Merge branch 'it18257632' into 'master'

Trained model

See merge request !53

backend/IT18257632/train.py

+import json
+import numpy as np
+import tensorflow.keras as keras
+from sklearn.model_selection import train_test_split
+
+DATA_PATH = "data.json"
+SAVE_MODEL_PATH = "model.h5"
+
+LEARNING_RATE = 0.0001
+EPOCHS = 40
+BATCH_SIZE = 32
+
+NUM_KEYWORDS = 15
+
+
+def load_dataset(data_path):
+    with open(data_path, "r") as fp:
+        data = json.load(fp)
+
+    # extract input and targets
+    X = np.array(data["MFCCs"])
+    y = np.array(data["labels"])
+
+    return X, y
+
+
+def get_data_splits(data_path, test_size=0.1, test_validation=0.1):
+    # load dataset
+    X, y = load_dataset(data_path)
+
+    # create train/validation/test splits
+    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=test_size)
+    X_train, X_validation, y_train, y_validation = train_test_split(X_train, y_train, test_size=test_validation)
+
+    # convert inputs from 2D TO 3D array
+    X_train = X_train[..., np.newaxis]
+    X_validation = X_validation[..., np.newaxis]
+    X_test = X_test[..., np.newaxis]
+
+    return X_train, X_validation, X_test, y_train, y_validation, y_test
+
+
+def build_model(input_shape, learning_rate, error="sparse_categorical_crossentropy"):
+    # build network
+    model = keras.Sequential()
+
+    # copy layer 1
+    model.add(keras.layers.Conv2D(64, (3, 3), activation="relu", input_shape=input_shape,
+                                  kernel_regularizer=keras.regularizers.l2(0.001)))
+    model.add(keras.layers.BatchNormalization())
+    model.add(keras.layers.MaxPool2D((3, 3), strides=(2, 2), padding="same"))
+
+    # copy layer 2
+    model.add(keras.layers.Conv2D(32, (3, 3), activation="relu", kernel_regularizer=keras.regularizers.l2(0.001)))
+    model.add(keras.layers.BatchNormalization())
+    model.add(keras.layers.MaxPool2D((3, 3), strides=(2, 2), padding="same"))
+
+    # copy layer 3
+    model.add(keras.layers.Conv2D(32, (2, 2), activation="relu", kernel_regularizer=keras.regularizers.l2(0.001)))
+    model.add(keras.layers.BatchNormalization())
+    model.add(keras.layers.MaxPool2D((2, 2), strides=(2, 2), padding="same"))
+
+    # flatten the output feed it into a dense layer
+    model.add(keras.layers.Flatten())
+    model.add(keras.layers.Dense(64, activation="relu"))
+    model.add(keras.layers.Dropout(0.3))
+
+    # softmax classifier
+    model.add(keras.layers.Dense(NUM_KEYWORDS, activation="softmax"))
+
+    # compile the model
+    optimiser = keras.optimizers.Adam(learning_rate=learning_rate)
+    model.compile(optimizer=optimiser, loss=error, metrics=["accuracy"])
+
+    # print model overview
+    model.summary()
+
+    return model
+
+
+def main():
+    # load train/validation/test data splits
+    X_train, X_validation, X_test, y_train, y_validation, y_test = get_data_splits(DATA_PATH)
+
+    # build the CNN model
+    input_shape = (X_train.shape[1], X_train.shape[2], X_train.shape[3],)
+    model = build_model(input_shape, LEARNING_RATE)
+
+    # train the model
+    model.fit(X_train, y_train, epochs=EPOCHS, batch_size=BATCH_SIZE, validation_data=(X_validation, y_validation))
+
+    # evaluate the model
+    test_error, test_accuracy = model.evaluate(X_test, y_test)
+    print(f"Test error: {test_error}, test accuracy: {test_accuracy}")
+
+    # save the model
+    model.save(SAVE_MODEL_PATH)
+
+
+if __name__ == "__main__":
+    main()