music_genre_classification.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "music_genre_classification.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "urkaWe-kSPNy",
        "colab_type": "code",
        "outputId": "980681fa-eacd-407b-9daa-5a9bbde1cc23",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 72
        }
      },
      "source": [
        "!pip install -q tensorflow==2.0.0-beta1"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[K     |████████████████████████████████| 87.9MB 570kB/s \n",
            "\u001b[K     |████████████████████████████████| 501kB 56.8MB/s \n",
            "\u001b[K     |████████████████████████████████| 3.1MB 49.3MB/s \n",
            "\u001b[?25h"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EJxH9yutSfuA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "import librosa\n",
        "import librosa.display\n",
        "import matplotlib.pyplot as plt\n",
        "import pandas as pd\n",
        "import os"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9x29knKvSpyC",
        "colab_type": "code",
        "outputId": "560365c6-caec-4461-d087-029513bdb677",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "tf.__version__"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'2.0.0-beta1'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "S7zr0vAaTfZG",
        "colab_type": "code",
        "outputId": "cfe7e615-fa72-453a-cbb0-309bd15bb29d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "data_root = tf.keras.utils.get_file('Audio', 'https://os.unil.cloud.switch.ch/fma/fma_small.zip' , extract=True)"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://os.unil.cloud.switch.ch/fma/fma_small.zip\n",
            "7679598592/7679594875 [==============================] - 285s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T2R_yQcKXlNB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!cp -r ~/.keras/datasets/fma_small /content"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RaJXAtnLxjNO",
        "colab_type": "code",
        "outputId": "ac8b421b-b22c-4279-cead-e24dcfbc3bf3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "meta_data = tf.keras.utils.get_file('Meta' , 'https://os.unil.cloud.switch.ch/fma/fma_metadata.zip' , extract=True)"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://os.unil.cloud.switch.ch/fma/fma_metadata.zip\n",
            "358416384/358412441 [==============================] - 14s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4D9JJexGcVDA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!cp ~/.keras/datasets/fma_metadata/tracks.csv /content"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mT1ga4uncbM0",
        "colab_type": "code",
        "outputId": "b74b24e3-a84f-4d37-afbc-53bb1c1f524c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 196
        }
      },
      "source": [
        "genres_df = pd.read_csv('genres.csv')\n",
        "genres_df.head()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "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>genre_id</th>\n",
              "      <th>#tracks</th>\n",
              "      <th>parent</th>\n",
              "      <th>title</th>\n",
              "      <th>top_level</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1</td>\n",
              "      <td>8693</td>\n",
              "      <td>38</td>\n",
              "      <td>Avant-Garde</td>\n",
              "      <td>38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>2</td>\n",
              "      <td>5271</td>\n",
              "      <td>0</td>\n",
              "      <td>International</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>3</td>\n",
              "      <td>1752</td>\n",
              "      <td>0</td>\n",
              "      <td>Blues</td>\n",
              "      <td>3</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>4</td>\n",
              "      <td>4126</td>\n",
              "      <td>0</td>\n",
              "      <td>Jazz</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>5</td>\n",
              "      <td>4106</td>\n",
              "      <td>0</td>\n",
              "      <td>Classical</td>\n",
              "      <td>5</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   genre_id  #tracks  parent          title  top_level\n",
              "0         1     8693      38    Avant-Garde         38\n",
              "1         2     5271       0  International          2\n",
              "2         3     1752       0          Blues          3\n",
              "3         4     4126       0           Jazz          4\n",
              "4         5     4106       0      Classical          5"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T6wv_dVmdRtA",
        "colab_type": "code",
        "outputId": "b25ac38e-8724-4f30-cbfc-6b530b119b42",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 258
        }
      },
      "source": [
        "tracks_df = pd.read_csv('tracks.csv' , index_col= 0 , header = [0,1])\n",
        "keep_cols = [('set', 'split'),\n",
        "('set', 'subset'),('track', 'genre_top')]\n",
        "\n",
        "df_all = tracks_df[keep_cols]\n",
        "df_all = df_all[df_all[('set', 'subset')] == 'small']\n",
        "\n",
        "df_all['track_id'] = df_all.index\n",
        "df_all.head()"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "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 tr th {\n",
              "        text-align: left;\n",
              "    }\n",
              "\n",
              "    .dataframe thead tr:last-of-type th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr>\n",
              "      <th></th>\n",
              "      <th colspan=\"2\" halign=\"left\">set</th>\n",
              "      <th>track</th>\n",
              "      <th>track_id</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th></th>\n",
              "      <th>split</th>\n",
              "      <th>subset</th>\n",
              "      <th>genre_top</th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>track_id</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>training</td>\n",
              "      <td>small</td>\n",
              "      <td>Hip-Hop</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>training</td>\n",
              "      <td>small</td>\n",
              "      <td>Hip-Hop</td>\n",
              "      <td>5</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10</th>\n",
              "      <td>training</td>\n",
              "      <td>small</td>\n",
              "      <td>Pop</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>140</th>\n",
              "      <td>training</td>\n",
              "      <td>small</td>\n",
              "      <td>Folk</td>\n",
              "      <td>140</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>141</th>\n",
              "      <td>training</td>\n",
              "      <td>small</td>\n",
              "      <td>Folk</td>\n",
              "      <td>141</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "               set            track track_id\n",
              "             split subset genre_top         \n",
              "track_id                                    \n",
              "2         training  small   Hip-Hop        2\n",
              "5         training  small   Hip-Hop        5\n",
              "10        training  small       Pop       10\n",
              "140       training  small      Folk      140\n",
              "141       training  small      Folk      141"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hFyFF7T4dcBO",
        "colab_type": "code",
        "outputId": "773a3ebc-d1dc-4068-e341-97e557a7bd27",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "df_all[('track' , 'genre_top')].unique()"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array(['Hip-Hop', 'Pop', 'Folk', 'Experimental', 'Rock', 'International',\n",
              "       'Electronic', 'Instrumental'], dtype=object)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "agxWfNGljQuP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def readAudioFile(trackid):\n",
        "  track = '{:06d}'.format(trackid)\n",
        "  return os.path.join('fma_small' , track[:3] , track + '.mp3')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MtoAijdWjMSc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def createSpectrogram(trackid):\n",
        "  audio = readAudioFile(trackid)\n",
        "  y, sr = librosa.load(audio , duration=2.97)\n",
        "  spect = librosa.feature.melspectrogram(y=y, sr=sr)\n",
        "  return spect"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hrUjpGP2kajj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def plotSpectrogram(trackid):\n",
        "  spectrogram = createSpectrogram(trackid)\n",
        "  print(spectrogram.shape)\n",
        "  librosa.display.specshow(spectrogram , y_axis='mel', x_axis='time')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C4Bg7jpJlLWe",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def getTracks(audio_dir):\n",
        "  tids = []\n",
        "  for _, dirnames, files in os.walk(audio_dir):\n",
        "      if dirnames == []:\n",
        "          tids.extend(int(file[:-4]) for file in files)\n",
        "  return tids"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "z4rSZPpclSkd",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "track_ids = getTracks('fma_small')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GN-zkCEpk2t8",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "plotSpectrogram(3400)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jpADQprkqZgq",
        "colab_type": "code",
        "outputId": "5d7ecd28-f465-4592-91c3-2a2794c9a63c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "df_train = df_all[df_all[('set', 'split')]=='training']\n",
        "df_valid = df_all[df_all[('set', 'split')]=='validation']\n",
        "df_test = df_all[df_all[('set', 'split')]=='test']\n",
        "\n",
        "print(df_train.shape, df_valid.shape, df_test.shape)"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(6400, 4) (800, 4) (800, 4)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XF5YTLbSiqXq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def create_dataset(df):\n",
        "  genres = []\n",
        "  X_spect = np.empty((0 , 128 , 128))\n",
        "  count = 0\n",
        "  for index, row in df.iterrows():\n",
        "    try:\n",
        "        count += 1\n",
        "        track_id = int(row['track_id'])\n",
        "        genre = str(row[('track', 'genre_top')])\n",
        "        spect = createSpectrogram(track_id)\n",
        "        X_spect = np.append(X_spect, [spect], axis=0)\n",
        "        genres.append(genre)\n",
        "        if count % 100 == 0:\n",
        "            print(\"Currently processing: \", count)\n",
        "    except:\n",
        "        print(\"Not processed : \" , count)\n",
        "        continue\n",
        "\n",
        "  y_arr = np.array(genres)\n",
        "  return X_spect, y_arr"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2M09vJ7dUu4O",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 163
        },
        "outputId": "2ed90cf8-8b38-4d7a-ad19-20e487e9f223"
      },
      "source": [
        "X_valid , y_valid = create_dataset(df_valid)"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Currently processing:  100\n",
            "Currently processing:  200\n",
            "Currently processing:  300\n",
            "Currently processing:  400\n",
            "Currently processing:  500\n",
            "Currently processing:  600\n",
            "Currently processing:  700\n",
            "Currently processing:  800\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vIW1dTX4q5we",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_train = create_dataset(df_train)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CbhFjgP_rQ4r",
        "colab_type": "code",
        "outputId": "e2c7e68e-d3ad-4926-8c4e-3c2f2e80206f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "X_train.shape"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(6394, 128, 128)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c9aggbz4rg9P",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "label_dict = {'Electronic':0, 'Experimental':1, 'Folk':2, 'Hip-Hop':3, \n",
        "               'Instrumental':4,'International':5, 'Pop' :6, 'Rock': 7  }"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MFiZQblesK41",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "genres = []"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1IGILcotqgKe",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for label in y_valid:\n",
        "  genres.append(label_dict[label])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y_5p_WbplA1E",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_valid = np.array(genres)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5pFF_4MCqnzI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "labels = np.array(genres)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t-9p1yunuHfb",
        "colab_type": "code",
        "outputId": "869fa652-312c-446b-8f6f-3c78470d8c60",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 33
        }
      },
      "source": [
        "labels.shape"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(6394,)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 51
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CUNjN3pDkvwN",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "3cadddd0-5a5d-444f-9aa5-45918879312d"
      },
      "source": [
        "y_valid.shape"
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(800,)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nqmzzlv4kzjC",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y_valid = tf.keras.utils.to_categorical(y_valid , 8)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0U6vmLaiuRlt",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "labels = tf.keras.utils.to_categorical(labels , 8)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N-TxrOP1uYp6",
        "colab_type": "code",
        "outputId": "02fe34f1-7886-4524-da02-eda06218860c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 33
        }
      },
      "source": [
        "labels[0]"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0., 0., 0., 1., 0., 0., 0., 0.], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 54
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UL71HkacvcLb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def getGenre(prob_vector):\n",
        "  value = np.argmax(prob_vector)\n",
        "  for genre , index in label_dict.items():\n",
        "    if index == value:\n",
        "      return genre"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3NrTDI_svutQ",
        "colab_type": "code",
        "outputId": "7c09e824-e1f2-4e75-8f39-c996595931e4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "print(getGenre([0., 0., 0., 1., 0., 0., 0., 0.]))"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Hip-Hop\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-9Ymzapj3x2I",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "labels = np.load('drive/My Drive/labels.npy')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RgpIPDb-w94K",
        "colab_type": "code",
        "outputId": "fbbdebf1-9dec-4d07-e37b-c781326f3766",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 728
        }
      },
      "source": [
        "z = 0\n",
        "fig , ax = plt.subplots(3 , 2 , figsize=(10 , 10))\n",
        "fig.tight_layout()\n",
        "for i in range(0,3):\n",
        "  for j in range(0,2):\n",
        "    plt.subplot(3 , 2 , z+1)\n",
        "    librosa.display.specshow(X_train[z])\n",
        "    plt.title(getGenre(labels[z]))\n",
        "    z = z + 1\n",
        "  "
      ],
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAArAAAALHCAYAAAB2Y3HPAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 720x720 with 6 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rxfBDeikwaPQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "np.save('spectrograms.npy' , X_train)\n",
        "np.save('labels.npy' , labels)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L37vvNO3EMgv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!cp labels.npy drive/My\\ Drive/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uJC-Zfcu1Nzh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X_valid = np.array([x.reshape((128 , 128 , 1)) for x in X_valid])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Tf9ClBoo2JmW",
        "colab_type": "code",
        "outputId": "b6f55af1-4c7b-4bdc-b55f-38cac450d376",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "X_valid.shape"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(800, 128, 128, 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SQucaoac2040",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = tf.keras.Sequential()\n",
        "input_shape = (128 , 128 , 1)\n",
        "\n",
        "model.add(tf.keras.layers.Conv2D(24 , (5 , 5) , strides=(1 , 1) , input_shape=input_shape))\n",
        "model.add(tf.keras.layers.MaxPooling2D((4,2) , strides=(4 , 2)))\n",
        "model.add(tf.keras.layers.Activation('relu'))\n",
        "\n",
        "model.add(tf.keras.layers.Conv2D(48 , (5 , 5) , padding='valid'))\n",
        "model.add(tf.keras.layers.MaxPooling2D((4 , 2) , strides=(4 , 2)))\n",
        "model.add(tf.keras.layers.Activation('relu'))\n",
        "\n",
        "model.add(tf.keras.layers.Conv2D(48 , (5 , 5) , padding='valid'))\n",
        "model.add(tf.keras.layers.Activation('relu'))\n",
        "\n",
        "model.add(tf.keras.layers.Flatten())\n",
        "model.add(tf.keras.layers.Dropout(0.5))\n",
        "\n",
        "model.add(tf.keras.layers.Dense(64))\n",
        "model.add(tf.keras.layers.Activation('relu'))\n",
        "model.add(tf.keras.layers.Dropout(0.5))\n",
        "\n",
        "model.add(tf.keras.layers.Dense(8))\n",
        "model.add(tf.keras.layers.Activation('softmax'))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Wbgb0Ij43Sb6",
        "colab_type": "code",
        "outputId": "52a980a5-b237-45d6-cbc7-cc27b13a3f4d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 708
        }
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 82,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_3 (Conv2D)            (None, 124, 124, 24)      624       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_2 (MaxPooling2 (None, 31, 62, 24)        0         \n",
            "_________________________________________________________________\n",
            "activation_5 (Activation)    (None, 31, 62, 24)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 27, 58, 48)        28848     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 6, 29, 48)         0         \n",
            "_________________________________________________________________\n",
            "activation_6 (Activation)    (None, 6, 29, 48)         0         \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 2, 25, 48)         57648     \n",
            "_________________________________________________________________\n",
            "activation_7 (Activation)    (None, 2, 25, 48)         0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 2400)              0         \n",
            "_________________________________________________________________\n",
            "dropout_2 (Dropout)          (None, 2400)              0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 64)                153664    \n",
            "_________________________________________________________________\n",
            "activation_8 (Activation)    (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dropout_3 (Dropout)          (None, 64)                0         \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 8)                 520       \n",
            "_________________________________________________________________\n",
            "activation_9 (Activation)    (None, 8)                 0         \n",
            "=================================================================\n",
            "Total params: 241,304\n",
            "Trainable params: 241,304\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TOoImYObWd3T",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "callbacks = [\n",
        "    tf.keras.callbacks.EarlyStopping(patience=10, verbose=1),\n",
        "    tf.keras.callbacks.ReduceLROnPlateau(factor=0.1, patience=10, min_lr=0.00001, verbose=1),\n",
        "    tf.keras.callbacks.ModelCheckpoint('model.h5', verbose=1, save_best_only=True, save_weights_only=True),\n",
        "]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xuVByM0U4-aP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.compile(\n",
        "    optimizer='Adam',\n",
        "    loss='categorical_crossentropy',\n",
        "    metrics = ['accuracy']\n",
        ")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ike-yRIi5TpY",
        "colab_type": "code",
        "outputId": "c2473d1e-eb2d-4a8c-ac33-572f8dcfd774",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "model.fit(\n",
        "    x=X_train,\n",
        "    y=labels,\n",
        "    epochs=50,\n",
        "    batch_size=128,\n",
        "    validation_data=(X_valid , y_valid),\n",
        "    callbacks=callbacks,\n",
        ")"
      ],
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: Logging before flag parsing goes to stderr.\n",
            "W0709 16:32:47.118226 140459465447296 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Train on 6394 samples, validate on 800 samples\n",
            "Epoch 1/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 3.4842 - accuracy: 0.1440\n",
            "Epoch 00001: val_loss improved from inf to 2.07237, saving model to model.h5\n",
            "6394/6394 [==============================] - 86s 13ms/sample - loss: 3.4572 - accuracy: 0.1444 - val_loss: 2.0724 - val_accuracy: 0.1400\n",
            "Epoch 2/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 2.0815 - accuracy: 0.1617\n",
            "Epoch 00002: val_loss improved from 2.07237 to 2.05376, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 2.0811 - accuracy: 0.1623 - val_loss: 2.0538 - val_accuracy: 0.1863\n",
            "Epoch 3/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 2.0559 - accuracy: 0.1819\n",
            "Epoch 00003: val_loss improved from 2.05376 to 2.01450, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 2.0540 - accuracy: 0.1836 - val_loss: 2.0145 - val_accuracy: 0.2138\n",
            "Epoch 4/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 2.0166 - accuracy: 0.2090\n",
            "Epoch 00004: val_loss improved from 2.01450 to 1.96949, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 2.0167 - accuracy: 0.2091 - val_loss: 1.9695 - val_accuracy: 0.2463\n",
            "Epoch 5/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.9666 - accuracy: 0.2298\n",
            "Epoch 00005: val_loss improved from 1.96949 to 1.93177, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.9669 - accuracy: 0.2293 - val_loss: 1.9318 - val_accuracy: 0.2475\n",
            "Epoch 6/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.9348 - accuracy: 0.2522\n",
            "Epoch 00006: val_loss improved from 1.93177 to 1.85219, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.9345 - accuracy: 0.2520 - val_loss: 1.8522 - val_accuracy: 0.2625\n",
            "Epoch 7/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.8826 - accuracy: 0.2757\n",
            "Epoch 00007: val_loss improved from 1.85219 to 1.82609, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.8814 - accuracy: 0.2751 - val_loss: 1.8261 - val_accuracy: 0.2675\n",
            "Epoch 8/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.8587 - accuracy: 0.2986\n",
            "Epoch 00008: val_loss improved from 1.82609 to 1.81861, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.8581 - accuracy: 0.2982 - val_loss: 1.8186 - val_accuracy: 0.2950\n",
            "Epoch 9/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.8324 - accuracy: 0.3036\n",
            "Epoch 00009: val_loss did not improve from 1.81861\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.8311 - accuracy: 0.3033 - val_loss: 1.8360 - val_accuracy: 0.2612\n",
            "Epoch 10/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.8043 - accuracy: 0.3160\n",
            "Epoch 00010: val_loss improved from 1.81861 to 1.80956, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.8042 - accuracy: 0.3158 - val_loss: 1.8096 - val_accuracy: 0.2688\n",
            "Epoch 11/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.7895 - accuracy: 0.3304\n",
            "Epoch 00011: val_loss did not improve from 1.80956\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.7891 - accuracy: 0.3297 - val_loss: 1.8189 - val_accuracy: 0.2875\n",
            "Epoch 12/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.7603 - accuracy: 0.3367\n",
            "Epoch 00012: val_loss improved from 1.80956 to 1.75153, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.7595 - accuracy: 0.3373 - val_loss: 1.7515 - val_accuracy: 0.3350\n",
            "Epoch 13/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.7367 - accuracy: 0.3512\n",
            "Epoch 00013: val_loss did not improve from 1.75153\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.7344 - accuracy: 0.3511 - val_loss: 1.7545 - val_accuracy: 0.3400\n",
            "Epoch 14/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.6942 - accuracy: 0.3678\n",
            "Epoch 00014: val_loss improved from 1.75153 to 1.72956, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.6935 - accuracy: 0.3675 - val_loss: 1.7296 - val_accuracy: 0.3438\n",
            "Epoch 15/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.6887 - accuracy: 0.3831\n",
            "Epoch 00015: val_loss did not improve from 1.72956\n",
            "6394/6394 [==============================] - 83s 13ms/sample - loss: 1.6886 - accuracy: 0.3827 - val_loss: 1.7670 - val_accuracy: 0.3438\n",
            "Epoch 16/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.6542 - accuracy: 0.3870\n",
            "Epoch 00016: val_loss did not improve from 1.72956\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.6574 - accuracy: 0.3844 - val_loss: 1.8137 - val_accuracy: 0.2975\n",
            "Epoch 17/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.6370 - accuracy: 0.3954\n",
            "Epoch 00017: val_loss did not improve from 1.72956\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.6367 - accuracy: 0.3944 - val_loss: 1.8152 - val_accuracy: 0.3413\n",
            "Epoch 18/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.6651 - accuracy: 0.3870\n",
            "Epoch 00018: val_loss improved from 1.72956 to 1.70977, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.6649 - accuracy: 0.3863 - val_loss: 1.7098 - val_accuracy: 0.3837\n",
            "Epoch 19/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.6069 - accuracy: 0.4072\n",
            "Epoch 00019: val_loss did not improve from 1.70977\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.6079 - accuracy: 0.4065 - val_loss: 1.7220 - val_accuracy: 0.3887\n",
            "Epoch 20/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.5734 - accuracy: 0.4190\n",
            "Epoch 00020: val_loss did not improve from 1.70977\n",
            "6394/6394 [==============================] - 83s 13ms/sample - loss: 1.5734 - accuracy: 0.4185 - val_loss: 1.7165 - val_accuracy: 0.4087\n",
            "Epoch 21/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.5578 - accuracy: 0.4267\n",
            "Epoch 00021: val_loss improved from 1.70977 to 1.69949, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.5589 - accuracy: 0.4266 - val_loss: 1.6995 - val_accuracy: 0.3487\n",
            "Epoch 22/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.5546 - accuracy: 0.4311\n",
            "Epoch 00022: val_loss did not improve from 1.69949\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.5544 - accuracy: 0.4318 - val_loss: 1.7201 - val_accuracy: 0.3725\n",
            "Epoch 23/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.5071 - accuracy: 0.4482\n",
            "Epoch 00023: val_loss did not improve from 1.69949\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.5100 - accuracy: 0.4471 - val_loss: 1.7360 - val_accuracy: 0.4013\n",
            "Epoch 24/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.5005 - accuracy: 0.4544\n",
            "Epoch 00024: val_loss did not improve from 1.69949\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.5006 - accuracy: 0.4536 - val_loss: 1.7390 - val_accuracy: 0.3925\n",
            "Epoch 25/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.4756 - accuracy: 0.4692\n",
            "Epoch 00025: val_loss improved from 1.69949 to 1.68352, saving model to model.h5\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.4736 - accuracy: 0.4698 - val_loss: 1.6835 - val_accuracy: 0.3925\n",
            "Epoch 26/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.4767 - accuracy: 0.4590\n",
            "Epoch 00026: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.4795 - accuracy: 0.4587 - val_loss: 1.7334 - val_accuracy: 0.4000\n",
            "Epoch 27/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.4679 - accuracy: 0.4664\n",
            "Epoch 00027: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.4666 - accuracy: 0.4664 - val_loss: 1.7067 - val_accuracy: 0.4250\n",
            "Epoch 28/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.4269 - accuracy: 0.4845\n",
            "Epoch 00028: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.4271 - accuracy: 0.4837 - val_loss: 1.7279 - val_accuracy: 0.3925\n",
            "Epoch 29/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.4218 - accuracy: 0.4809\n",
            "Epoch 00029: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.4208 - accuracy: 0.4808 - val_loss: 1.7528 - val_accuracy: 0.4000\n",
            "Epoch 30/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.4138 - accuracy: 0.4850\n",
            "Epoch 00030: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.4171 - accuracy: 0.4847 - val_loss: 1.7557 - val_accuracy: 0.3925\n",
            "Epoch 31/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.3787 - accuracy: 0.4973\n",
            "Epoch 00031: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.3800 - accuracy: 0.4966 - val_loss: 1.7803 - val_accuracy: 0.3650\n",
            "Epoch 32/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.3743 - accuracy: 0.5022\n",
            "Epoch 00032: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.3765 - accuracy: 0.5016 - val_loss: 1.7955 - val_accuracy: 0.4075\n",
            "Epoch 33/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.3417 - accuracy: 0.5118\n",
            "Epoch 00033: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.3446 - accuracy: 0.5099 - val_loss: 1.7643 - val_accuracy: 0.4263\n",
            "Epoch 34/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.3444 - accuracy: 0.5172\n",
            "Epoch 00034: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.3448 - accuracy: 0.5164 - val_loss: 1.7307 - val_accuracy: 0.4125\n",
            "Epoch 35/50\n",
            "6272/6394 [============================>.] - ETA: 1s - loss: 1.3304 - accuracy: 0.5254\n",
            "Epoch 00035: ReduceLROnPlateau reducing learning rate to 0.00010000000474974513.\n",
            "\n",
            "Epoch 00035: val_loss did not improve from 1.68352\n",
            "6394/6394 [==============================] - 84s 13ms/sample - loss: 1.3286 - accuracy: 0.5261 - val_loss: 1.7977 - val_accuracy: 0.4125\n",
            "Epoch 00035: early stopping\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7fbef39f47f0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 48
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "huECGjnR-3Jy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "  for index , row in df_test.iterrows():\n",
        "    track_id = int(row['track_id'])\n",
        "    gen1 = str(row[('track', 'genre_top')])\n",
        "    spect1 = createSpectrogram(track_id)\n",
        "    break"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p7Ys369N_bCZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test = np.array(specter).reshape((128 , 128 , 1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hwyUYKr6__r7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test1 = np.expand_dims(test , axis=0)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RwAODiCmAJum",
        "colab_type": "code",
        "outputId": "21e0c3d4-dc71-4bf8-b2f9-1ddf872ae755",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "test1.shape"
      ],
      "execution_count": 170,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(1, 128, 128, 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 170
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KUVYTCWX5nHW",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "res = model.predict(test2)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iL7VxkPE_8TO",
        "colab_type": "code",
        "outputId": "329fcfcc-3bdf-4a3a-ee72-27be7b335361",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        }
      },
      "source": [
        "getGenre(res)"
      ],
      "execution_count": 197,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Pop'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 197
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "g1RBNEHjC9cK",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def getGenreFromValue(value):\n",
        "  for i , j in label_dict.items():\n",
        "    if j == value:\n",
        "      return i\n",
        "    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "03YEq9NJAPSc",
        "colab_type": "code",
        "outputId": "2e57dc84-6130-4c68-a0b3-bd2d396f6e27",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 308
        }
      },
      "source": [
        "for i in range(0,8):\n",
        "  print('{} : {}\\n'.format(getGenreFromValue(i) , res[0][i]))"
      ],
      "execution_count": 198,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Electronic : 0.09282361716032028\n",
            "\n",
            "Experimental : 0.04594282805919647\n",
            "\n",
            "Folk : 0.08038511127233505\n",
            "\n",
            "Hip-Hop : 0.051831576973199844\n",
            "\n",
            "Instrumental : 0.02783791720867157\n",
            "\n",
            "International : 0.10592349618673325\n",
            "\n",
            "Pop : 0.3225657045841217\n",
            "\n",
            "Rock : 0.2726898193359375\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LXhjDvTkAVVD",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "y, sr = librosa.load('Coldplay - Hymn For The Weekend (Official Video).mp3' , offset=120.0 , duration=2.97)\n",
        "specter = librosa.feature.melspectrogram(y=y, sr=sr)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6Dq3dUrjBUun",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.save('genre-model.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iRdID1qY4T21",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = tf.keras.models.load_model('genre-model.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HgrtJu_34og2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model.load_weights('model.h5')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BDS38YI26RZJ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "ab872e50-8925-40e9-86f4-b0635e0995b2"
      },
      "source": [
        "model.outputs"
      ],
      "execution_count": 178,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<tf.Tensor 'activation_4_2/Identity:0' shape=(None, 8) dtype=float32>]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 178
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YJJENlB2FS6-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "tf.saved_model.save(model , \"/tmp/genre/1/\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-4Ir-HT9Fdgn",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 217
        },
        "outputId": "d470c467-3a85-42e3-8fe2-6ad6146e2a51"
      },
      "source": [
        "!saved_model_cli show --dir /tmp/genre/1 --tag_set serve --signature_def serving_default"
      ],
      "execution_count": 185,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "The given SavedModel SignatureDef contains the following input(s):\n",
            "  inputs['conv2d_input'] tensor_info:\n",
            "      dtype: DT_FLOAT\n",
            "      shape: (-1, 128, 128, 1)\n",
            "      name: serving_default_conv2d_input:0\n",
            "The given SavedModel SignatureDef contains the following output(s):\n",
            "  outputs['activation_4'] tensor_info:\n",
            "      dtype: DT_FLOAT\n",
            "      shape: (-1, 8)\n",
            "      name: StatefulPartitionedCall:0\n",
            "Method name is: tensorflow/serving/predict\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "79uG1cTbHAn1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "173928e5-2372-4f71-f6d0-1be1de455014"
      },
      "source": [
        "loaded = tf.saved_model.load(\"/tmp/genre/1/\")\n",
        "print(list(loaded.signatures.keys())) "
      ],
      "execution_count": 186,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['serving_default']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HMhsXAfpHM1Q",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "b5ef5929-4903-4668-e256-03c997c05676"
      },
      "source": [
        "infer = loaded.signatures[\"serving_default\"]\n",
        "print(infer.structured_outputs)"
      ],
      "execution_count": 187,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "{'activation_4': TensorSpec(shape=(None, 8), dtype=tf.float32, name='activation_4')}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TKbE78QUHtQo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "11bbc14c-daed-468e-d807-aabe1fb55e73"
      },
      "source": [
        "model.output_names[0]"
      ],
      "execution_count": 188,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'activation_4'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 188
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Q3blYzRcIJ36",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "test2 = test1.astype('float32')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZAQFC0RFIZUX",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "7961a580-dbf9-4cc7-cb7c-72b7a200c0f7"
      },
      "source": [
        "test2.shape"
      ],
      "execution_count": 209,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(1, 128, 128, 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 209
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gBjDJ5YWHQgQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "labeling = infer(tf.constant(test2))[model.output_names[0]]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UjLttAzgHwLd",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "5706bf63-6a0e-4b18-a472-77e640afacd5"
      },
      "source": [
        "getGenre(labeling)"
      ],
      "execution_count": 211,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "'Pop'"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 211
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9nSgxfLXIuxy",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!nohup tensorflow_model_server \\\n",
        "  --rest_api_port=8501 \\\n",
        "  --model_name=genre \\\n",
        "  --model_base_path=\"/tmp/genre\" >server.log 2>&1"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pu9ZWO-MI8HV",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!pip install -q requests\n",
        "import json\n",
        "import numpy\n",
        "import requests\n",
        "data = json.dumps({\"signature_name\": \"serving_default\",\n",
        "                   \"instances\": test2.tolist()})\n",
        "headers = {\"content-type\": \"application/json\"}\n",
        "json_response = requests.post('http://localhost:8501/v1/models/genre:predict',\n",
        "                              data=data, headers=headers)\n",
        "predictions = numpy.array(json.loads(json_response.text)[\"predictions\"])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QotL1ZYtLMCF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import subprocess"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DTtq4Ft8JVcI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!echo \"deb http://storage.googleapis.com/tensorflow-serving-apt stable tensorflow-model-server tensorflow-model-server-universal\" | tee /etc/apt/sources.list.d/tensorflow-serving.list && curl https://storage.googleapis.com/tensorflow-serving-apt/tensorflow-serving.release.pub.gpg | apt-key add -\n",
        "!apt update"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sOZHhMvLJ-gr",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 90
        },
        "outputId": "34bfcfbf-211b-48be-d06c-476ebdd9d5ba"
      },
      "source": [
        "!apt-get install tensorflow-model-server\n"
      ],
      "execution_count": 216,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Reading package lists... Done\n",
            "Building dependency tree       \n",
            "Reading state information... Done\n",
            "E: Unable to locate package tensorflow-model-server\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oYb50na-KIdE",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!cp -r /tmp /content/drive/My\\ Drive/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d5wFo3hDMs04",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}