multi-dodality on gujrati digits.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Ensembling on gujrati digits.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyPruy0JKRq+1wRnWugieOjn",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/turab45/Audd-multi-modality-for-urdu-audio/blob/master/multi-dodality%20on%20gujrati%20digits.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UHZ4fwe2fljG",
        "outputId": "00dcae4a-0e03-4d9a-9bb5-c50000361ceb"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ],
      "source": [
        "import matplotlib.pyplot as plt\n",
        "from scipy import signal\n",
        "from scipy.io import wavfile\n",
        "import os\n",
        "import numpy as np\n",
        "import librosa\n",
        "import librosa.display\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "import cv2\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")\n",
        "\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "!git clone https://github.com/Nikunj1729/free-spoken-gujarati-digit-dataset"
      ],
      "metadata": {
        "id": "_A3lLsYjf3Cu"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "cd free-spoken-gujarati-digit-dataset/"
      ],
      "metadata": {
        "id": "XU4rrZFrf3-7"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Feature extraction methods\n",
        "\n",
        "def CalculateZeroCrossingRate(file_location):\n",
        "    y, sr = librosa.load(file_location)\n",
        "    zCross = librosa.feature.zero_crossing_rate(y=y)\n",
        "    dim = (32, 32)\n",
        "    resized = cv2.resize(zCross, dim, interpolation = cv2.INTER_AREA)\n",
        "    return resized\n",
        "\n",
        "def CalculateMelSpectrogram(file_location):\n",
        "    y, sr = librosa.load(file_location)\n",
        "    melSpec = librosa.feature.melspectrogram(y=y, sr=sr)\n",
        "    melSpec_dB = librosa.power_to_db(melSpec)\n",
        "    dim = (32, 32)\n",
        "    resized = cv2.resize(melSpec_dB, dim, interpolation = cv2.INTER_AREA)\n",
        "    return resized\n",
        "\n",
        "\n",
        "def CalculateMFCC(file_location):\n",
        "    y, sr = librosa.load(file_location)\n",
        "    melSpec = librosa.feature.mfcc(y=y, sr=sr)\n",
        "    dim = (32, 32)\n",
        "    resized = cv2.resize(melSpec, dim, interpolation = cv2.INTER_AREA)\n",
        "    return resized"
      ],
      "metadata": {
        "id": "G9S0AOOVf-VG"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "files_path = '/content/drive/MyDrive/audd saved files/gujrati dataset'\n",
        "if not os.path.exists(\"processed_data\"):\n",
        "  os.mkdir(\"processed_data\")\n",
        "x=[]\n",
        "y=[]\n",
        "Total=23500\n",
        "count=0\n",
        "for i in range(5):\n",
        "  j = \"R\"+str(i+1)\n",
        "  for k in os.listdir(j):\n",
        "    if \".DS_\" in k:\n",
        "      continue \n",
        "    for t in os.listdir(j+\"/\"+k):\n",
        "      if \".DS_\" in t:\n",
        "        continue \n",
        "      print(count,t)\n",
        "      x.append(CalculateZeroCrossingRate((j+\"/\"+k+\"/\"+t)))\n",
        "      y.append(int(t.split(\"D\")[1][0]))\n",
        "      count+=1\n",
        "    if count%300==0:\n",
        "      print(Total-count)\n",
        "  \n",
        "x=np.array(x)\n",
        "y=np.array(y)\n",
        "np.save(files_path+\"/X_zero_crossing_rate\",x)\n",
        "np.save(files_path+\"/y_zero_crossing_rate\",y)"
      ],
      "metadata": {
        "id": "hQlatr2ngAjn"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(np.unique(y))"
      ],
      "metadata": {
        "id": "vpgnD64lgFAx"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "files_path = '/content/drive/MyDrive/audd saved files/gujrati dataset'\n",
        "\n",
        "x = np.load(files_path+\"/X_mfcc.npy\", allow_pickle=True)\n",
        "y = np.load(files_path+\"/y_mfcc.npy\", allow_pickle=True)\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.20, random_state=42)\n",
        "np.save(files_path+\"/x_mfcc_train\",x_train)\n",
        "np.save(files_path+\"/y_mfcc_train\",y_train)\n",
        "np.save(files_path+\"/x_mfcc_test\",x_test)\n",
        "np.save(files_path+\"/y_mfcc_test\",y_test)\n",
        "\n",
        "# Dono change this, not important"
      ],
      "metadata": {
        "id": "kunNeR_YgFq9"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "import numpy as np \n",
        "\n",
        "files_path = '/content/drive/MyDrive/audd saved files/gujrati dataset'\n",
        "\n",
        "features = ['mel_spectrogram', 'mfcc', 'zero_crossing_rate']\n",
        "\n",
        "x_train = np.load(files_path+\"/x_{}_train.npy\".format(features[2]), allow_pickle=True)\n",
        "y_train = np.load(files_path+\"/y_{}_train.npy\".format(features[2]), allow_pickle=True)\n",
        "\n",
        "x_test = np.load(files_path+\"/x_{}_test.npy\".format(features[2]),allow_pickle=True) \n",
        "y_test = np.load(files_path+\"/y_{}_test.npy\".format(features[2]),allow_pickle=True)"
      ],
      "metadata": {
        "id": "IXII2uUXgIdw"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "x_train =x_train.reshape((x_train.shape[0],32,32,1))\n",
        "x_test =x_test.reshape((x_test.shape[0],32,32,1))\n"
      ],
      "metadata": {
        "id": "AnTQg-VEgM8w"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Train model\n",
        "import tensorflow as tf\n",
        "\n",
        "from tensorflow.keras import datasets, layers, models\n",
        "def GetCNN():\n",
        "  model = models.Sequential()\n",
        "  \n",
        "  model.add(layers.Conv2D(64, (3, 3), activation='relu', input_shape=(32, 32, 1)))\n",
        "  model.add(layers.MaxPooling2D((2, 2)))\n",
        "  model.add(layers.BatchNormalization())\n",
        "  \n",
        "  model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
        "  model.add(layers.MaxPooling2D((2, 2)))\n",
        "  model.add(layers.BatchNormalization())\n",
        "\n",
        "  model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
        "  model.add(layers.MaxPooling2D((2, 2)))\n",
        "  model.add(layers.BatchNormalization())\n",
        "\n",
        "  model.add(layers.Dropout(0.1))\n",
        "  model.add(layers.Flatten())\n",
        "  model.add(layers.Dense(256, activation='relu'))# fully connected\n",
        "  model.add(layers.Dropout(0.1))\n",
        "  model.add(layers.Dense(128, activation='relu'))\n",
        "  model.add(layers.Dropout(0.1))\n",
        "  \n",
        "  model.add(layers.Dense(10,activation='softmax', use_bias=True))\n",
        "  return model "
      ],
      "metadata": {
        "id": "4DBcB0BugNwe"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.preprocessing import LabelBinarizer\n",
        "label_as_binary = LabelBinarizer()\n",
        "train_y = label_as_binary.fit_transform(y_train)\n",
        "test_y = label_as_binary.fit_transform(y_test)\n"
      ],
      "metadata": {
        "id": "iWyDJfmugRdh"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "model_path = '/content/drive/MyDrive/audd saved files/gujrati dataset/models'\n",
        "\n",
        "from tensorflow.keras.callbacks import ModelCheckpoint\n",
        "cnn_model = GetCNN()\n",
        "print(cnn_model.summary())\n",
        "cnn_model.compile(loss  = tf.keras.losses.CategoricalCrossentropy(),\n",
        "              metrics = [\"accuracy\"],\n",
        "              optimizer = tf.keras.optimizers.Adam(learning_rate=0.01))\n",
        "callbacks = ModelCheckpoint(model_path+\"/{}_.h5\".format(features[2]), monitor='val_accuracy',\n",
        "                                    mode='max',\n",
        "                                    save_best_only=True,\n",
        "                                    verbose=1)\n",
        "\n",
        "history = cnn_model.fit(x_train, train_y, batch_size=64, epochs=150, verbose=1, validation_split=0.1, callbacks=[callbacks]).history\n",
        "\n",
        "\n",
        "import keras\n",
        "zcr_model = keras.models.load_model(model_path+\"/{}_.h5\".format(features[2]))\n",
        "print(\"Test Accuracy is \", cnn_model.evaluate(x_test,test_y)[1])\n"
      ],
      "metadata": {
        "id": "GUZtjdiwgVLA",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "62370cf5-a2cb-4645-9c1a-abd09c97fc7a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d_6 (Conv2D)           (None, 30, 30, 64)        640       \n",
            "                                                                 \n",
            " max_pooling2d_6 (MaxPooling  (None, 15, 15, 64)       0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " batch_normalization_6 (Batc  (None, 15, 15, 64)       256       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " conv2d_7 (Conv2D)           (None, 13, 13, 64)        36928     \n",
            "                                                                 \n",
            " max_pooling2d_7 (MaxPooling  (None, 6, 6, 64)         0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " batch_normalization_7 (Batc  (None, 6, 6, 64)         256       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " conv2d_8 (Conv2D)           (None, 4, 4, 64)          36928     \n",
            "                                                                 \n",
            " max_pooling2d_8 (MaxPooling  (None, 2, 2, 64)         0         \n",
            " 2D)                                                             \n",
            "                                                                 \n",
            " batch_normalization_8 (Batc  (None, 2, 2, 64)         256       \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dropout_6 (Dropout)         (None, 2, 2, 64)          0         \n",
            "                                                                 \n",
            " flatten_2 (Flatten)         (None, 256)               0         \n",
            "                                                                 \n",
            " dense_6 (Dense)             (None, 256)               65792     \n",
            "                                                                 \n",
            " dropout_7 (Dropout)         (None, 256)               0         \n",
            "                                                                 \n",
            " dense_7 (Dense)             (None, 128)               32896     \n",
            "                                                                 \n",
            " dropout_8 (Dropout)         (None, 128)               0         \n",
            "                                                                 \n",
            " dense_8 (Dense)             (None, 10)                1290      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 175,242\n",
            "Trainable params: 174,858\n",
            "Non-trainable params: 384\n",
            "_________________________________________________________________\n",
            "None\n",
            "Epoch 1/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 1.7939 - accuracy: 0.4026\n",
            "Epoch 1: val_accuracy improved from -inf to 0.21795, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 4s 155ms/step - loss: 1.7939 - accuracy: 0.4026 - val_loss: 2.9159 - val_accuracy: 0.2179\n",
            "Epoch 2/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 1.2960 - accuracy: 0.5272\n",
            "Epoch 2: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 4s 163ms/step - loss: 1.2960 - accuracy: 0.5272 - val_loss: 2.5434 - val_accuracy: 0.1474\n",
            "Epoch 3/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 1.1592 - accuracy: 0.5523\n",
            "Epoch 3: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 124ms/step - loss: 1.1592 - accuracy: 0.5523 - val_loss: 2.5791 - val_accuracy: 0.1603\n",
            "Epoch 4/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 1.1343 - accuracy: 0.5731\n",
            "Epoch 4: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 1.1343 - accuracy: 0.5731 - val_loss: 2.8159 - val_accuracy: 0.2179\n",
            "Epoch 5/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 1.1154 - accuracy: 0.5996\n",
            "Epoch 5: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 1.1154 - accuracy: 0.5996 - val_loss: 2.7950 - val_accuracy: 0.1410\n",
            "Epoch 6/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 1.0566 - accuracy: 0.6089\n",
            "Epoch 6: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 1.0566 - accuracy: 0.6089 - val_loss: 2.6313 - val_accuracy: 0.1859\n",
            "Epoch 7/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.9899 - accuracy: 0.6519\n",
            "Epoch 7: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.9899 - accuracy: 0.6519 - val_loss: 2.9353 - val_accuracy: 0.1538\n",
            "Epoch 8/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.9942 - accuracy: 0.6447\n",
            "Epoch 8: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.9942 - accuracy: 0.6447 - val_loss: 3.4789 - val_accuracy: 0.1218\n",
            "Epoch 9/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.9766 - accuracy: 0.6354\n",
            "Epoch 9: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.9766 - accuracy: 0.6354 - val_loss: 2.7480 - val_accuracy: 0.1346\n",
            "Epoch 10/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.9669 - accuracy: 0.6268\n",
            "Epoch 10: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.9669 - accuracy: 0.6268 - val_loss: 3.5889 - val_accuracy: 0.1731\n",
            "Epoch 11/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.9665 - accuracy: 0.6612\n",
            "Epoch 11: val_accuracy did not improve from 0.21795\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.9665 - accuracy: 0.6612 - val_loss: 3.6208 - val_accuracy: 0.1603\n",
            "Epoch 12/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.9387 - accuracy: 0.6554\n",
            "Epoch 12: val_accuracy improved from 0.21795 to 0.26282, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 144ms/step - loss: 0.9387 - accuracy: 0.6554 - val_loss: 2.6027 - val_accuracy: 0.2628\n",
            "Epoch 13/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.9344 - accuracy: 0.6583\n",
            "Epoch 13: val_accuracy did not improve from 0.26282\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.9344 - accuracy: 0.6583 - val_loss: 4.0381 - val_accuracy: 0.1731\n",
            "Epoch 14/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.8802 - accuracy: 0.6698\n",
            "Epoch 14: val_accuracy did not improve from 0.26282\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.8802 - accuracy: 0.6698 - val_loss: 2.8693 - val_accuracy: 0.1474\n",
            "Epoch 15/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.8057 - accuracy: 0.7013\n",
            "Epoch 15: val_accuracy did not improve from 0.26282\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.8057 - accuracy: 0.7013 - val_loss: 2.5376 - val_accuracy: 0.1538\n",
            "Epoch 16/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.8401 - accuracy: 0.6798\n",
            "Epoch 16: val_accuracy did not improve from 0.26282\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.8401 - accuracy: 0.6798 - val_loss: 3.3511 - val_accuracy: 0.1410\n",
            "Epoch 17/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.8546 - accuracy: 0.6913\n",
            "Epoch 17: val_accuracy did not improve from 0.26282\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.8546 - accuracy: 0.6913 - val_loss: 3.0841 - val_accuracy: 0.2628\n",
            "Epoch 18/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.8112 - accuracy: 0.6905\n",
            "Epoch 18: val_accuracy did not improve from 0.26282\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.8112 - accuracy: 0.6905 - val_loss: 3.2154 - val_accuracy: 0.2115\n",
            "Epoch 19/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.7860 - accuracy: 0.6991\n",
            "Epoch 19: val_accuracy did not improve from 0.26282\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.7860 - accuracy: 0.6991 - val_loss: 3.1993 - val_accuracy: 0.2436\n",
            "Epoch 20/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.7929 - accuracy: 0.7092\n",
            "Epoch 20: val_accuracy improved from 0.26282 to 0.27564, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 144ms/step - loss: 0.7929 - accuracy: 0.7092 - val_loss: 3.1002 - val_accuracy: 0.2756\n",
            "Epoch 21/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.7405 - accuracy: 0.7357\n",
            "Epoch 21: val_accuracy improved from 0.27564 to 0.41026, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.7405 - accuracy: 0.7357 - val_loss: 2.3248 - val_accuracy: 0.4103\n",
            "Epoch 22/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.7691 - accuracy: 0.7070\n",
            "Epoch 22: val_accuracy did not improve from 0.41026\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.7691 - accuracy: 0.7070 - val_loss: 1.9310 - val_accuracy: 0.4103\n",
            "Epoch 23/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.7576 - accuracy: 0.7335\n",
            "Epoch 23: val_accuracy improved from 0.41026 to 0.50641, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 135ms/step - loss: 0.7576 - accuracy: 0.7335 - val_loss: 1.6466 - val_accuracy: 0.5064\n",
            "Epoch 24/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.7571 - accuracy: 0.7256\n",
            "Epoch 24: val_accuracy did not improve from 0.50641\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.7571 - accuracy: 0.7256 - val_loss: 1.7397 - val_accuracy: 0.4359\n",
            "Epoch 25/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.7139 - accuracy: 0.7400\n",
            "Epoch 25: val_accuracy improved from 0.50641 to 0.52564, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.7139 - accuracy: 0.7400 - val_loss: 1.5239 - val_accuracy: 0.5256\n",
            "Epoch 26/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6565 - accuracy: 0.7665\n",
            "Epoch 26: val_accuracy did not improve from 0.52564\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.6565 - accuracy: 0.7665 - val_loss: 1.5797 - val_accuracy: 0.5192\n",
            "Epoch 27/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6580 - accuracy: 0.7550\n",
            "Epoch 27: val_accuracy improved from 0.52564 to 0.62179, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 133ms/step - loss: 0.6580 - accuracy: 0.7550 - val_loss: 1.3740 - val_accuracy: 0.6218\n",
            "Epoch 28/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6822 - accuracy: 0.7486\n",
            "Epoch 28: val_accuracy did not improve from 0.62179\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.6822 - accuracy: 0.7486 - val_loss: 1.3931 - val_accuracy: 0.6154\n",
            "Epoch 29/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6540 - accuracy: 0.7622\n",
            "Epoch 29: val_accuracy did not improve from 0.62179\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.6540 - accuracy: 0.7622 - val_loss: 1.1700 - val_accuracy: 0.5833\n",
            "Epoch 30/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6371 - accuracy: 0.7514\n",
            "Epoch 30: val_accuracy did not improve from 0.62179\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.6371 - accuracy: 0.7514 - val_loss: 1.3211 - val_accuracy: 0.6090\n",
            "Epoch 31/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6004 - accuracy: 0.7715\n",
            "Epoch 31: val_accuracy improved from 0.62179 to 0.64103, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 147ms/step - loss: 0.6004 - accuracy: 0.7715 - val_loss: 1.3693 - val_accuracy: 0.6410\n",
            "Epoch 32/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6314 - accuracy: 0.7658\n",
            "Epoch 32: val_accuracy improved from 0.64103 to 0.65385, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 132ms/step - loss: 0.6314 - accuracy: 0.7658 - val_loss: 1.1746 - val_accuracy: 0.6538\n",
            "Epoch 33/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6093 - accuracy: 0.7744\n",
            "Epoch 33: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.6093 - accuracy: 0.7744 - val_loss: 1.2762 - val_accuracy: 0.6474\n",
            "Epoch 34/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5598 - accuracy: 0.7908\n",
            "Epoch 34: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.5598 - accuracy: 0.7908 - val_loss: 1.7009 - val_accuracy: 0.5385\n",
            "Epoch 35/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5803 - accuracy: 0.7908\n",
            "Epoch 35: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.5803 - accuracy: 0.7908 - val_loss: 2.0588 - val_accuracy: 0.4808\n",
            "Epoch 36/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5768 - accuracy: 0.7894\n",
            "Epoch 36: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.5768 - accuracy: 0.7894 - val_loss: 1.5118 - val_accuracy: 0.6026\n",
            "Epoch 37/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5441 - accuracy: 0.7937\n",
            "Epoch 37: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.5441 - accuracy: 0.7937 - val_loss: 1.4418 - val_accuracy: 0.5833\n",
            "Epoch 38/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5424 - accuracy: 0.7980\n",
            "Epoch 38: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.5424 - accuracy: 0.7980 - val_loss: 1.9710 - val_accuracy: 0.5449\n",
            "Epoch 39/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5281 - accuracy: 0.8001\n",
            "Epoch 39: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.5281 - accuracy: 0.8001 - val_loss: 1.6188 - val_accuracy: 0.6090\n",
            "Epoch 40/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.6313 - accuracy: 0.7658\n",
            "Epoch 40: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.6313 - accuracy: 0.7658 - val_loss: 1.4618 - val_accuracy: 0.5769\n",
            "Epoch 41/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5782 - accuracy: 0.7930\n",
            "Epoch 41: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.5782 - accuracy: 0.7930 - val_loss: 1.7579 - val_accuracy: 0.5641\n",
            "Epoch 42/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4700 - accuracy: 0.8295\n",
            "Epoch 42: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.4700 - accuracy: 0.8295 - val_loss: 1.8857 - val_accuracy: 0.5000\n",
            "Epoch 43/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4875 - accuracy: 0.8209\n",
            "Epoch 43: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4875 - accuracy: 0.8209 - val_loss: 1.8744 - val_accuracy: 0.4808\n",
            "Epoch 44/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5320 - accuracy: 0.8023\n",
            "Epoch 44: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.5320 - accuracy: 0.8023 - val_loss: 1.5898 - val_accuracy: 0.5449\n",
            "Epoch 45/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5222 - accuracy: 0.8095\n",
            "Epoch 45: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.5222 - accuracy: 0.8095 - val_loss: 1.6839 - val_accuracy: 0.5577\n",
            "Epoch 46/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4505 - accuracy: 0.8324\n",
            "Epoch 46: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4505 - accuracy: 0.8324 - val_loss: 2.0506 - val_accuracy: 0.5321\n",
            "Epoch 47/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5266 - accuracy: 0.8052\n",
            "Epoch 47: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.5266 - accuracy: 0.8052 - val_loss: 1.8351 - val_accuracy: 0.5641\n",
            "Epoch 48/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4522 - accuracy: 0.8381\n",
            "Epoch 48: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4522 - accuracy: 0.8381 - val_loss: 1.6288 - val_accuracy: 0.5769\n",
            "Epoch 49/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5530 - accuracy: 0.7901\n",
            "Epoch 49: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.5530 - accuracy: 0.7901 - val_loss: 2.0091 - val_accuracy: 0.4679\n",
            "Epoch 50/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4600 - accuracy: 0.8259\n",
            "Epoch 50: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4600 - accuracy: 0.8259 - val_loss: 1.5793 - val_accuracy: 0.6026\n",
            "Epoch 51/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5392 - accuracy: 0.8123\n",
            "Epoch 51: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.5392 - accuracy: 0.8123 - val_loss: 1.6694 - val_accuracy: 0.5705\n",
            "Epoch 52/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4693 - accuracy: 0.8173\n",
            "Epoch 52: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4693 - accuracy: 0.8173 - val_loss: 2.1549 - val_accuracy: 0.5321\n",
            "Epoch 53/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.5346 - accuracy: 0.8037\n",
            "Epoch 53: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.5346 - accuracy: 0.8037 - val_loss: 1.7840 - val_accuracy: 0.4936\n",
            "Epoch 54/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4320 - accuracy: 0.8460\n",
            "Epoch 54: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.4320 - accuracy: 0.8460 - val_loss: 1.6748 - val_accuracy: 0.5897\n",
            "Epoch 55/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4924 - accuracy: 0.8274\n",
            "Epoch 55: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4924 - accuracy: 0.8274 - val_loss: 1.4123 - val_accuracy: 0.6346\n",
            "Epoch 56/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4265 - accuracy: 0.8467\n",
            "Epoch 56: val_accuracy did not improve from 0.65385\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4265 - accuracy: 0.8467 - val_loss: 1.6345 - val_accuracy: 0.6218\n",
            "Epoch 57/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4427 - accuracy: 0.8360\n",
            "Epoch 57: val_accuracy improved from 0.65385 to 0.66667, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 142ms/step - loss: 0.4427 - accuracy: 0.8360 - val_loss: 1.5085 - val_accuracy: 0.6667\n",
            "Epoch 58/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4280 - accuracy: 0.8553\n",
            "Epoch 58: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.4280 - accuracy: 0.8553 - val_loss: 1.4277 - val_accuracy: 0.6218\n",
            "Epoch 59/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4000 - accuracy: 0.8553\n",
            "Epoch 59: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.4000 - accuracy: 0.8553 - val_loss: 2.2167 - val_accuracy: 0.5256\n",
            "Epoch 60/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3992 - accuracy: 0.8639\n",
            "Epoch 60: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3992 - accuracy: 0.8639 - val_loss: 1.4916 - val_accuracy: 0.5705\n",
            "Epoch 61/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4006 - accuracy: 0.8496\n",
            "Epoch 61: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.4006 - accuracy: 0.8496 - val_loss: 1.4034 - val_accuracy: 0.6410\n",
            "Epoch 62/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3539 - accuracy: 0.8596\n",
            "Epoch 62: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3539 - accuracy: 0.8596 - val_loss: 1.5744 - val_accuracy: 0.5897\n",
            "Epoch 63/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3448 - accuracy: 0.8782\n",
            "Epoch 63: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.3448 - accuracy: 0.8782 - val_loss: 1.4639 - val_accuracy: 0.6603\n",
            "Epoch 64/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3601 - accuracy: 0.8617\n",
            "Epoch 64: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3601 - accuracy: 0.8617 - val_loss: 2.7659 - val_accuracy: 0.5385\n",
            "Epoch 65/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3322 - accuracy: 0.8754\n",
            "Epoch 65: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3322 - accuracy: 0.8754 - val_loss: 1.6733 - val_accuracy: 0.5962\n",
            "Epoch 66/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3318 - accuracy: 0.8804\n",
            "Epoch 66: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3318 - accuracy: 0.8804 - val_loss: 3.2811 - val_accuracy: 0.4808\n",
            "Epoch 67/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3938 - accuracy: 0.8582\n",
            "Epoch 67: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.3938 - accuracy: 0.8582 - val_loss: 2.1120 - val_accuracy: 0.5705\n",
            "Epoch 68/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3647 - accuracy: 0.8668\n",
            "Epoch 68: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.3647 - accuracy: 0.8668 - val_loss: 3.2817 - val_accuracy: 0.4487\n",
            "Epoch 69/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3730 - accuracy: 0.8653\n",
            "Epoch 69: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3730 - accuracy: 0.8653 - val_loss: 2.4067 - val_accuracy: 0.4936\n",
            "Epoch 70/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3528 - accuracy: 0.8660\n",
            "Epoch 70: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3528 - accuracy: 0.8660 - val_loss: 2.3361 - val_accuracy: 0.5769\n",
            "Epoch 71/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3674 - accuracy: 0.8732\n",
            "Epoch 71: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.3674 - accuracy: 0.8732 - val_loss: 2.1242 - val_accuracy: 0.5897\n",
            "Epoch 72/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3648 - accuracy: 0.8761\n",
            "Epoch 72: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 126ms/step - loss: 0.3648 - accuracy: 0.8761 - val_loss: 2.6618 - val_accuracy: 0.5256\n",
            "Epoch 73/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3480 - accuracy: 0.8768\n",
            "Epoch 73: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3480 - accuracy: 0.8768 - val_loss: 1.8650 - val_accuracy: 0.6410\n",
            "Epoch 74/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2841 - accuracy: 0.9019\n",
            "Epoch 74: val_accuracy did not improve from 0.66667\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.2841 - accuracy: 0.9019 - val_loss: 2.1134 - val_accuracy: 0.6538\n",
            "Epoch 75/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3227 - accuracy: 0.8832\n",
            "Epoch 75: val_accuracy improved from 0.66667 to 0.67308, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 144ms/step - loss: 0.3227 - accuracy: 0.8832 - val_loss: 1.4645 - val_accuracy: 0.6731\n",
            "Epoch 76/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3419 - accuracy: 0.8890\n",
            "Epoch 76: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3419 - accuracy: 0.8890 - val_loss: 1.5240 - val_accuracy: 0.6538\n",
            "Epoch 77/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.4845 - accuracy: 0.8489\n",
            "Epoch 77: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.4845 - accuracy: 0.8489 - val_loss: 1.6544 - val_accuracy: 0.6282\n",
            "Epoch 78/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3705 - accuracy: 0.8732\n",
            "Epoch 78: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3705 - accuracy: 0.8732 - val_loss: 1.9586 - val_accuracy: 0.5705\n",
            "Epoch 79/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3879 - accuracy: 0.8560\n",
            "Epoch 79: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3879 - accuracy: 0.8560 - val_loss: 1.6365 - val_accuracy: 0.6731\n",
            "Epoch 80/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3874 - accuracy: 0.8646\n",
            "Epoch 80: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3874 - accuracy: 0.8646 - val_loss: 1.6827 - val_accuracy: 0.6538\n",
            "Epoch 81/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3516 - accuracy: 0.8739\n",
            "Epoch 81: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3516 - accuracy: 0.8739 - val_loss: 2.0731 - val_accuracy: 0.6282\n",
            "Epoch 82/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3284 - accuracy: 0.8861\n",
            "Epoch 82: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3284 - accuracy: 0.8861 - val_loss: 1.5661 - val_accuracy: 0.6474\n",
            "Epoch 83/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3426 - accuracy: 0.8789\n",
            "Epoch 83: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3426 - accuracy: 0.8789 - val_loss: 1.6826 - val_accuracy: 0.5897\n",
            "Epoch 84/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3661 - accuracy: 0.8546\n",
            "Epoch 84: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3661 - accuracy: 0.8546 - val_loss: 1.7094 - val_accuracy: 0.6410\n",
            "Epoch 85/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3880 - accuracy: 0.8761\n",
            "Epoch 85: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.3880 - accuracy: 0.8761 - val_loss: 2.0090 - val_accuracy: 0.5962\n",
            "Epoch 86/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3044 - accuracy: 0.8911\n",
            "Epoch 86: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3044 - accuracy: 0.8911 - val_loss: 1.6293 - val_accuracy: 0.6603\n",
            "Epoch 87/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3065 - accuracy: 0.8990\n",
            "Epoch 87: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3065 - accuracy: 0.8990 - val_loss: 1.4910 - val_accuracy: 0.6731\n",
            "Epoch 88/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2892 - accuracy: 0.8990\n",
            "Epoch 88: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2892 - accuracy: 0.8990 - val_loss: 1.5763 - val_accuracy: 0.6603\n",
            "Epoch 89/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2894 - accuracy: 0.8983\n",
            "Epoch 89: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2894 - accuracy: 0.8983 - val_loss: 1.7887 - val_accuracy: 0.6282\n",
            "Epoch 90/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2964 - accuracy: 0.8976\n",
            "Epoch 90: val_accuracy did not improve from 0.67308\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.2964 - accuracy: 0.8976 - val_loss: 1.8201 - val_accuracy: 0.6603\n",
            "Epoch 91/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3228 - accuracy: 0.8897\n",
            "Epoch 91: val_accuracy improved from 0.67308 to 0.68590, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 142ms/step - loss: 0.3228 - accuracy: 0.8897 - val_loss: 1.5216 - val_accuracy: 0.6859\n",
            "Epoch 92/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3196 - accuracy: 0.8911\n",
            "Epoch 92: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3196 - accuracy: 0.8911 - val_loss: 2.0001 - val_accuracy: 0.6090\n",
            "Epoch 93/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2830 - accuracy: 0.9026\n",
            "Epoch 93: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.2830 - accuracy: 0.9026 - val_loss: 1.9056 - val_accuracy: 0.5962\n",
            "Epoch 94/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3172 - accuracy: 0.8954\n",
            "Epoch 94: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.3172 - accuracy: 0.8954 - val_loss: 1.5461 - val_accuracy: 0.6410\n",
            "Epoch 95/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3368 - accuracy: 0.8832\n",
            "Epoch 95: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.3368 - accuracy: 0.8832 - val_loss: 1.6095 - val_accuracy: 0.6538\n",
            "Epoch 96/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2679 - accuracy: 0.9033\n",
            "Epoch 96: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.2679 - accuracy: 0.9033 - val_loss: 1.8174 - val_accuracy: 0.6154\n",
            "Epoch 97/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2884 - accuracy: 0.9112\n",
            "Epoch 97: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.2884 - accuracy: 0.9112 - val_loss: 1.5654 - val_accuracy: 0.6346\n",
            "Epoch 98/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2656 - accuracy: 0.9011\n",
            "Epoch 98: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2656 - accuracy: 0.9011 - val_loss: 1.4750 - val_accuracy: 0.6346\n",
            "Epoch 99/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2629 - accuracy: 0.9105\n",
            "Epoch 99: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2629 - accuracy: 0.9105 - val_loss: 1.5517 - val_accuracy: 0.6667\n",
            "Epoch 100/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2570 - accuracy: 0.9076\n",
            "Epoch 100: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2570 - accuracy: 0.9076 - val_loss: 1.8537 - val_accuracy: 0.6026\n",
            "Epoch 101/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2631 - accuracy: 0.9062\n",
            "Epoch 101: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2631 - accuracy: 0.9062 - val_loss: 1.7795 - val_accuracy: 0.6346\n",
            "Epoch 102/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2241 - accuracy: 0.9140\n",
            "Epoch 102: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2241 - accuracy: 0.9140 - val_loss: 1.9195 - val_accuracy: 0.6410\n",
            "Epoch 103/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2946 - accuracy: 0.9019\n",
            "Epoch 103: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.2946 - accuracy: 0.9019 - val_loss: 3.0433 - val_accuracy: 0.5385\n",
            "Epoch 104/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2886 - accuracy: 0.9083\n",
            "Epoch 104: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 127ms/step - loss: 0.2886 - accuracy: 0.9083 - val_loss: 2.2412 - val_accuracy: 0.5513\n",
            "Epoch 105/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3312 - accuracy: 0.8961\n",
            "Epoch 105: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.3312 - accuracy: 0.8961 - val_loss: 2.3699 - val_accuracy: 0.5769\n",
            "Epoch 106/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2896 - accuracy: 0.8933\n",
            "Epoch 106: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.2896 - accuracy: 0.8933 - val_loss: 2.1356 - val_accuracy: 0.6026\n",
            "Epoch 107/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2237 - accuracy: 0.9269\n",
            "Epoch 107: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2237 - accuracy: 0.9269 - val_loss: 1.8290 - val_accuracy: 0.6282\n",
            "Epoch 108/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1735 - accuracy: 0.9341\n",
            "Epoch 108: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.1735 - accuracy: 0.9341 - val_loss: 2.1527 - val_accuracy: 0.5962\n",
            "Epoch 109/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1893 - accuracy: 0.9319\n",
            "Epoch 109: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.1893 - accuracy: 0.9319 - val_loss: 2.1778 - val_accuracy: 0.6410\n",
            "Epoch 110/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2435 - accuracy: 0.9226\n",
            "Epoch 110: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.2435 - accuracy: 0.9226 - val_loss: 2.0259 - val_accuracy: 0.5962\n",
            "Epoch 111/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2342 - accuracy: 0.9248\n",
            "Epoch 111: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2342 - accuracy: 0.9248 - val_loss: 1.9783 - val_accuracy: 0.5962\n",
            "Epoch 112/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2776 - accuracy: 0.9090\n",
            "Epoch 112: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2776 - accuracy: 0.9090 - val_loss: 1.8756 - val_accuracy: 0.6667\n",
            "Epoch 113/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2595 - accuracy: 0.9191\n",
            "Epoch 113: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 128ms/step - loss: 0.2595 - accuracy: 0.9191 - val_loss: 1.7063 - val_accuracy: 0.6218\n",
            "Epoch 114/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2434 - accuracy: 0.9183\n",
            "Epoch 114: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.2434 - accuracy: 0.9183 - val_loss: 2.1722 - val_accuracy: 0.5833\n",
            "Epoch 115/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2495 - accuracy: 0.9076\n",
            "Epoch 115: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.2495 - accuracy: 0.9076 - val_loss: 2.1504 - val_accuracy: 0.6346\n",
            "Epoch 116/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2214 - accuracy: 0.9298\n",
            "Epoch 116: val_accuracy did not improve from 0.68590\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.2214 - accuracy: 0.9298 - val_loss: 1.7829 - val_accuracy: 0.6731\n",
            "Epoch 117/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2684 - accuracy: 0.9176\n",
            "Epoch 117: val_accuracy improved from 0.68590 to 0.69231, saving model to /content/drive/MyDrive/audd saved files/gujrati dataset/models/zero_crossing_rate_.h5\n",
            "22/22 [==============================] - 3s 147ms/step - loss: 0.2684 - accuracy: 0.9176 - val_loss: 1.7872 - val_accuracy: 0.6923\n",
            "Epoch 118/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2789 - accuracy: 0.9105\n",
            "Epoch 118: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.2789 - accuracy: 0.9105 - val_loss: 2.1655 - val_accuracy: 0.6218\n",
            "Epoch 119/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2787 - accuracy: 0.9026\n",
            "Epoch 119: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2787 - accuracy: 0.9026 - val_loss: 1.7783 - val_accuracy: 0.6731\n",
            "Epoch 120/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2857 - accuracy: 0.9090\n",
            "Epoch 120: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 132ms/step - loss: 0.2857 - accuracy: 0.9090 - val_loss: 1.6707 - val_accuracy: 0.6154\n",
            "Epoch 121/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2666 - accuracy: 0.9069\n",
            "Epoch 121: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 134ms/step - loss: 0.2666 - accuracy: 0.9069 - val_loss: 1.7077 - val_accuracy: 0.6731\n",
            "Epoch 122/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2704 - accuracy: 0.9026\n",
            "Epoch 122: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 132ms/step - loss: 0.2704 - accuracy: 0.9026 - val_loss: 2.1469 - val_accuracy: 0.6538\n",
            "Epoch 123/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2501 - accuracy: 0.9105\n",
            "Epoch 123: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2501 - accuracy: 0.9105 - val_loss: 2.1547 - val_accuracy: 0.6474\n",
            "Epoch 124/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2075 - accuracy: 0.9291\n",
            "Epoch 124: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.2075 - accuracy: 0.9291 - val_loss: 2.0588 - val_accuracy: 0.6474\n",
            "Epoch 125/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2067 - accuracy: 0.9291\n",
            "Epoch 125: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.2067 - accuracy: 0.9291 - val_loss: 1.9114 - val_accuracy: 0.6731\n",
            "Epoch 126/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1793 - accuracy: 0.9291\n",
            "Epoch 126: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.1793 - accuracy: 0.9291 - val_loss: 2.0669 - val_accuracy: 0.6474\n",
            "Epoch 127/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1994 - accuracy: 0.9348\n",
            "Epoch 127: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.1994 - accuracy: 0.9348 - val_loss: 2.6582 - val_accuracy: 0.6346\n",
            "Epoch 128/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2528 - accuracy: 0.9219\n",
            "Epoch 128: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 132ms/step - loss: 0.2528 - accuracy: 0.9219 - val_loss: 2.3516 - val_accuracy: 0.5833\n",
            "Epoch 129/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2426 - accuracy: 0.9140\n",
            "Epoch 129: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 133ms/step - loss: 0.2426 - accuracy: 0.9140 - val_loss: 2.6394 - val_accuracy: 0.5833\n",
            "Epoch 130/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2962 - accuracy: 0.9097\n",
            "Epoch 130: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2962 - accuracy: 0.9097 - val_loss: 2.2981 - val_accuracy: 0.5769\n",
            "Epoch 131/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1814 - accuracy: 0.9334\n",
            "Epoch 131: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.1814 - accuracy: 0.9334 - val_loss: 2.0977 - val_accuracy: 0.6090\n",
            "Epoch 132/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1545 - accuracy: 0.9477\n",
            "Epoch 132: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 132ms/step - loss: 0.1545 - accuracy: 0.9477 - val_loss: 2.6104 - val_accuracy: 0.6346\n",
            "Epoch 133/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1941 - accuracy: 0.9463\n",
            "Epoch 133: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.1941 - accuracy: 0.9463 - val_loss: 2.6851 - val_accuracy: 0.6282\n",
            "Epoch 134/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1605 - accuracy: 0.9463\n",
            "Epoch 134: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 132ms/step - loss: 0.1605 - accuracy: 0.9463 - val_loss: 2.6400 - val_accuracy: 0.6410\n",
            "Epoch 135/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1765 - accuracy: 0.9413\n",
            "Epoch 135: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.1765 - accuracy: 0.9413 - val_loss: 2.4692 - val_accuracy: 0.6538\n",
            "Epoch 136/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2061 - accuracy: 0.9413\n",
            "Epoch 136: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.2061 - accuracy: 0.9413 - val_loss: 2.3905 - val_accuracy: 0.6603\n",
            "Epoch 137/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2113 - accuracy: 0.9319\n",
            "Epoch 137: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2113 - accuracy: 0.9319 - val_loss: 2.9477 - val_accuracy: 0.5769\n",
            "Epoch 138/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2286 - accuracy: 0.9291\n",
            "Epoch 138: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2286 - accuracy: 0.9291 - val_loss: 2.8141 - val_accuracy: 0.6218\n",
            "Epoch 139/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2765 - accuracy: 0.9083\n",
            "Epoch 139: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 133ms/step - loss: 0.2765 - accuracy: 0.9083 - val_loss: 2.2456 - val_accuracy: 0.6154\n",
            "Epoch 140/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2512 - accuracy: 0.9226\n",
            "Epoch 140: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2512 - accuracy: 0.9226 - val_loss: 2.2214 - val_accuracy: 0.6410\n",
            "Epoch 141/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2393 - accuracy: 0.9191\n",
            "Epoch 141: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2393 - accuracy: 0.9191 - val_loss: 2.1001 - val_accuracy: 0.6667\n",
            "Epoch 142/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2466 - accuracy: 0.9291\n",
            "Epoch 142: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2466 - accuracy: 0.9291 - val_loss: 3.5424 - val_accuracy: 0.5513\n",
            "Epoch 143/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2276 - accuracy: 0.9269\n",
            "Epoch 143: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.2276 - accuracy: 0.9269 - val_loss: 2.2857 - val_accuracy: 0.6410\n",
            "Epoch 144/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2003 - accuracy: 0.9284\n",
            "Epoch 144: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.2003 - accuracy: 0.9284 - val_loss: 1.9618 - val_accuracy: 0.6154\n",
            "Epoch 145/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.3011 - accuracy: 0.9083\n",
            "Epoch 145: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.3011 - accuracy: 0.9083 - val_loss: 2.4759 - val_accuracy: 0.6346\n",
            "Epoch 146/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2157 - accuracy: 0.9284\n",
            "Epoch 146: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.2157 - accuracy: 0.9284 - val_loss: 2.1533 - val_accuracy: 0.6090\n",
            "Epoch 147/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2136 - accuracy: 0.9348\n",
            "Epoch 147: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.2136 - accuracy: 0.9348 - val_loss: 2.0877 - val_accuracy: 0.6346\n",
            "Epoch 148/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1546 - accuracy: 0.9513\n",
            "Epoch 148: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 130ms/step - loss: 0.1546 - accuracy: 0.9513 - val_loss: 2.3462 - val_accuracy: 0.6603\n",
            "Epoch 149/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.1865 - accuracy: 0.9427\n",
            "Epoch 149: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 129ms/step - loss: 0.1865 - accuracy: 0.9427 - val_loss: 2.1503 - val_accuracy: 0.6218\n",
            "Epoch 150/150\n",
            "22/22 [==============================] - ETA: 0s - loss: 0.2027 - accuracy: 0.9420\n",
            "Epoch 150: val_accuracy did not improve from 0.69231\n",
            "22/22 [==============================] - 3s 131ms/step - loss: 0.2027 - accuracy: 0.9420 - val_loss: 2.4909 - val_accuracy: 0.5769\n",
            "13/13 [==============================] - 0s 18ms/step - loss: 2.3852 - accuracy: 0.6134\n",
            "Test Accuracy is  0.6134020686149597\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# save history\n",
        "np.save(model_path+\"/{}_history.npy\".format(features[2]),history)\n",
        "#print(history['accuracy'])"
      ],
      "metadata": {
        "id": "7tE7NLwtbMYp"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#plot history\n",
        "import matplotlib.pylab as plt\n",
        "from matplotlib.pyplot import figure\n",
        "figure(figsize=(8, 6))\n",
        "plt.plot(history['loss'])\n",
        "plt.plot(history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'test'], loc='upper left')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 404
        },
        "id": "6YNQdTg2b4r0",
        "outputId": "2c2edbd9-51a1-4d0e-eb82-42a87b4a45f3"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x432 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Accuracy plot\n",
        "plt.plot(history['accuracy'])\n",
        "plt.plot(history['val_accuracy'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='upper left')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "sygRhLYCchBY",
        "outputId": "91889537-55ef-4c8a-e590-028ed156a7d7"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Load all models\n",
        "import keras\n",
        "model_path = '/content/drive/MyDrive/audd saved files/gujrati dataset/models'\n",
        "\n",
        "mp_model = keras.models.load_model(model_path+\"/mel_spectrogram_.h5\")\n",
        "mfcc_model = keras.models.load_model(model_path+\"/mfcc_.h5\")\n",
        "zc_model = keras.models.load_model(model_path+\"/zero_crossing_rate_.h5\")"
      ],
      "metadata": {
        "id": "dGZWpcaFgYje"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Model Ensembling\n",
        "from keras.models import Model\n",
        "from keras import layers\n",
        "\n",
        "def ensembleModels(models, model_input):\n",
        "    # collect outputs of models in a list\n",
        "    model_outputs=[model(model_input) for model in models] \n",
        "    # averaging outputs\n",
        "    ensemble_output=layers.average(model_outputs) \n",
        "    # build model from same input and avg output\n",
        "    modelEns = Model(inputs=model_input, outputs=ensemble_output,    name='ensemble')  \n",
        "   \n",
        "    return modelEns"
      ],
      "metadata": {
        "id": "1KcLLNOZgaaP"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Ensembling\n",
        "from keras.models import Input\n",
        "models=[mp_model,mfcc_model,zc_model]\n",
        "\n",
        "model_input = Input(shape=models[0].input_shape[1:]) # c*h*w\n",
        "modelEns = ensembleModels(models, model_input)\n",
        "modelEns.summary()\n",
        "\n",
        "modelEns.save(model_path+\"/ensemble_.h5\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "09afdhrcgceY",
        "outputId": "1f116276-34cf-41af-a4dd-e8d05a1441f9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"ensemble\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_2 (InputLayer)           [(None, 32, 32, 1)]  0           []                               \n",
            "                                                                                                  \n",
            " sequential (Sequential)        (None, 10)           175242      ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " sequential_5 (Sequential)      (None, 10)           175242      ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " sequential_3 (Sequential)      (None, 10)           175242      ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " average (Average)              (None, 10)           0           ['sequential[1][0]',             \n",
            "                                                                  'sequential_5[1][0]',           \n",
            "                                                                  'sequential_3[1][0]']           \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 525,726\n",
            "Trainable params: 524,574\n",
            "Non-trainable params: 1,152\n",
            "__________________________________________________________________________________________________\n",
            "WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Load model and evaluate\n",
        "models_path = '/content/drive/MyDrive/audd saved files/gujrati dataset/models'\n",
        "model = keras.models.load_model(models_path+\"/ensemble_.h5\")\n",
        "model.summary()\n",
        "\n",
        "print(\"Test Accuracy is \", model.evaluate(x_test,test_y)[1])\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 727
        },
        "id": "44NN_p6ng-BF",
        "outputId": "c60502b7-c70a-4e1b-e3a2-1dacf77fa7a6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "WARNING:tensorflow:No training configuration found in the save file, so the model was *not* compiled. Compile it manually.\n",
            "Model: \"ensemble\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_2 (InputLayer)           [(None, 32, 32, 1)]  0           []                               \n",
            "                                                                                                  \n",
            " sequential (Sequential)        (None, 10)           175242      ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " sequential_5 (Sequential)      (None, 10)           175242      ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " sequential_3 (Sequential)      (None, 10)           175242      ['input_2[0][0]']                \n",
            "                                                                                                  \n",
            " average (Average)              (None, 10)           0           ['sequential[0][0]',             \n",
            "                                                                  'sequential_5[0][0]',           \n",
            "                                                                  'sequential_3[0][0]']           \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 525,726\n",
            "Trainable params: 524,574\n",
            "Non-trainable params: 1,152\n",
            "__________________________________________________________________________________________________\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "RuntimeError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-16-4746c05c81da>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msummary\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Test Accuracy is \"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mevaluate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx_test\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mtest_y\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/keras/utils/traceback_utils.py\u001b[0m in \u001b[0;36merror_handler\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     65\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0mException\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m  \u001b[0;31m# pylint: disable=broad-except\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     66\u001b[0m       \u001b[0mfiltered_tb\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0m_process_traceback_frames\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m__traceback__\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 67\u001b[0;31m       \u001b[0;32mraise\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwith_traceback\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfiltered_tb\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     68\u001b[0m     \u001b[0;32mfinally\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     69\u001b[0m       \u001b[0;32mdel\u001b[0m \u001b[0mfiltered_tb\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/usr/local/lib/python3.7/dist-packages/keras/engine/training.py\u001b[0m in \u001b[0;36m_assert_compile_was_called\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   3057\u001b[0m     \u001b[0;31m# (i.e. whether the model is built and its inputs/outputs are set).\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   3058\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_is_compiled\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 3059\u001b[0;31m       raise RuntimeError('You must compile your model before '\n\u001b[0m\u001b[1;32m   3060\u001b[0m                          \u001b[0;34m'training/testing. '\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   3061\u001b[0m                          'Use `model.compile(optimizer, loss)`.')\n",
            "\u001b[0;31mRuntimeError\u001b[0m: You must compile your model before training/testing. Use `model.compile(optimizer, loss)`."
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from sklearn.preprocessing import LabelBinarizer\n",
        "\n",
        "def evaluateAll(model, x_test, test_y):\n",
        "  label_as_binary = LabelBinarizer()\n",
        "  test_y = label_as_binary.fit_transform(test_y)\n",
        "  return model.evaluate(x_test,test_y)\n",
        "\n"
      ],
      "metadata": {
        "id": "8TsCO492QhVT"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#Test data\n",
        "import keras\n",
        "files_path = '/content/drive/MyDrive/audd saved files/gujrati dataset'\n",
        "\n",
        "x_test = np.load(files_path+\"/x_mel_spectrogram_test.npy\",allow_pickle=True) \n",
        "y_test = np.load(files_path+\"/y_mel_spectrogram_test.npy\",allow_pickle=True)\n",
        "\n",
        "# model\n",
        "model_path = '/content/drive/MyDrive/audd saved files/gujrati dataset/models'\n",
        "\n",
        "mp_model = keras.models.load_model(model_path+\"/mel_spectrogram_.h5\")\n",
        "\n",
        "model_result = evaluateAll(mp_model, x_test, y_test)\n",
        "print(model_result)\n",
        "print(\"model history \", mp_model.history)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "a3QPbhDCUH_x",
        "outputId": "8f6da6ef-c8a4-4860-ee39-b41486553bba"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "13/13 [==============================] - 0s 12ms/step - loss: 0.2081 - accuracy: 0.9536\n",
            "[0.20806215703487396, 0.9536082744598389]\n",
            "model history  <keras.callbacks.History object at 0x7f1796f2c7d0>\n"
          ]
        }
      ]
    }
  ]
}