Skip to content

Instantly share code, notes, and snippets.

@c0ldf1recsgo

c0ldf1recsgo/colab-github-demo.ipynb

Created November 27, 2020 02:18
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save c0ldf1recsgo/2c0e1cb4fabd98c5494cccd68d18ce85 to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save c0ldf1recsgo/2c0e1cb4fabd98c5494cccd68d18ce85 to your computer and use it in GitHub Desktop.
Download ZIP
colab-github-demo.ipynb
Display the source blob
Display the rendered blob
Raw
colab-github-demo.ipynb
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"name": "colab-github-demo.ipynb",
"provenance": [],
"collapsed_sections": [],
"include_colab_link": true
},
"kernelspec": {
"display_name": "Python 3",
"name": "python3"
}
},
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "view-in-github",
"colab_type": "text"
},
"source": [
"<a href=\"https://colab.research.google.com/gist/c0ldf1recsgo/2c0e1cb4fabd98c5494cccd68d18ce85/colab-github-demo.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "Pv0dCqitKua3",
"outputId": "7f627294-9682-479b-b633-a38f40e001c3"
},
"source": [
"!git clone https://github.com/ruohoruotsi/LSTM-Music-Genre-Classification.git"
],
"execution_count": 1,
"outputs": [
{
"output_type": "stream",
"text": [
"Cloning into 'LSTM-Music-Genre-Classification'...\n",
"remote: Enumerating objects: 752, done.\u001b[K\n",
"remote: Total 752 (delta 0), reused 0 (delta 0), pack-reused 752\u001b[K\n",
"Receiving objects: 100% (752/752), 727.61 MiB | 33.27 MiB/s, done.\n",
"Resolving deltas: 100% (77/77), done.\n",
"Checking out files: 100% (624/624), done.\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "1npS93-_LRJj",
"outputId": "6bc92adf-3780-4ef4-8e8c-7f9413b0ef07"
},
"source": [
"!ls LSTM-Music-Genre-Classification"
],
"execution_count": 8,
"outputs": [
{
"output_type": "stream",
"text": [
"audio\t\t lstm_genre_classifier_keras.py requirements.txt\n",
"GenreFeatureData.py lstm_genre_classifier_pytorch.py weights\n",
"gtzan\t\t predict_example.py\n",
"LICENSE\t\t README.md\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "7bdgeSEMLuLS"
},
"source": [
"import librosa\n",
"import math\n",
"import os\n",
"import re\n",
"\n",
"import numpy as np\n",
"\n",
"\n",
"class GenreFeatureData:\n",
"\n",
" \"Music audio features for genre classification\"\n",
" hop_length = None\n",
" genre_list = [\n",
" \"classical\",\n",
" \"country\",\n",
" \"disco\",\n",
" \"hiphop\",\n",
" \"jazz\",\n",
" \"metal\",\n",
" \"pop\",\n",
" \"reggae\",\n",
" ]\n",
"\n",
" dir_trainfolder = \"./LSTM-Music-Genre-Classification/gtzan/_train\"\n",
" dir_devfolder = \"./LSTM-Music-Genre-Classification/gtzan/_validation\"\n",
" dir_testfolder = \"./LSTM-Music-Genre-Classification/gtzan/_test\"\n",
" dir_all_files = \"./LSTM-Music-Genre-Classification/gtzan\"\n",
"\n",
" train_X_preprocessed_data = \"./LSTM-Music-Genre-Classification/gtzan/data_train_input.npy\"\n",
" train_Y_preprocessed_data = \"./LSTM-Music-Genre-Classification/gtzan/data_train_target.npy\"\n",
" dev_X_preprocessed_data = \"./LSTM-Music-Genre-Classification/gtzan/data_validation_input.npy\"\n",
" dev_Y_preprocessed_data = \"./LSTM-Music-Genre-Classification/gtzan/data_validation_target.npy\"\n",
" test_X_preprocessed_data = \"./LSTM-Music-Genre-Classification/gtzan/data_test_input.npy\"\n",
" test_Y_preprocessed_data = \"./LSTM-Music-Genre-Classification/gtzan/data_test_target.npy\"\n",
"\n",
" train_X = train_Y = None\n",
" dev_X = dev_Y = None\n",
" test_X = test_Y = None\n",
"\n",
" def __init__(self):\n",
" self.hop_length = 512\n",
"\n",
" self.timeseries_length_list = []\n",
" self.trainfiles_list = self.path_to_audiofiles(self.dir_trainfolder)\n",
" self.devfiles_list = self.path_to_audiofiles(self.dir_devfolder)\n",
" self.testfiles_list = self.path_to_audiofiles(self.dir_testfolder)\n",
"\n",
" self.all_files_list = []\n",
" self.all_files_list.extend(self.trainfiles_list)\n",
" self.all_files_list.extend(self.devfiles_list)\n",
" self.all_files_list.extend(self.testfiles_list)\n",
"\n",
" # compute minimum timeseries length, slow to compute, caching pre-computed value of 1290\n",
" # self.precompute_min_timeseries_len()\n",
" # print(\"min(self.timeseries_length_list) ==\" + str(min(self.timeseries_length_list)))\n",
" # self.timeseries_length = min(self.timeseries_length_list)\n",
"\n",
" self.timeseries_length = (\n",
" 128\n",
" ) # sequence length == 128, default fftsize == 2048 & hop == 512 @ SR of 22050\n",
" # equals 128 overlapped windows that cover approx ~3.065 seconds of audio, which is a bit small!\n",
"\n",
" def load_preprocess_data(self):\n",
" print(\"[DEBUG] total number of files: \" + str(len(self.timeseries_length_list)))\n",
"\n",
" # Training set\n",
" self.train_X, self.train_Y = self.extract_audio_features(self.trainfiles_list)\n",
" with open(self.train_X_preprocessed_data, \"wb\") as f:\n",
" np.save(f, self.train_X)\n",
" with open(self.train_Y_preprocessed_data, \"wb\") as f:\n",
" self.train_Y = self.one_hot(self.train_Y)\n",
" np.save(f, self.train_Y)\n",
"\n",
" # Validation set\n",
" self.dev_X, self.dev_Y = self.extract_audio_features(self.devfiles_list)\n",
" with open(self.dev_X_preprocessed_data, \"wb\") as f:\n",
" np.save(f, self.dev_X)\n",
" with open(self.dev_Y_preprocessed_data, \"wb\") as f:\n",
" self.dev_Y = self.one_hot(self.dev_Y)\n",
" np.save(f, self.dev_Y)\n",
"\n",
" # Test set\n",
" self.test_X, self.test_Y = self.extract_audio_features(self.testfiles_list)\n",
" with open(self.test_X_preprocessed_data, \"wb\") as f:\n",
" np.save(f, self.test_X)\n",
" with open(self.test_Y_preprocessed_data, \"wb\") as f:\n",
" self.test_Y = self.one_hot(self.test_Y)\n",
" np.save(f, self.test_Y)\n",
"\n",
" def load_deserialize_data(self):\n",
"\n",
" self.train_X = np.load(self.train_X_preprocessed_data)\n",
" self.train_Y = np.load(self.train_Y_preprocessed_data)\n",
"\n",
" self.dev_X = np.load(self.dev_X_preprocessed_data)\n",
" self.dev_Y = np.load(self.dev_Y_preprocessed_data)\n",
"\n",
" self.test_X = np.load(self.test_X_preprocessed_data)\n",
" self.test_Y = np.load(self.test_Y_preprocessed_data)\n",
"\n",
" def precompute_min_timeseries_len(self):\n",
" for file in self.all_files_list:\n",
" print(\"Loading \" + str(file))\n",
" y, sr = librosa.load(file)\n",
" self.timeseries_length_list.append(math.ceil(len(y) / self.hop_length))\n",
"\n",
" def extract_audio_features(self, list_of_audiofiles):\n",
"\n",
" data = np.zeros(\n",
" (len(list_of_audiofiles), self.timeseries_length, 33), dtype=np.float64\n",
" )\n",
" target = []\n",
"\n",
" for i, file in enumerate(list_of_audiofiles):\n",
" y, sr = librosa.load(file)\n",
" mfcc = librosa.feature.mfcc(\n",
" y=y, sr=sr, hop_length=self.hop_length, n_mfcc=13\n",
" )\n",
" spectral_center = librosa.feature.spectral_centroid(\n",
" y=y, sr=sr, hop_length=self.hop_length\n",
" )\n",
" chroma = librosa.feature.chroma_stft(y=y, sr=sr, hop_length=self.hop_length)\n",
" spectral_contrast = librosa.feature.spectral_contrast(\n",
" y=y, sr=sr, hop_length=self.hop_length\n",
" )\n",
"\n",
" splits = re.split(\"[ .]\", file)\n",
" genre = re.split(\"[ /]\", splits[1])[3]\n",
" target.append(genre)\n",
"\n",
" data[i, :, 0:13] = mfcc.T[0:self.timeseries_length, :]\n",
" data[i, :, 13:14] = spectral_center.T[0:self.timeseries_length, :]\n",
" data[i, :, 14:26] = chroma.T[0:self.timeseries_length, :]\n",
" data[i, :, 26:33] = spectral_contrast.T[0:self.timeseries_length, :]\n",
"\n",
" print(\n",
" \"Extracted features audio track %i of %i.\"\n",
" % (i + 1, len(list_of_audiofiles))\n",
" )\n",
"\n",
" return data, np.expand_dims(np.asarray(target), axis=1)\n",
"\n",
" def one_hot(self, Y_genre_strings):\n",
" y_one_hot = np.zeros((Y_genre_strings.shape[0], len(self.genre_list)))\n",
" for i, genre_string in enumerate(Y_genre_strings):\n",
" index = self.genre_list.index(genre_string)\n",
" y_one_hot[i, index] = 1\n",
" return y_one_hot\n",
"\n",
" @staticmethod\n",
" def path_to_audiofiles(dir_folder):\n",
" list_of_audio = []\n",
" for file in os.listdir(dir_folder):\n",
" if file.endswith(\".au\"):\n",
" directory = \"%s/%s\" % (dir_folder, file)\n",
" list_of_audio.append(directory)\n",
" return list_of_audio"
],
"execution_count": 9,
"outputs": []
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "VCXg6JS1MWJx",
"outputId": "a263784e-42f4-4d96-d5e5-7be82cd3739a"
},
"source": [
"import logging\n",
"import os\n",
"from keras.models import Sequential\n",
"#from keras.layers.recurrent import LSTM\n",
"from keras.layers import LSTM\n",
"from keras.layers import Dense\n",
"from keras.optimizers import Adam\n",
"from keras.layers import Bidirectional\n",
"\n",
"#from GenreFeatureData import (\n",
"# GenreFeatureData,\n",
"#) # local python class with Audio feature extraction (librosa)\n",
"\n",
"# set logging level\n",
"logging.getLogger(\"tensorflow\").setLevel(logging.ERROR)\n",
"\n",
"genre_features = GenreFeatureData()\n",
"\n",
"# if all of the preprocessed files do not exist, regenerate them all for self-consistency\n",
"if (\n",
" os.path.isfile(genre_features.train_X_preprocessed_data)\n",
" and os.path.isfile(genre_features.train_Y_preprocessed_data)\n",
" and os.path.isfile(genre_features.dev_X_preprocessed_data)\n",
" and os.path.isfile(genre_features.dev_Y_preprocessed_data)\n",
" and os.path.isfile(genre_features.test_X_preprocessed_data)\n",
" and os.path.isfile(genre_features.test_Y_preprocessed_data)\n",
"):\n",
" print(\"Preprocessed files exist, deserializing npy files\")\n",
" genre_features.load_deserialize_data()\n",
"else:\n",
" print(\"Preprocessing raw audio files\")\n",
" genre_features.load_preprocess_data()\n",
"\n",
"print(\"Training X shape: \" + str(genre_features.train_X.shape))\n",
"print(\"Training Y shape: \" + str(genre_features.train_Y.shape))\n",
"print(\"Dev X shape: \" + str(genre_features.dev_X.shape))\n",
"print(\"Dev Y shape: \" + str(genre_features.dev_Y.shape))\n",
"print(\"Test X shape: \" + str(genre_features.test_X.shape))\n",
"print(\"Test Y shape: \" + str(genre_features.test_Y.shape))\n",
"\n",
"input_shape = (genre_features.train_X.shape[1], genre_features.train_X.shape[2])\n",
"print(\"Build LSTM RNN model ...\")\n",
"model = Sequential()\n",
"\n",
"#forward_layer = LSTM(units=128, dropout=0.05, recurrent_dropout=0.35, return_sequences=True)\n",
"#backward_layer = LSTM(units=32, dropout=0.05, return_sequences=True, go_backwards=True)\n",
"#model.add(Bidirectional(forward_layer, backward_layer=backward_layer, input_shape=input_shape))\n",
"\n",
"model.add(Bidirectional(LSTM(units=128, dropout=0.05, recurrent_dropout=0.35, return_sequences=True, go_backwards=True), input_shape=input_shape))\n",
"model.add(Bidirectional(LSTM(units=32, dropout=0.05, recurrent_dropout=0.35, return_sequences=False, go_backwards=True)))\n",
"model.add(Dense(units=genre_features.train_Y.shape[1], activation=\"sigmoid\"))\n",
"\n",
"#model.add(LSTM(units=128, dropout=0.05, recurrent_dropout=0.35, return_sequences=True, input_shape=input_shape))\n",
"#model.add(LSTM(units=32, dropout=0.05, recurrent_dropout=0.35, return_sequences=False))\n",
"#model.add(Dense(units=genre_features.train_Y.shape[1], activation=\"softmax\"))\n",
"\n",
"print(\"Compiling ...\")\n",
"# Keras optimizer defaults:\n",
"# Adam : lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-8, decay=0.\n",
"# RMSprop: lr=0.001, rho=0.9, epsilon=1e-8, decay=0.\n",
"# SGD : lr=0.01, momentum=0., decay=0.\n",
"opt = Adam()\n",
"#model.compile(loss=\"categorical_crossentropy\", optimizer=opt, metrics=[\"accuracy\"])\n",
"model.compile(loss=\"binary_crossentropy\", optimizer=opt, metrics=[\"accuracy\"])\n",
"model.summary()\n",
"\n",
"print(\"Training ...\")\n",
"batch_size = 35 # num of training examples per minibatch\n",
"num_epochs = 400\n",
"model.fit(\n",
" genre_features.train_X,\n",
" genre_features.train_Y,\n",
" batch_size=batch_size,\n",
" epochs=num_epochs,\n",
")\n",
"\n",
"print(\"\\nValidating ...\")\n",
"score, accuracy = model.evaluate(\n",
" genre_features.dev_X, genre_features.dev_Y, batch_size=batch_size, verbose=1\n",
")\n",
"print(\"Dev loss: \", score)\n",
"print(\"Dev accuracy: \", accuracy)\n",
"\n",
"\n",
"print(\"\\nTesting ...\")\n",
"score, accuracy = model.evaluate(\n",
" genre_features.test_X, genre_features.test_Y, batch_size=batch_size, verbose=1\n",
")\n",
"print(\"Test loss: \", score)\n",
"print(\"Test accuracy: \", accuracy)\n",
"\n",
"# serialize model to JSON\n",
"json_filename = \"model.json\"\n",
"print(\"\\nSaving model to JSON: \" + json_filename)\n",
"model_json = model.to_json()\n",
"with open(json_filename, \"w\") as json_file:\n",
" json_file.write(model_json)\n",
"\n",
"# Creates a HDF5 file 'lstm_genre_classifier.h5'\n",
"model_filename = \"lstm_genre_classifier_lstm.h5\"\n",
"print(\"\\nSaving model: \" + model_filename)\n",
"model.save(model_filename)"
],
"execution_count": 15,
"outputs": [
{
"output_type": "stream",
"text": [
"Preprocessed files exist, deserializing npy files\n",
"Training X shape: (420, 128, 33)\n",
"Training Y shape: (420, 8)\n",
"Dev X shape: (120, 128, 33)\n",
"Dev Y shape: (120, 8)\n",
"Test X shape: (60, 128, 33)\n",
"Test Y shape: (60, 8)\n",
"Build LSTM RNN model ...\n",
"Compiling ...\n",
"Model: \"sequential_1\"\n",
"_________________________________________________________________\n",
"Layer (type) Output Shape Param # \n",
"=================================================================\n",
"bidirectional_2 (Bidirection (None, 128, 256) 165888 \n",
"_________________________________________________________________\n",
"bidirectional_3 (Bidirection (None, 64) 73984 \n",
"_________________________________________________________________\n",
"dense_1 (Dense) (None, 8) 520 \n",
"=================================================================\n",
"Total params: 240,392\n",
"Trainable params: 240,392\n",
"Non-trainable params: 0\n",
"_________________________________________________________________\n",
"Training ...\n",
"Epoch 1/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.4737 - accuracy: 0.1524\n",
"Epoch 2/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.3457 - accuracy: 0.2405\n",
"Epoch 3/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.3232 - accuracy: 0.3857\n",
"Epoch 4/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.3014 - accuracy: 0.4500\n",
"Epoch 5/400\n",
"12/12 [==============================] - 9s 776ms/step - loss: 0.2846 - accuracy: 0.5000\n",
"Epoch 6/400\n",
"12/12 [==============================] - 10s 792ms/step - loss: 0.2727 - accuracy: 0.5048\n",
"Epoch 7/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.2687 - accuracy: 0.5405\n",
"Epoch 8/400\n",
"12/12 [==============================] - 9s 759ms/step - loss: 0.2600 - accuracy: 0.5333\n",
"Epoch 9/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.2447 - accuracy: 0.5524\n",
"Epoch 10/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.2347 - accuracy: 0.5810\n",
"Epoch 11/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.2262 - accuracy: 0.6143\n",
"Epoch 12/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.2202 - accuracy: 0.6024\n",
"Epoch 13/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.2251 - accuracy: 0.6071\n",
"Epoch 14/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.2157 - accuracy: 0.6143\n",
"Epoch 15/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.2121 - accuracy: 0.6143\n",
"Epoch 16/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.2093 - accuracy: 0.6429\n",
"Epoch 17/400\n",
"12/12 [==============================] - 12s 1s/step - loss: 0.2063 - accuracy: 0.6429\n",
"Epoch 18/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.2019 - accuracy: 0.6452\n",
"Epoch 19/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.2036 - accuracy: 0.6452\n",
"Epoch 20/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.1890 - accuracy: 0.6738\n",
"Epoch 21/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.1929 - accuracy: 0.6643\n",
"Epoch 22/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.1941 - accuracy: 0.6476\n",
"Epoch 23/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.1880 - accuracy: 0.6881\n",
"Epoch 24/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.1966 - accuracy: 0.6214\n",
"Epoch 25/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.1846 - accuracy: 0.6976\n",
"Epoch 26/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.1901 - accuracy: 0.6548\n",
"Epoch 27/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.1860 - accuracy: 0.6881\n",
"Epoch 28/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.1792 - accuracy: 0.7048\n",
"Epoch 29/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.1823 - accuracy: 0.6881\n",
"Epoch 30/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.1738 - accuracy: 0.7095\n",
"Epoch 31/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.1790 - accuracy: 0.7024\n",
"Epoch 32/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.1730 - accuracy: 0.7167\n",
"Epoch 33/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.1711 - accuracy: 0.7143\n",
"Epoch 34/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.1660 - accuracy: 0.7500\n",
"Epoch 35/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.1658 - accuracy: 0.7262\n",
"Epoch 36/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.1612 - accuracy: 0.7381\n",
"Epoch 37/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.1705 - accuracy: 0.7143\n",
"Epoch 38/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.1664 - accuracy: 0.7190\n",
"Epoch 39/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.1579 - accuracy: 0.7524\n",
"Epoch 40/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.1557 - accuracy: 0.7643\n",
"Epoch 41/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.1446 - accuracy: 0.7976\n",
"Epoch 42/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.1526 - accuracy: 0.7548\n",
"Epoch 43/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.1528 - accuracy: 0.7548\n",
"Epoch 44/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.1456 - accuracy: 0.7738\n",
"Epoch 45/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.1435 - accuracy: 0.7667\n",
"Epoch 46/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.1420 - accuracy: 0.7762\n",
"Epoch 47/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.1440 - accuracy: 0.7690\n",
"Epoch 48/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.1395 - accuracy: 0.8000\n",
"Epoch 49/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.1481 - accuracy: 0.7619\n",
"Epoch 50/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.1386 - accuracy: 0.8119\n",
"Epoch 51/400\n",
"12/12 [==============================] - 9s 783ms/step - loss: 0.1390 - accuracy: 0.7810\n",
"Epoch 52/400\n",
"12/12 [==============================] - 9s 780ms/step - loss: 0.1370 - accuracy: 0.7833\n",
"Epoch 53/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.1448 - accuracy: 0.7690\n",
"Epoch 54/400\n",
"12/12 [==============================] - 9s 784ms/step - loss: 0.1418 - accuracy: 0.7690\n",
"Epoch 55/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.1235 - accuracy: 0.8095\n",
"Epoch 56/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.1198 - accuracy: 0.8167\n",
"Epoch 57/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.1235 - accuracy: 0.7952\n",
"Epoch 58/400\n",
"12/12 [==============================] - 9s 778ms/step - loss: 0.1186 - accuracy: 0.8238\n",
"Epoch 59/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.1151 - accuracy: 0.8071\n",
"Epoch 60/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.1329 - accuracy: 0.7929\n",
"Epoch 61/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.1405 - accuracy: 0.7905\n",
"Epoch 62/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.1646 - accuracy: 0.7381\n",
"Epoch 63/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.1655 - accuracy: 0.7286\n",
"Epoch 64/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.1608 - accuracy: 0.7310\n",
"Epoch 65/400\n",
"12/12 [==============================] - 9s 787ms/step - loss: 0.1432 - accuracy: 0.7738\n",
"Epoch 66/400\n",
"12/12 [==============================] - 10s 794ms/step - loss: 0.1398 - accuracy: 0.7786\n",
"Epoch 67/400\n",
"12/12 [==============================] - 9s 789ms/step - loss: 0.1290 - accuracy: 0.7952\n",
"Epoch 68/400\n",
"12/12 [==============================] - 10s 830ms/step - loss: 0.1219 - accuracy: 0.8214\n",
"Epoch 69/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.1142 - accuracy: 0.8262\n",
"Epoch 70/400\n",
"12/12 [==============================] - 10s 830ms/step - loss: 0.1126 - accuracy: 0.8357\n",
"Epoch 71/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.1118 - accuracy: 0.8333\n",
"Epoch 72/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.1114 - accuracy: 0.8333\n",
"Epoch 73/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.1125 - accuracy: 0.8310\n",
"Epoch 74/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.1096 - accuracy: 0.8452\n",
"Epoch 75/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.1134 - accuracy: 0.8286\n",
"Epoch 76/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.1024 - accuracy: 0.8405\n",
"Epoch 77/400\n",
"12/12 [==============================] - 12s 1s/step - loss: 0.0995 - accuracy: 0.8667\n",
"Epoch 78/400\n",
"12/12 [==============================] - 9s 780ms/step - loss: 0.0975 - accuracy: 0.8524\n",
"Epoch 79/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.1149 - accuracy: 0.8071\n",
"Epoch 80/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.1072 - accuracy: 0.8238\n",
"Epoch 81/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0987 - accuracy: 0.8500\n",
"Epoch 82/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0935 - accuracy: 0.8738\n",
"Epoch 83/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0875 - accuracy: 0.8786\n",
"Epoch 84/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0951 - accuracy: 0.8738\n",
"Epoch 85/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0913 - accuracy: 0.8762\n",
"Epoch 86/400\n",
"12/12 [==============================] - 9s 776ms/step - loss: 0.1038 - accuracy: 0.8333\n",
"Epoch 87/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0989 - accuracy: 0.8429\n",
"Epoch 88/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.1012 - accuracy: 0.8452\n",
"Epoch 89/400\n",
"12/12 [==============================] - 9s 776ms/step - loss: 0.1001 - accuracy: 0.8262\n",
"Epoch 90/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0925 - accuracy: 0.8595\n",
"Epoch 91/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0948 - accuracy: 0.8619\n",
"Epoch 92/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0935 - accuracy: 0.8643\n",
"Epoch 93/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0894 - accuracy: 0.8643\n",
"Epoch 94/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0888 - accuracy: 0.8690\n",
"Epoch 95/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0884 - accuracy: 0.8714\n",
"Epoch 96/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0880 - accuracy: 0.8690\n",
"Epoch 97/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0840 - accuracy: 0.8786\n",
"Epoch 98/400\n",
"12/12 [==============================] - 9s 786ms/step - loss: 0.0836 - accuracy: 0.8762\n",
"Epoch 99/400\n",
"12/12 [==============================] - 9s 788ms/step - loss: 0.0906 - accuracy: 0.8571\n",
"Epoch 100/400\n",
"12/12 [==============================] - 10s 830ms/step - loss: 0.0841 - accuracy: 0.8762\n",
"Epoch 101/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0823 - accuracy: 0.8833\n",
"Epoch 102/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0863 - accuracy: 0.8643\n",
"Epoch 103/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0962 - accuracy: 0.8452\n",
"Epoch 104/400\n",
"12/12 [==============================] - 9s 759ms/step - loss: 0.0878 - accuracy: 0.8548\n",
"Epoch 105/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.1000 - accuracy: 0.8381\n",
"Epoch 106/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0963 - accuracy: 0.8643\n",
"Epoch 107/400\n",
"12/12 [==============================] - 9s 757ms/step - loss: 0.0934 - accuracy: 0.8595\n",
"Epoch 108/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0833 - accuracy: 0.8786\n",
"Epoch 109/400\n",
"12/12 [==============================] - 9s 759ms/step - loss: 0.0899 - accuracy: 0.8619\n",
"Epoch 110/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0825 - accuracy: 0.8667\n",
"Epoch 111/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0817 - accuracy: 0.8881\n",
"Epoch 112/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0739 - accuracy: 0.8786\n",
"Epoch 113/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0864 - accuracy: 0.8714\n",
"Epoch 114/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0813 - accuracy: 0.8643\n",
"Epoch 115/400\n",
"12/12 [==============================] - 9s 759ms/step - loss: 0.0760 - accuracy: 0.8738\n",
"Epoch 116/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0742 - accuracy: 0.8929\n",
"Epoch 117/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0880 - accuracy: 0.8690\n",
"Epoch 118/400\n",
"12/12 [==============================] - 9s 757ms/step - loss: 0.0828 - accuracy: 0.8833\n",
"Epoch 119/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0715 - accuracy: 0.9024\n",
"Epoch 120/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0740 - accuracy: 0.8881\n",
"Epoch 121/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0663 - accuracy: 0.9024\n",
"Epoch 122/400\n",
"12/12 [==============================] - 9s 778ms/step - loss: 0.0709 - accuracy: 0.9048\n",
"Epoch 123/400\n",
"12/12 [==============================] - 9s 782ms/step - loss: 0.0733 - accuracy: 0.9024\n",
"Epoch 124/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0736 - accuracy: 0.8881\n",
"Epoch 125/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0722 - accuracy: 0.8976\n",
"Epoch 126/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0751 - accuracy: 0.8976\n",
"Epoch 127/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0827 - accuracy: 0.8714\n",
"Epoch 128/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0675 - accuracy: 0.9048\n",
"Epoch 129/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0623 - accuracy: 0.9095\n",
"Epoch 130/400\n",
"12/12 [==============================] - 10s 802ms/step - loss: 0.0632 - accuracy: 0.9048\n",
"Epoch 131/400\n",
"12/12 [==============================] - 10s 866ms/step - loss: 0.0635 - accuracy: 0.9119\n",
"Epoch 132/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0759 - accuracy: 0.8833\n",
"Epoch 133/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0688 - accuracy: 0.8976\n",
"Epoch 134/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0660 - accuracy: 0.8976\n",
"Epoch 135/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0696 - accuracy: 0.9000\n",
"Epoch 136/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0679 - accuracy: 0.8929\n",
"Epoch 137/400\n",
"12/12 [==============================] - 12s 1s/step - loss: 0.0662 - accuracy: 0.9048\n",
"Epoch 138/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0712 - accuracy: 0.8881\n",
"Epoch 139/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0705 - accuracy: 0.9000\n",
"Epoch 140/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0708 - accuracy: 0.8976\n",
"Epoch 141/400\n",
"12/12 [==============================] - 9s 781ms/step - loss: 0.0849 - accuracy: 0.8643\n",
"Epoch 142/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0713 - accuracy: 0.8881\n",
"Epoch 143/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0662 - accuracy: 0.9048\n",
"Epoch 144/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0715 - accuracy: 0.8905\n",
"Epoch 145/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0668 - accuracy: 0.9024\n",
"Epoch 146/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0665 - accuracy: 0.9119\n",
"Epoch 147/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0579 - accuracy: 0.9333\n",
"Epoch 148/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0586 - accuracy: 0.9024\n",
"Epoch 149/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0599 - accuracy: 0.9071\n",
"Epoch 150/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0584 - accuracy: 0.9095\n",
"Epoch 151/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0510 - accuracy: 0.9381\n",
"Epoch 152/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0570 - accuracy: 0.9214\n",
"Epoch 153/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0562 - accuracy: 0.9190\n",
"Epoch 154/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0497 - accuracy: 0.9286\n",
"Epoch 155/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0551 - accuracy: 0.9119\n",
"Epoch 156/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0627 - accuracy: 0.9024\n",
"Epoch 157/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0540 - accuracy: 0.9238\n",
"Epoch 158/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0488 - accuracy: 0.9405\n",
"Epoch 159/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0456 - accuracy: 0.9405\n",
"Epoch 160/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0490 - accuracy: 0.9357\n",
"Epoch 161/400\n",
"12/12 [==============================] - 9s 785ms/step - loss: 0.0496 - accuracy: 0.9429\n",
"Epoch 162/400\n",
"12/12 [==============================] - 9s 782ms/step - loss: 0.0526 - accuracy: 0.9238\n",
"Epoch 163/400\n",
"12/12 [==============================] - 9s 781ms/step - loss: 0.0616 - accuracy: 0.9167\n",
"Epoch 164/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0659 - accuracy: 0.8976\n",
"Epoch 165/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0669 - accuracy: 0.9000\n",
"Epoch 166/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0599 - accuracy: 0.9024\n",
"Epoch 167/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0529 - accuracy: 0.9238\n",
"Epoch 168/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0535 - accuracy: 0.9214\n",
"Epoch 169/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0574 - accuracy: 0.9167\n",
"Epoch 170/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0551 - accuracy: 0.9143\n",
"Epoch 171/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0587 - accuracy: 0.9095\n",
"Epoch 172/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0532 - accuracy: 0.9405\n",
"Epoch 173/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0627 - accuracy: 0.9048\n",
"Epoch 174/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0539 - accuracy: 0.9214\n",
"Epoch 175/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0501 - accuracy: 0.9405\n",
"Epoch 176/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0538 - accuracy: 0.9143\n",
"Epoch 177/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0680 - accuracy: 0.9048\n",
"Epoch 178/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0480 - accuracy: 0.9310\n",
"Epoch 179/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0473 - accuracy: 0.9405\n",
"Epoch 180/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0489 - accuracy: 0.9310\n",
"Epoch 181/400\n",
"12/12 [==============================] - 9s 759ms/step - loss: 0.0450 - accuracy: 0.9429\n",
"Epoch 182/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0487 - accuracy: 0.9333\n",
"Epoch 183/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0434 - accuracy: 0.9452\n",
"Epoch 184/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0504 - accuracy: 0.9262\n",
"Epoch 185/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0422 - accuracy: 0.9452\n",
"Epoch 186/400\n",
"12/12 [==============================] - 9s 757ms/step - loss: 0.0472 - accuracy: 0.9286\n",
"Epoch 187/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0607 - accuracy: 0.8976\n",
"Epoch 188/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0489 - accuracy: 0.9357\n",
"Epoch 189/400\n",
"12/12 [==============================] - 9s 758ms/step - loss: 0.0509 - accuracy: 0.9286\n",
"Epoch 190/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0584 - accuracy: 0.9167\n",
"Epoch 191/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0459 - accuracy: 0.9429\n",
"Epoch 192/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0493 - accuracy: 0.9333\n",
"Epoch 193/400\n",
"12/12 [==============================] - 9s 786ms/step - loss: 0.0453 - accuracy: 0.9286\n",
"Epoch 194/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0445 - accuracy: 0.9429\n",
"Epoch 195/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0421 - accuracy: 0.9405\n",
"Epoch 196/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0476 - accuracy: 0.9262\n",
"Epoch 197/400\n",
"12/12 [==============================] - 12s 1s/step - loss: 0.0427 - accuracy: 0.9500\n",
"Epoch 198/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0447 - accuracy: 0.9310\n",
"Epoch 199/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0452 - accuracy: 0.9333\n",
"Epoch 200/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0426 - accuracy: 0.9476\n",
"Epoch 201/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0440 - accuracy: 0.9500\n",
"Epoch 202/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0539 - accuracy: 0.9190\n",
"Epoch 203/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0565 - accuracy: 0.9119\n",
"Epoch 204/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0490 - accuracy: 0.9381\n",
"Epoch 205/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0500 - accuracy: 0.9262\n",
"Epoch 206/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0440 - accuracy: 0.9310\n",
"Epoch 207/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0388 - accuracy: 0.9500\n",
"Epoch 208/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0329 - accuracy: 0.9548\n",
"Epoch 209/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0289 - accuracy: 0.9643\n",
"Epoch 210/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0432 - accuracy: 0.9381\n",
"Epoch 211/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0424 - accuracy: 0.9357\n",
"Epoch 212/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0496 - accuracy: 0.9214\n",
"Epoch 213/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0459 - accuracy: 0.9310\n",
"Epoch 214/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0353 - accuracy: 0.9524\n",
"Epoch 215/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0382 - accuracy: 0.9452\n",
"Epoch 216/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0383 - accuracy: 0.9476\n",
"Epoch 217/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0384 - accuracy: 0.9452\n",
"Epoch 218/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0343 - accuracy: 0.9619\n",
"Epoch 219/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0289 - accuracy: 0.9619\n",
"Epoch 220/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0293 - accuracy: 0.9619\n",
"Epoch 221/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0271 - accuracy: 0.9667\n",
"Epoch 222/400\n",
"12/12 [==============================] - 10s 801ms/step - loss: 0.0334 - accuracy: 0.9571\n",
"Epoch 223/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0329 - accuracy: 0.9548\n",
"Epoch 224/400\n",
"12/12 [==============================] - 10s 795ms/step - loss: 0.0322 - accuracy: 0.9595\n",
"Epoch 225/400\n",
"12/12 [==============================] - 10s 805ms/step - loss: 0.0260 - accuracy: 0.9690\n",
"Epoch 226/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0380 - accuracy: 0.9548\n",
"Epoch 227/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0628 - accuracy: 0.9071\n",
"Epoch 228/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0455 - accuracy: 0.9310\n",
"Epoch 229/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0322 - accuracy: 0.9643\n",
"Epoch 230/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0323 - accuracy: 0.9667\n",
"Epoch 231/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0358 - accuracy: 0.9548\n",
"Epoch 232/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0327 - accuracy: 0.9524\n",
"Epoch 233/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0345 - accuracy: 0.9524\n",
"Epoch 234/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0270 - accuracy: 0.9619\n",
"Epoch 235/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0355 - accuracy: 0.9571\n",
"Epoch 236/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0350 - accuracy: 0.9500\n",
"Epoch 237/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0390 - accuracy: 0.9405\n",
"Epoch 238/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0402 - accuracy: 0.9548\n",
"Epoch 239/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0351 - accuracy: 0.9548\n",
"Epoch 240/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0331 - accuracy: 0.9524\n",
"Epoch 241/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0337 - accuracy: 0.9548\n",
"Epoch 242/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0369 - accuracy: 0.9524\n",
"Epoch 243/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.0550 - accuracy: 0.9214\n",
"Epoch 244/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0461 - accuracy: 0.9405\n",
"Epoch 245/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0395 - accuracy: 0.9524\n",
"Epoch 246/400\n",
"12/12 [==============================] - 9s 756ms/step - loss: 0.0381 - accuracy: 0.9476\n",
"Epoch 247/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0640 - accuracy: 0.9048\n",
"Epoch 248/400\n",
"12/12 [==============================] - 9s 760ms/step - loss: 0.0383 - accuracy: 0.9524\n",
"Epoch 249/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0473 - accuracy: 0.9357\n",
"Epoch 250/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0431 - accuracy: 0.9381\n",
"Epoch 251/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0331 - accuracy: 0.9690\n",
"Epoch 252/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0279 - accuracy: 0.9643\n",
"Epoch 253/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0292 - accuracy: 0.9595\n",
"Epoch 254/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0272 - accuracy: 0.9690\n",
"Epoch 255/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0318 - accuracy: 0.9595\n",
"Epoch 256/400\n",
"12/12 [==============================] - 9s 789ms/step - loss: 0.0273 - accuracy: 0.9643\n",
"Epoch 257/400\n",
"12/12 [==============================] - 12s 997ms/step - loss: 0.0266 - accuracy: 0.9643\n",
"Epoch 258/400\n",
"12/12 [==============================] - 10s 852ms/step - loss: 0.0308 - accuracy: 0.9714\n",
"Epoch 259/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0383 - accuracy: 0.9476\n",
"Epoch 260/400\n",
"12/12 [==============================] - 9s 780ms/step - loss: 0.0328 - accuracy: 0.9452\n",
"Epoch 261/400\n",
"12/12 [==============================] - 9s 776ms/step - loss: 0.0458 - accuracy: 0.9405\n",
"Epoch 262/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0350 - accuracy: 0.9619\n",
"Epoch 263/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0392 - accuracy: 0.9500\n",
"Epoch 264/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0265 - accuracy: 0.9643\n",
"Epoch 265/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0277 - accuracy: 0.9690\n",
"Epoch 266/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0293 - accuracy: 0.9667\n",
"Epoch 267/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0253 - accuracy: 0.9667\n",
"Epoch 268/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0269 - accuracy: 0.9643\n",
"Epoch 269/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0275 - accuracy: 0.9548\n",
"Epoch 270/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0214 - accuracy: 0.9786\n",
"Epoch 271/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0240 - accuracy: 0.9667\n",
"Epoch 272/400\n",
"12/12 [==============================] - 9s 783ms/step - loss: 0.0383 - accuracy: 0.9476\n",
"Epoch 273/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0252 - accuracy: 0.9667\n",
"Epoch 274/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0300 - accuracy: 0.9619\n",
"Epoch 275/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0469 - accuracy: 0.9452\n",
"Epoch 276/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0543 - accuracy: 0.9310\n",
"Epoch 277/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0485 - accuracy: 0.9167\n",
"Epoch 278/400\n",
"12/12 [==============================] - 9s 776ms/step - loss: 0.0320 - accuracy: 0.9643\n",
"Epoch 279/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0282 - accuracy: 0.9643\n",
"Epoch 280/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0246 - accuracy: 0.9667\n",
"Epoch 281/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0277 - accuracy: 0.9595\n",
"Epoch 282/400\n",
"12/12 [==============================] - 9s 787ms/step - loss: 0.0335 - accuracy: 0.9524\n",
"Epoch 283/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0254 - accuracy: 0.9643\n",
"Epoch 284/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0252 - accuracy: 0.9667\n",
"Epoch 285/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0289 - accuracy: 0.9690\n",
"Epoch 286/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0253 - accuracy: 0.9738\n",
"Epoch 287/400\n",
"12/12 [==============================] - 9s 784ms/step - loss: 0.0266 - accuracy: 0.9667\n",
"Epoch 288/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0292 - accuracy: 0.9643\n",
"Epoch 289/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0273 - accuracy: 0.9762\n",
"Epoch 290/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0276 - accuracy: 0.9690\n",
"Epoch 291/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0272 - accuracy: 0.9619\n",
"Epoch 292/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0382 - accuracy: 0.9500\n",
"Epoch 293/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0354 - accuracy: 0.9524\n",
"Epoch 294/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0294 - accuracy: 0.9548\n",
"Epoch 295/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0235 - accuracy: 0.9714\n",
"Epoch 296/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0233 - accuracy: 0.9667\n",
"Epoch 297/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0226 - accuracy: 0.9690\n",
"Epoch 298/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0238 - accuracy: 0.9738\n",
"Epoch 299/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0203 - accuracy: 0.9762\n",
"Epoch 300/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0162 - accuracy: 0.9786\n",
"Epoch 301/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0193 - accuracy: 0.9762\n",
"Epoch 302/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0211 - accuracy: 0.9690\n",
"Epoch 303/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0204 - accuracy: 0.9786\n",
"Epoch 304/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0209 - accuracy: 0.9690\n",
"Epoch 305/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0281 - accuracy: 0.9667\n",
"Epoch 306/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0215 - accuracy: 0.9786\n",
"Epoch 307/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0195 - accuracy: 0.9762\n",
"Epoch 308/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0182 - accuracy: 0.9857\n",
"Epoch 309/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0197 - accuracy: 0.9738\n",
"Epoch 310/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0163 - accuracy: 0.9810\n",
"Epoch 311/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0199 - accuracy: 0.9738\n",
"Epoch 312/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0284 - accuracy: 0.9619\n",
"Epoch 313/400\n",
"12/12 [==============================] - 9s 781ms/step - loss: 0.0651 - accuracy: 0.8833\n",
"Epoch 314/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0402 - accuracy: 0.9452\n",
"Epoch 315/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0388 - accuracy: 0.9524\n",
"Epoch 316/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0333 - accuracy: 0.9524\n",
"Epoch 317/400\n",
"12/12 [==============================] - 12s 1s/step - loss: 0.0235 - accuracy: 0.9738\n",
"Epoch 318/400\n",
"12/12 [==============================] - 10s 799ms/step - loss: 0.0181 - accuracy: 0.9929\n",
"Epoch 319/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0309 - accuracy: 0.9690\n",
"Epoch 320/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0268 - accuracy: 0.9619\n",
"Epoch 321/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0282 - accuracy: 0.9643\n",
"Epoch 322/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0251 - accuracy: 0.9667\n",
"Epoch 323/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0256 - accuracy: 0.9643\n",
"Epoch 324/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0174 - accuracy: 0.9810\n",
"Epoch 325/400\n",
"12/12 [==============================] - 9s 778ms/step - loss: 0.0216 - accuracy: 0.9762\n",
"Epoch 326/400\n",
"12/12 [==============================] - 10s 869ms/step - loss: 0.0262 - accuracy: 0.9786\n",
"Epoch 327/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0229 - accuracy: 0.9738\n",
"Epoch 328/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0228 - accuracy: 0.9810\n",
"Epoch 329/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0209 - accuracy: 0.9714\n",
"Epoch 330/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0204 - accuracy: 0.9762\n",
"Epoch 331/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0169 - accuracy: 0.9833\n",
"Epoch 332/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0174 - accuracy: 0.9833\n",
"Epoch 333/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0234 - accuracy: 0.9714\n",
"Epoch 334/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0272 - accuracy: 0.9714\n",
"Epoch 335/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0222 - accuracy: 0.9714\n",
"Epoch 336/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0216 - accuracy: 0.9762\n",
"Epoch 337/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0203 - accuracy: 0.9714\n",
"Epoch 338/400\n",
"12/12 [==============================] - 9s 768ms/step - loss: 0.0159 - accuracy: 0.9833\n",
"Epoch 339/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0161 - accuracy: 0.9833\n",
"Epoch 340/400\n",
"12/12 [==============================] - 9s 770ms/step - loss: 0.0172 - accuracy: 0.9786\n",
"Epoch 341/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0189 - accuracy: 0.9738\n",
"Epoch 342/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0190 - accuracy: 0.9833\n",
"Epoch 343/400\n",
"12/12 [==============================] - 9s 789ms/step - loss: 0.0177 - accuracy: 0.9762\n",
"Epoch 344/400\n",
"12/12 [==============================] - 9s 762ms/step - loss: 0.0230 - accuracy: 0.9667\n",
"Epoch 345/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0186 - accuracy: 0.9714\n",
"Epoch 346/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0161 - accuracy: 0.9786\n",
"Epoch 347/400\n",
"12/12 [==============================] - 9s 758ms/step - loss: 0.0159 - accuracy: 0.9786\n",
"Epoch 348/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0173 - accuracy: 0.9762\n",
"Epoch 349/400\n",
"12/12 [==============================] - 9s 792ms/step - loss: 0.0245 - accuracy: 0.9690\n",
"Epoch 350/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.0186 - accuracy: 0.9786\n",
"Epoch 351/400\n",
"12/12 [==============================] - 9s 767ms/step - loss: 0.0216 - accuracy: 0.9738\n",
"Epoch 352/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0205 - accuracy: 0.9786\n",
"Epoch 353/400\n",
"12/12 [==============================] - 9s 764ms/step - loss: 0.0217 - accuracy: 0.9762\n",
"Epoch 354/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0203 - accuracy: 0.9762\n",
"Epoch 355/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0153 - accuracy: 0.9833\n",
"Epoch 356/400\n",
"12/12 [==============================] - 9s 766ms/step - loss: 0.0179 - accuracy: 0.9786\n",
"Epoch 357/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0150 - accuracy: 0.9786\n",
"Epoch 358/400\n",
"12/12 [==============================] - 9s 761ms/step - loss: 0.0230 - accuracy: 0.9690\n",
"Epoch 359/400\n",
"12/12 [==============================] - 9s 763ms/step - loss: 0.0173 - accuracy: 0.9762\n",
"Epoch 360/400\n",
"12/12 [==============================] - 9s 765ms/step - loss: 0.0282 - accuracy: 0.9595\n",
"Epoch 361/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0226 - accuracy: 0.9786\n",
"Epoch 362/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0205 - accuracy: 0.9762\n",
"Epoch 363/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0161 - accuracy: 0.9786\n",
"Epoch 364/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0204 - accuracy: 0.9738\n",
"Epoch 365/400\n",
"12/12 [==============================] - 9s 780ms/step - loss: 0.0149 - accuracy: 0.9857\n",
"Epoch 366/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.0176 - accuracy: 0.9762\n",
"Epoch 367/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0337 - accuracy: 0.9619\n",
"Epoch 368/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0452 - accuracy: 0.9405\n",
"Epoch 369/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.0597 - accuracy: 0.9048\n",
"Epoch 370/400\n",
"12/12 [==============================] - 9s 783ms/step - loss: 0.0492 - accuracy: 0.9310\n",
"Epoch 371/400\n",
"12/12 [==============================] - 10s 793ms/step - loss: 0.0415 - accuracy: 0.9429\n",
"Epoch 372/400\n",
"12/12 [==============================] - 9s 778ms/step - loss: 0.0243 - accuracy: 0.9619\n",
"Epoch 373/400\n",
"12/12 [==============================] - 10s 842ms/step - loss: 0.0148 - accuracy: 0.9881\n",
"Epoch 374/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0185 - accuracy: 0.9833\n",
"Epoch 375/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0228 - accuracy: 0.9643\n",
"Epoch 376/400\n",
"12/12 [==============================] - 9s 778ms/step - loss: 0.0186 - accuracy: 0.9762\n",
"Epoch 377/400\n",
"12/12 [==============================] - 14s 1s/step - loss: 0.0246 - accuracy: 0.9667\n",
"Epoch 378/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.0220 - accuracy: 0.9762\n",
"Epoch 379/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0206 - accuracy: 0.9738\n",
"Epoch 380/400\n",
"12/12 [==============================] - 10s 822ms/step - loss: 0.0145 - accuracy: 0.9905\n",
"Epoch 381/400\n",
"12/12 [==============================] - 10s 844ms/step - loss: 0.0188 - accuracy: 0.9786\n",
"Epoch 382/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0191 - accuracy: 0.9738\n",
"Epoch 383/400\n",
"12/12 [==============================] - 9s 779ms/step - loss: 0.0210 - accuracy: 0.9786\n",
"Epoch 384/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0193 - accuracy: 0.9714\n",
"Epoch 385/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0182 - accuracy: 0.9833\n",
"Epoch 386/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0167 - accuracy: 0.9738\n",
"Epoch 387/400\n",
"12/12 [==============================] - 9s 781ms/step - loss: 0.0150 - accuracy: 0.9810\n",
"Epoch 388/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0182 - accuracy: 0.9738\n",
"Epoch 389/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0140 - accuracy: 0.9786\n",
"Epoch 390/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0141 - accuracy: 0.9881\n",
"Epoch 391/400\n",
"12/12 [==============================] - 9s 775ms/step - loss: 0.0195 - accuracy: 0.9762\n",
"Epoch 392/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0179 - accuracy: 0.9786\n",
"Epoch 393/400\n",
"12/12 [==============================] - 9s 773ms/step - loss: 0.0165 - accuracy: 0.9786\n",
"Epoch 394/400\n",
"12/12 [==============================] - 9s 769ms/step - loss: 0.0135 - accuracy: 0.9881\n",
"Epoch 395/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0178 - accuracy: 0.9738\n",
"Epoch 396/400\n",
"12/12 [==============================] - 9s 771ms/step - loss: 0.0151 - accuracy: 0.9929\n",
"Epoch 397/400\n",
"12/12 [==============================] - 9s 774ms/step - loss: 0.0142 - accuracy: 0.9857\n",
"Epoch 398/400\n",
"12/12 [==============================] - 9s 772ms/step - loss: 0.0147 - accuracy: 0.9833\n",
"Epoch 399/400\n",
"12/12 [==============================] - 9s 776ms/step - loss: 0.0194 - accuracy: 0.9810\n",
"Epoch 400/400\n",
"12/12 [==============================] - 9s 777ms/step - loss: 0.0208 - accuracy: 0.9786\n",
"\n",
"Validating ...\n",
"4/4 [==============================] - 0s 77ms/step - loss: 0.2979 - accuracy: 0.6667\n",
"Dev loss: 0.2978794276714325\n",
"Dev accuracy: 0.6666666865348816\n",
"\n",
"Testing ...\n",
"2/2 [==============================] - 0s 49ms/step - loss: 0.3714 - accuracy: 0.6000\n",
"Test loss: 0.37135106325149536\n",
"Test accuracy: 0.6000000238418579\n",
"\n",
"Saving model to JSON: model.json\n",
"\n",
"Saving model: lstm_genre_classifier_lstm.h5\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "_IRUgFR4eioe",
"outputId": "00aec362-a5e7-438f-a892-0984b1daf2b9"
},
"source": [
"!ls"
],
"execution_count": 16,
"outputs": [
{
"output_type": "stream",
"text": [
"lstm_genre_classifier_lstm.h5\t model.json\n",
"LSTM-Music-Genre-Classification sample_data\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "ObxYdoIqejWM",
"outputId": "691de971-3f77-4af8-cf7b-591f6d1c88fc"
},
"source": [
"\"An example of predicting a music genre from a custom audio file\"\n",
"import librosa\n",
"import logging\n",
"import sys\n",
"import numpy as np\n",
"\n",
"from keras.models import model_from_json\n",
"#from GenreFeatureData import (\n",
"# GenreFeatureData,\n",
"#) # local python class with Audio feature extraction and genre list\n",
"\n",
"# set logging level\n",
"logging.getLogger(\"tensorflow\").setLevel(logging.ERROR)\n",
"\n",
"\n",
"def load_model(model_path, weights_path):\n",
" \"Load the trained LSTM model from directory for genre classification\"\n",
" with open(model_path, \"r\") as model_file:\n",
" trained_model = model_from_json(model_file.read())\n",
" trained_model.load_weights(weights_path)\n",
" trained_model.compile(\n",
" loss=\"categorical_crossentropy\", optimizer=\"adam\", metrics=[\"accuracy\"]\n",
" )\n",
" return trained_model\n",
"\n",
"\n",
"def extract_audio_features(file):\n",
" \"Extract audio features from an audio file for genre classification\"\n",
" timeseries_length = 128\n",
" features = np.zeros((1, timeseries_length, 33), dtype=np.float64)\n",
"\n",
" y, sr = librosa.load(file)\n",
" mfcc = librosa.feature.mfcc(y=y, sr=sr, hop_length=512, n_mfcc=13)\n",
" spectral_center = librosa.feature.spectral_centroid(y=y, sr=sr, hop_length=512)\n",
" chroma = librosa.feature.chroma_stft(y=y, sr=sr, hop_length=512)\n",
" spectral_contrast = librosa.feature.spectral_contrast(y=y, sr=sr, hop_length=512)\n",
"\n",
" features[0, :, 0:13] = mfcc.T[0:timeseries_length, :]\n",
" features[0, :, 13:14] = spectral_center.T[0:timeseries_length, :]\n",
" features[0, :, 14:26] = chroma.T[0:timeseries_length, :]\n",
" features[0, :, 26:33] = spectral_contrast.T[0:timeseries_length, :]\n",
" return features\n",
"\n",
"\n",
"def get_genre(model, music_path):\n",
" \"Predict genre of music using a trained model\"\n",
" prediction = model.predict(extract_audio_features(music_path))\n",
" predict_genre = GenreFeatureData().genre_list[np.argmax(prediction)]\n",
" return predict_genre\n",
"\n",
"\n",
"if __name__ == \"__main__\":\n",
" PATH = \"./LSTM-Music-Genre-Classification/audio/country_music.mp3\"\n",
" #MODEL = load_model(\"./model.json\", \"./weights/model_weights.h5\")\n",
" M_PATH = \"./model.json\"\n",
" W_PATH = \"./lstm_genre_classifier_lstm.h5\"\n",
" MODEL = load_model(M_PATH, W_PATH)\n",
" GENRE = get_genre(MODEL, PATH)\n",
" print(\"Model predict: {}\".format(GENRE))"
],
"execution_count": 21,
"outputs": [
{
"output_type": "stream",
"text": [
"Model predict: metal\n"
],
"name": "stdout"
}
]
},
{
"cell_type": "code",
"metadata": {
"id": "A9UlGJtNfdlG"
},
"source": [
""
],
"execution_count": null,
"outputs": []
}
]
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment