feature_extract.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "metadata": {
        "id": "f7IzYuDP49ck"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
        "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
        "trusted": true,
        "id": "56eEjGFt4ART"
      },
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "import librosa as lb\n",
        "import librosa.display as lbd\n",
        "import os"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a",
        "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
        "trusted": true,
        "id": "Xgq0yNgR4ARU",
        "outputId": "dc36af76-a40d-4b28-f3a7-88f99276dc69"
      },
      "cell_type": "code",
      "source": [
        "train=pd.read_csv('/content/drive/MyDrive/Respiratory_Sound_Database/train.csv')\n",
        "val=pd.read_csv('/content/drive/MyDrive/Respiratory_Sound_Database/val.csv')\n",
        "train.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 3,
          "data": {
            "text/plain": "   Unnamed: 0  pid                      filename  disease\n0        1481  176  176_1b3_Ll_mc_AKGC417L_3.wav     COPD\n1        6197  144  144_1b1_Al_sc_Meditron_9.wav  Healthy\n2         499  193  193_1b2_Ll_mc_AKGC417L_0.wav     COPD\n3        6559  208  208_1b1_Ll_sc_Meditron_8.wav  Healthy\n4        1238  154  154_3b3_Al_mc_AKGC417L_3.wav     COPD",
            "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>Unnamed: 0</th>\n      <th>pid</th>\n      <th>filename</th>\n      <th>disease</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>1481</td>\n      <td>176</td>\n      <td>176_1b3_Ll_mc_AKGC417L_3.wav</td>\n      <td>COPD</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>6197</td>\n      <td>144</td>\n      <td>144_1b1_Al_sc_Meditron_9.wav</td>\n      <td>Healthy</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>499</td>\n      <td>193</td>\n      <td>193_1b2_Ll_mc_AKGC417L_0.wav</td>\n      <td>COPD</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>6559</td>\n      <td>208</td>\n      <td>208_1b1_Ll_sc_Meditron_8.wav</td>\n      <td>Healthy</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>1238</td>\n      <td>154</td>\n      <td>154_3b3_Al_mc_AKGC417L_3.wav</td>\n      <td>COPD</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "lk_PyxkU4ARV",
        "outputId": "5d7662e5-2809-41ce-8397-5dd20730afa6"
      },
      "cell_type": "code",
      "source": [
        "ytrain=train.disease\n",
        "yval=val.disease\n",
        "yval"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 4,
          "data": {
            "text/plain": "0       COPD\n1       COPD\n2       COPD\n3       COPD\n4       COPD\n        ... \n1720    COPD\n1721    COPD\n1722    COPD\n1723    COPD\n1724    COPD\nName: disease, Length: 1725, dtype: object"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "IUKC0gjj4ARW"
      },
      "cell_type": "code",
      "source": [
        "from sklearn.preprocessing import LabelEncoder\n",
        "le=LabelEncoder()\n",
        "ytrain=le.fit_transform(ytrain)\n",
        "yval=le.transform(yval)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "id": "P-YqlJ2o4ARW"
      },
      "cell_type": "markdown",
      "source": [
        " Librosa's feature extraction methods\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "aByaakd54ARW"
      },
      "cell_type": "code",
      "source": [
        "def getFeatures(path):\n",
        "    soundArr,sample_rate=lb.load(path)\n",
        "    mfcc=lb.feature.mfcc(y=soundArr,sr=sample_rate)\n",
        "    cstft=lb.feature.chroma_stft(y=soundArr,sr=sample_rate)\n",
        "    mSpec=lb.feature.melspectrogram(y=soundArr,sr=sample_rate)\n",
        "\n",
        "    return mfcc,cstft,mSpec"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "id": "LyHHirpA4ARW"
      },
      "cell_type": "markdown",
      "source": [
        "> iterating over train and val dataset,converting them to numpy arrays\n"
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "xlGZ8tm44ARW",
        "outputId": "926eda2d-502b-4dc4-d24a-7e88c02d53df"
      },
      "cell_type": "code",
      "source": [
        "root='/content/drive/MyDrive/Respiratory_Sound_Database/processed_audio_files/'\n",
        "mfcc,cstft,mSpec=[],[],[]\n",
        "i=0\n",
        "for idx,row in val.iterrows():\n",
        "    path=root + row['filename']\n",
        "    a,b,c=getFeatures(path)\n",
        "    mfcc.append(a)\n",
        "    cstft.append(b)\n",
        "    mSpec.append(c)\n",
        "\n",
        "mfcc_val=np.array(mfcc)\n",
        "cstft_val=np.array(cstft)\n",
        "mSpec_val=np.array(mSpec)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": "/opt/conda/lib/python3.7/site-packages/librosa/core/pitch.py:153: UserWarning: Trying to estimate tuning from empty frequency set.\n  warnings.warn(\"Trying to estimate tuning from empty frequency set.\")\n",
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "Ijsvbe3-4ARW"
      },
      "cell_type": "code",
      "source": [
        "root='/content/drive/MyDrive/Respiratory_Sound_Database/processed_audio_files/'\n",
        "mfcc,cstft,mSpec=[],[],[]\n",
        "i=0\n",
        "for idx,row in train.iterrows():\n",
        "    path=root + row['filename']\n",
        "    a,b,c=getFeatures(path)\n",
        "    mfcc.append(a)\n",
        "    cstft.append(b)\n",
        "    mSpec.append(c)\n",
        "\n",
        "mfcc_train=np.array(mfcc)\n",
        "cstft_train=np.array(cstft)\n",
        "mSpec_train=np.array(mSpec)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "id": "bZGbOrH54ARW"
      },
      "cell_type": "markdown",
      "source": [
        "> Discarded few features beacuse they were not important to performance"
      ]
    },
    {
      "metadata": {
        "id": "3Dewrn974ARX"
      },
      "cell_type": "markdown",
      "source": [
        "# Testing Features\n",
        "\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "ytbr8xNy4ARX"
      },
      "cell_type": "code",
      "source": [
        "my_callbacks = [\n",
        "    tf.keras.callbacks.EarlyStopping(patience=5),\n",
        "    tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1,\n",
        "                              patience=3, min_lr=0.00001,mode='min'),\n",
        "    tf.keras.callbacks.ModelCheckpoint('hbNet', monitor='val_loss', verbose=0, save_best_only=True,)\n",
        "]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "id": "yiUcOcED4ARX"
      },
      "cell_type": "markdown",
      "source": [
        "* Here i m creating **three** different **CNN** and then Combining them to a **Dense Network**."
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "XyH2NrcE4ARX"
      },
      "cell_type": "code",
      "source": [
        "mfcc_input=keras.layers.Input(shape=(20,259,1),name=\"mfccInput\")\n",
        "x=keras.layers.Conv2D(32,5,strides=(1,3),padding='same')(mfcc_input)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(64,3,strides=(1,2),padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(96,2,padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(128,2,padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "mfcc_output=keras.layers.GlobalMaxPooling2D()(x)\n",
        "\n",
        "mfcc_model=keras.Model(mfcc_input, mfcc_output, name=\"mfccModel\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "1VUl81Ok4ARX",
        "outputId": "42ce8752-6e9a-4ce1-e10a-ab2353ddd6ec"
      },
      "cell_type": "code",
      "source": [
        "mfcc_model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Model: \"mfccModel\"\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\nmfccInput (InputLayer)       [(None, 20, 259, 1)]      0         \n_________________________________________________________________\nconv2d (Conv2D)              (None, 20, 87, 32)        832       \n_________________________________________________________________\nbatch_normalization (BatchNo (None, 20, 87, 32)        128       \n_________________________________________________________________\nactivation (Activation)      (None, 20, 87, 32)        0         \n_________________________________________________________________\nmax_pooling2d (MaxPooling2D) (None, 10, 43, 32)        0         \n_________________________________________________________________\nconv2d_1 (Conv2D)            (None, 10, 22, 64)        18496     \n_________________________________________________________________\nbatch_normalization_1 (Batch (None, 10, 22, 64)        256       \n_________________________________________________________________\nactivation_1 (Activation)    (None, 10, 22, 64)        0         \n_________________________________________________________________\nmax_pooling2d_1 (MaxPooling2 (None, 5, 11, 64)         0         \n_________________________________________________________________\nconv2d_2 (Conv2D)            (None, 5, 11, 96)         24672     \n_________________________________________________________________\nbatch_normalization_2 (Batch (None, 5, 11, 96)         384       \n_________________________________________________________________\nactivation_2 (Activation)    (None, 5, 11, 96)         0         \n_________________________________________________________________\nmax_pooling2d_2 (MaxPooling2 (None, 2, 5, 96)          0         \n_________________________________________________________________\nconv2d_3 (Conv2D)            (None, 2, 5, 128)         49280     \n_________________________________________________________________\nbatch_normalization_3 (Batch (None, 2, 5, 128)         512       \n_________________________________________________________________\nactivation_3 (Activation)    (None, 2, 5, 128)         0         \n_________________________________________________________________\nglobal_max_pooling2d (Global (None, 128)               0         \n=================================================================\nTotal params: 94,560\nTrainable params: 93,920\nNon-trainable params: 640\n_________________________________________________________________\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "fd30O8UP4ARX"
      },
      "cell_type": "code",
      "source": [
        "croma_input=keras.layers.Input(shape=(12,259,1),name=\"cromaInput\")\n",
        "x=keras.layers.Conv2D(32,5,strides=(1,3),padding='same')(croma_input)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(64,3,strides=(1,2),padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(128,2,padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "croma_output=keras.layers.GlobalMaxPooling2D()(x)\n",
        "\n",
        "croma_model=keras.Model(croma_input, croma_output, name=\"cromaModel\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "IHRrpmKo4ARX",
        "outputId": "8d0554a7-eaef-4da9-c9c0-3facbc4a9efd"
      },
      "cell_type": "code",
      "source": [
        "croma_model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Model: \"cromaModel\"\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\ncromaInput (InputLayer)      [(None, 12, 259, 1)]      0         \n_________________________________________________________________\nconv2d_4 (Conv2D)            (None, 12, 87, 32)        832       \n_________________________________________________________________\nbatch_normalization_4 (Batch (None, 12, 87, 32)        128       \n_________________________________________________________________\nactivation_4 (Activation)    (None, 12, 87, 32)        0         \n_________________________________________________________________\nmax_pooling2d_3 (MaxPooling2 (None, 6, 43, 32)         0         \n_________________________________________________________________\nconv2d_5 (Conv2D)            (None, 6, 22, 64)         18496     \n_________________________________________________________________\nbatch_normalization_5 (Batch (None, 6, 22, 64)         256       \n_________________________________________________________________\nactivation_5 (Activation)    (None, 6, 22, 64)         0         \n_________________________________________________________________\nmax_pooling2d_4 (MaxPooling2 (None, 3, 11, 64)         0         \n_________________________________________________________________\nconv2d_6 (Conv2D)            (None, 3, 11, 128)        32896     \n_________________________________________________________________\nbatch_normalization_6 (Batch (None, 3, 11, 128)        512       \n_________________________________________________________________\nactivation_6 (Activation)    (None, 3, 11, 128)        0         \n_________________________________________________________________\nglobal_max_pooling2d_1 (Glob (None, 128)               0         \n=================================================================\nTotal params: 53,120\nTrainable params: 52,672\nNon-trainable params: 448\n_________________________________________________________________\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "oTOKYkuf4ARX"
      },
      "cell_type": "code",
      "source": [
        "mSpec_input=keras.layers.Input(shape=(128,259,1),name=\"mSpecInput\")\n",
        "x=keras.layers.Conv2D(32,5,strides=(2,3),padding='same')(mSpec_input)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(64,3,strides=(2,2),padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(96,2,padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "x=keras.layers.MaxPooling2D(pool_size=2,padding='valid')(x)\n",
        "\n",
        "x=keras.layers.Conv2D(128,2,padding='same')(x)\n",
        "x=keras.layers.BatchNormalization()(x)\n",
        "x=keras.layers.Activation(keras.activations.relu)(x)\n",
        "mSpec_output=keras.layers.GlobalMaxPooling2D()(x)\n",
        "\n",
        "mSpec_model=keras.Model(mSpec_input, mSpec_output, name=\"mSpecModel\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "7CMecY4C4ARX",
        "outputId": "270084e2-0626-4687-b4fe-65b79eb269a0"
      },
      "cell_type": "code",
      "source": [
        "mSpec_model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Model: \"mSpecModel\"\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\nmSpecInput (InputLayer)      [(None, 128, 259, 1)]     0         \n_________________________________________________________________\nconv2d_7 (Conv2D)            (None, 64, 87, 32)        832       \n_________________________________________________________________\nbatch_normalization_7 (Batch (None, 64, 87, 32)        128       \n_________________________________________________________________\nactivation_7 (Activation)    (None, 64, 87, 32)        0         \n_________________________________________________________________\nmax_pooling2d_5 (MaxPooling2 (None, 32, 43, 32)        0         \n_________________________________________________________________\nconv2d_8 (Conv2D)            (None, 16, 22, 64)        18496     \n_________________________________________________________________\nbatch_normalization_8 (Batch (None, 16, 22, 64)        256       \n_________________________________________________________________\nactivation_8 (Activation)    (None, 16, 22, 64)        0         \n_________________________________________________________________\nmax_pooling2d_6 (MaxPooling2 (None, 8, 11, 64)         0         \n_________________________________________________________________\nconv2d_9 (Conv2D)            (None, 8, 11, 96)         24672     \n_________________________________________________________________\nbatch_normalization_9 (Batch (None, 8, 11, 96)         384       \n_________________________________________________________________\nactivation_9 (Activation)    (None, 8, 11, 96)         0         \n_________________________________________________________________\nmax_pooling2d_7 (MaxPooling2 (None, 4, 5, 96)          0         \n_________________________________________________________________\nconv2d_10 (Conv2D)           (None, 4, 5, 128)         49280     \n_________________________________________________________________\nbatch_normalization_10 (Batc (None, 4, 5, 128)         512       \n_________________________________________________________________\nactivation_10 (Activation)   (None, 4, 5, 128)         0         \n_________________________________________________________________\nglobal_max_pooling2d_2 (Glob (None, 128)               0         \n=================================================================\nTotal params: 94,560\nTrainable params: 93,920\nNon-trainable params: 640\n_________________________________________________________________\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "X9V5gzaq4ARY"
      },
      "cell_type": "code",
      "source": [
        "input_mfcc=keras.layers.Input(shape=(20,259,1),name=\"mfcc\")\n",
        "mfcc=mfcc_model(input_mfcc)\n",
        "\n",
        "input_croma=keras.layers.Input(shape=(12,259,1),name=\"croma\")\n",
        "croma=croma_model(input_croma)\n",
        "\n",
        "input_mSpec=keras.layers.Input(shape=(128,259,1),name=\"mspec\")\n",
        "mSpec=mSpec_model(input_mSpec)\n",
        "\n",
        "\n",
        "concat=keras.layers.concatenate([mfcc,croma,mSpec])\n",
        "hidden=keras.layers.Dropout(0.2)(concat)\n",
        "hidden=keras.layers.Dense(50,activation='relu')(concat)\n",
        "hidden=keras.layers.Dropout(0.3)(hidden)\n",
        "hidden=keras.layers.Dense(25,activation='relu')(hidden)\n",
        "hidden=keras.layers.Dropout(0.3)(hidden)\n",
        "output=keras.layers.Dense(8,activation='softmax')(hidden)\n",
        "\n",
        "net=keras.Model([input_mfcc,input_croma,input_mSpec], output, name=\"Net\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "YidWezft4ARY",
        "outputId": "1c21f09e-89a0-4824-9886-f63631594eae"
      },
      "cell_type": "code",
      "source": [
        "net.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Model: \"Net\"\n__________________________________________________________________________________________________\nLayer (type)                    Output Shape         Param #     Connected to                     \n==================================================================================================\nmfcc (InputLayer)               [(None, 20, 259, 1)] 0                                            \n__________________________________________________________________________________________________\ncroma (InputLayer)              [(None, 12, 259, 1)] 0                                            \n__________________________________________________________________________________________________\nmspec (InputLayer)              [(None, 128, 259, 1) 0                                            \n__________________________________________________________________________________________________\nmfccModel (Functional)          (None, 128)          94560       mfcc[0][0]                       \n__________________________________________________________________________________________________\ncromaModel (Functional)         (None, 128)          53120       croma[0][0]                      \n__________________________________________________________________________________________________\nmSpecModel (Functional)         (None, 128)          94560       mspec[0][0]                      \n__________________________________________________________________________________________________\nconcatenate (Concatenate)       (None, 384)          0           mfccModel[0][0]                  \n                                                                 cromaModel[0][0]                 \n                                                                 mSpecModel[0][0]                 \n__________________________________________________________________________________________________\ndense (Dense)                   (None, 50)           19250       concatenate[0][0]                \n__________________________________________________________________________________________________\ndropout_1 (Dropout)             (None, 50)           0           dense[0][0]                      \n__________________________________________________________________________________________________\ndense_1 (Dense)                 (None, 25)           1275        dropout_1[0][0]                  \n__________________________________________________________________________________________________\ndropout_2 (Dropout)             (None, 25)           0           dense_1[0][0]                    \n__________________________________________________________________________________________________\ndense_2 (Dense)                 (None, 8)            208         dropout_2[0][0]                  \n==================================================================================================\nTotal params: 262,973\nTrainable params: 261,245\nNon-trainable params: 1,728\n__________________________________________________________________________________________________\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "1Wxp27tE4ARY",
        "outputId": "50ece498-bbbe-42d9-a455-4546079af5cd"
      },
      "cell_type": "code",
      "source": [
        "keras.utils.plot_model(net, \"net.png\", show_shapes=True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 18,
          "data": {
            "image/png": "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\n",
            "text/plain": "<IPython.core.display.Image object>"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "UDZPDRuT4ARY"
      },
      "cell_type": "code",
      "source": [
        "accuracy='sparse_categorical_accuracy'\n",
        "sparseLoss=keras.losses.SparseCategoricalCrossentropy()\n",
        "\n",
        "from keras import backend as K\n",
        "K.clear_session()\n",
        "net.compile(optimizer='nadam', loss=sparseLoss,metrics=[accuracy])\n",
        "K.set_value(net.optimizer.learning_rate, 0.001)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "-1n54p0Y4ARY"
      },
      "cell_type": "code",
      "source": [
        "history=net.fit(\n",
        "    {\"mfcc\":mfcc_train,\"croma\":cstft_train,\"mspec\":mSpec_train},\n",
        "    ytrain,\n",
        "    validation_data=({\"mfcc\":mfcc_val,\"croma\":cstft_val,\"mspec\":mSpec_val},yval),\n",
        "    epochs=100,verbose=0,\n",
        "    callbacks=my_callbacks\n",
        ")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "KcFrPNBf4ARY",
        "outputId": "c41cfe0e-f367-4131-e109-019d10cb1bb6"
      },
      "cell_type": "code",
      "source": [
        "pd.DataFrame(history.history).plot()\n",
        "plt.grid(True)\n",
        "plt.gca().set_ylim(-0.1,1.1)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": "<Figure size 432x288 with 1 Axes>",
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "YE2FRl1n4ARY",
        "outputId": "233bbade-5135-4356-b3c5-e0801c988290"
      },
      "cell_type": "code",
      "source": [
        "net.evaluate({\"mfcc\":mfcc_val,\"croma\":cstft_val,\"mspec\":mSpec_val},yval)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": "54/54 [==============================] - 3s 49ms/step - loss: 0.1729 - sparse_categorical_accuracy: 0.9449\n",
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "execution_count": 82,
          "data": {
            "text/plain": "[0.17292781174182892, 0.9449275135993958]"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "trusted": true,
        "id": "-8esvOHk4ARY"
      },
      "cell_type": "code",
      "source": [
        "model=keras.models.load_model('./hbNet')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "27w--dxw4ARZ"
      },
      "cell_type": "code",
      "source": [
        "model.compile(optimizer='nadam', loss=sparseLoss,metrics=[accuracy])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "metadata": {
        "trusted": true,
        "id": "Pz9PhyB24ARZ",
        "outputId": "a9921017-6b0e-49e3-dea4-579544328c2b"
      },
      "cell_type": "code",
      "source": [
        "loss, acc = model.evaluate({\"mfcc\":mfcc_val,\"croma\":cstft_val,\"mspec\":mSpec_val},yval, verbose=2)\n",
        "print('Restored model, accuracy: {:5.2f}%'.format(100*acc))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": "54/54 - 3s - loss: 0.1660 - sparse_categorical_accuracy: 0.9443\nRestored model, accuracy: 94.43%\n",
          "name": "stdout"
        }
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3",
      "language": "python"
    },
    "language_info": {
      "name": "python",
      "version": "3.7.6",
      "mimetype": "text/x-python",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "pygments_lexer": "ipython3",
      "nbconvert_exporter": "python",
      "file_extension": ".py"
    },
    "colab": {
      "provenance": []
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}