3rd hidden layer
        keras.layers.Dense(64, activation="relu",kernel_regularizer=keras.regularizers.l2(0.001)),
        keras.layers.Dropout(0.3),

        # output layer
        keras.layers.Dense(10, activation="softmax")  # use 10 norons because we want to split 10 categories
    ])

    # compile model
    optimizer = keras.optimizers.Adam(learning_rate=0.0001)
    model.compile(optimizer=optimizer,
                  loss="sparse_categorical_crossentropy",
                  metrics=["accuracy"])

    model.summary()  # The model summary table reports the strength of the relationship between the model and the dependent variable

    # train model
    history=model.fit(inputs_train, targets_train,          #solving overfitting: add variable and method
              validation_data=(inputs_test, targets_test),
              epochs=50,
              batch_size=32)

    #plots accuracy and error over the epochs
    plot_history(history)