Classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "mQ-9Vdq6TiBM"
      },
      "outputs": [],
      "source": [
        "# import necessary libraries\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "from sklearn.model_selection import train_test_split\n",
        "from keras.utils import to_categorical\n",
        "from keras.models import Model\n",
        "from keras.layers import Input, Dense, Dropout, LeakyReLU, Add\n",
        "from keras.layers import Conv1D, Flatten, BatchNormalization, LocallyConnected1D \n",
        "from keras.optimizers import SGD\n",
        "from sklearn.metrics import accuracy_score, confusion_matrix, ConfusionMatrixDisplay\n",
        "import matplotlib.pyplot as plt"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# load the US8K dataset\n",
        "features = np.load('./drive/MyDrive/dataset_US.npy',allow_pickle= True)\n",
        "# load the data labels from the metadata file \n",
        "df = pd.read_csv('./drive/MyDrive/UrbanSound8K.csv', usecols=['classID'], dtype={'classID': np.int32})\n",
        "data_class = df.to_numpy().flatten()\n",
        "# split the data into testset and trainset\n",
        "X_train, X_test, y_train, y_test = train_test_split(features, data_class, test_size=0.1, random_state=75)\n",
        "# reshape the data \n",
        "x_tr = X_train.reshape(X_train.shape[0], X_train.shape[1], 1)\n",
        "x_ts = X_test.reshape(X_test.shape[0], X_test.shape[1], 1)"
      ],
      "metadata": {
        "id": "9Rvq-MMNq796"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# describe the neural network\n",
        "def ConvBlock(inputs, filters, kernel_size):\n",
        "    x = Conv1D(filters, kernel_size, padding='same')(inputs)\n",
        "    x = BatchNormalization()(x)\n",
        "    x = LeakyReLU(alpha=0.05)(x)\n",
        "    x = Conv1D(filters, kernel_size, padding='same')(x)\n",
        "    x = BatchNormalization()(x)\n",
        "    x = LeakyReLU(alpha=0.05)(x)\n",
        "    return x\n",
        "\n",
        "def LocalBlock(inputs, filters, kernel_size):\n",
        "    x = LocallyConnected1D(filters, kernel_size)(inputs)\n",
        "    x = BatchNormalization()(x)\n",
        "    x = LeakyReLU(alpha=0.05)(x)\n",
        "    x = Conv1D(filters, kernel_size, padding='same')(x)\n",
        "    x = BatchNormalization()(x)\n",
        "    x = LeakyReLU(alpha=0.05)(x)\n",
        "    return x\n",
        "\n",
        "def DenseBlock(inputs, neuron_count):\n",
        "    x = Dense(neuron_count)(inputs)\n",
        "    x = LeakyReLU(alpha=0.05)(x)\n",
        "    x = Dropout(0.5)(x)\n",
        "    return x\n",
        "\n",
        "# specify the NN's input\n",
        "f_inp = Input(shape=(features.shape[1],1))\n",
        "\n",
        "# define four parallel streams for feature extraction\n",
        "x_0 = ConvBlock(f_inp, 64, 3)\n",
        "x_0 = LocalBlock(x_0, 16, 3)\n",
        "\n",
        "x_1 = ConvBlock(f_inp, 64, 7)\n",
        "x_1 = LocalBlock(x_1, 16, 3)\n",
        "\n",
        "x_2 = ConvBlock(f_inp, 64, 11)\n",
        "x_2 = LocalBlock(x_2, 16, 3)\n",
        "\n",
        "x_3 = ConvBlock(f_inp, 64, 17)\n",
        "x_3 = LocalBlock(x_3, 16, 3)\n",
        "\n",
        "x_4 = ConvBlock(f_inp, 64, 23)\n",
        "x_4 = LocalBlock(x_4, 16, 3)\n",
        "\n",
        "# merge the streams via addition\n",
        "x = Add()([x_0, x_1, x_2, x_3, x_4])\n",
        "\n",
        "# extract the final feature map\n",
        "x = LocalBlock(x, 32, 3)\n",
        "\n",
        "# flatten the feature map           \n",
        "x = Flatten()(x)\n",
        "\n",
        "# classify the data\n",
        "x = DenseBlock(x, 200)\n",
        "x_out = Dense(10, activation='softmax')(x)"
      ],
      "metadata": {
        "id": "xKIi6Ex8JjAC"
      },
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# NN instantiation\n",
        "model = Model(inputs=f_inp, outputs=x_out)\n",
        "\n",
        "# load model weights  \n",
        "model.load_weights('./model')\n",
        "\n",
        "# test the NN\n",
        "pred = model.predict(x_ts)\n",
        "pred_y = pred.argmax(axis=-1)\n",
        "\n",
        "# report the classification accuracy\n",
        "print(f\"accuracy = {accuracy_score(y_test, pred_y)}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EkxOP2AnQRDU",
        "outputId": "75cc4cfe-5ad1-4102-8437-c62c7b28bf68"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "28/28 [==============================] - 56s 48ms/step\n",
            "accuracy = 0.9805491990846682\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# display the confusion matrix\n",
        "con_matrix = confusion_matrix(y_test, pred_y)\n",
        "cmn = con_matrix.astype('float') / con_matrix.sum(axis=1)[:, np.newaxis]\n",
        "fig, ax = plt.subplots(figsize=(5,5))\n",
        "cmd = ConfusionMatrixDisplay(confusion_matrix=cmn)\n",
        "cmd.plot(ax=ax)\n",
        "# save the confusion matrix\n",
        "plt.savefig(\"cm.png\", format='png', dpi=600, bbox_inches='tight')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 313
        },
        "id": "2K1RrJofKI6t",
        "outputId": "b5de8ccd-656f-4d29-80da-f57bda6fe813"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 360x360 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}