1_data_preprocessing.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5yvxhl-D_cJz"
      },
      "source": [
        "1. <u>Imports and downloads</u>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "tCtU9dvp_cmf"
      },
      "outputs": [],
      "source": [
        "#!pip install mido\n",
        "#!pip install midi2audio\n",
        "#!pip install google-cloud-storage\n",
        "#!apt-get install -y fluidsynth"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "YCdqOmjI_eSj"
      },
      "outputs": [],
      "source": [
        "# Imports\n",
        "\n",
        "# Calc\n",
        "import numpy as np\n",
        "import random\n",
        "import pickle\n",
        "\n",
        "# Audio\n",
        "from mido import MidiFile, MidiTrack, Message\n",
        "from midi2audio import FluidSynth\n",
        "import librosa\n",
        "import pygame\n",
        "import mido\n",
        "\n",
        "# Image\n",
        "from PIL import Image\n",
        "\n",
        "# Visualisation\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# Cloud\n",
        "from io import BytesIO\n",
        "from google.colab import auth\n",
        "from google.cloud import storage\n",
        "\n",
        "# File management\n",
        "import os\n",
        "import tempfile"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3s-Ub7Dn_fv_"
      },
      "outputs": [],
      "source": [
        "# Cloud authentification\n",
        "\n",
        "auth.authenticate_user()\n",
        "client = storage.Client()\n",
        "bucket = client.get_bucket('piano-transcription')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "qjim5dJa_mvn"
      },
      "outputs": [],
      "source": [
        "# Useful directories & paths\n",
        "\n",
        "soundfont_path = 'soundfont/FluidR3_GM.sf2'\n",
        "midi_directory = 'midi'\n",
        "wav_directory = 'wav'\n",
        "spectrograms_directory = 'spectrograms'"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CKUXpRcuuAvC"
      },
      "source": [
        "<u>2. Useful functions</u>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mBEb0_5tTFUE"
      },
      "outputs": [],
      "source": [
        "# Manage file storage on GCS\n",
        "\n",
        "def save_pkl_to_gcp(data_list, destination_blob):\n",
        "    \"\"\"\n",
        "    Save a a .pkl file to GCS.\n",
        "    \"\"\"\n",
        "\n",
        "    pkl_list = pickle.dumps(data_list)\n",
        "    blob = bucket.blob(destination_blob)\n",
        "    blob.upload_from_string(pkl_list, content_type = 'application/octet-stream')\n",
        "\n",
        "    print(f\"File {destination_blob} uploaded to GCS.\")\n",
        "    return None\n",
        "\n",
        "def load_pkl_from_gcp(source_blob):\n",
        "    \"\"\"\n",
        "    Load a .pkl file stored in GCS.\n",
        "    \"\"\"\n",
        "\n",
        "    blob = bucket.blob(source_blob)\n",
        "    pkl_data = blob.download_as_string()\n",
        "    data_list = pickle.loads(pkl_data)\n",
        "\n",
        "    print(f\"File {source_blob} loaded from GCS.\")\n",
        "    return data_list\n",
        "\n",
        "def download_blob_to_temp_file(blob_name):\n",
        "    \"\"\"\n",
        "    Download a file from GCS to a local temp file.\n",
        "    \"\"\"\n",
        "    with tempfile.NamedTemporaryFile(delete = False) as temp_file:\n",
        "        blob = bucket.blob(blob_name)\n",
        "        blob.download_to_filename(temp_file.name)\n",
        "    return temp_file.name\n",
        "\n",
        "def upload_file_to_bucket(local_file_path, blob_name):\n",
        "    \"\"\"\n",
        "    Upload a local file to GCS.\n",
        "    \"\"\"\n",
        "    blob = bucket.blob(blob_name)\n",
        "    blob.upload_from_filename(local_file_path)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Q1FqBj8k_tmC"
      },
      "outputs": [],
      "source": [
        "# From midi (.mid) to audio (.wav)\n",
        "\n",
        "def convert_midi_to_wav(midi_directory = midi_directory, wav_directory = wav_directory, soundfont_path = soundfont_path):\n",
        "\n",
        "    \"\"\"\n",
        "    Convert an midi file (.mid) into an audio file (.wav)\n",
        "    \"\"\"\n",
        "\n",
        "    soundfont_local_path = download_blob_to_temp_file(soundfont_path)\n",
        "    fs = FluidSynth(soundfont_local_path)\n",
        "\n",
        "    midi_blobs = sorted([blob.name for blob in bucket.list_blobs(prefix = midi_directory) if blob.name.endswith('.mid')])\n",
        "\n",
        "    for midi_blob in midi_blobs:\n",
        "\n",
        "        midi_local_path = download_blob_to_temp_file(midi_blob)\n",
        "        wav_blob_name = midi_blob.replace(midi_directory, wav_directory).replace('.mid', '.wav')\n",
        "\n",
        "        with tempfile.NamedTemporaryFile(delete=False, suffix='.wav') as wav_temp_file:\n",
        "            wav_local_path = wav_temp_file.name\n",
        "            fs.midi_to_audio(midi_local_path, wav_local_path)\n",
        "\n",
        "        upload_file_to_bucket(wav_local_path, wav_blob_name)\n",
        "\n",
        "        os.remove(midi_local_path)\n",
        "        os.remove(wav_local_path)\n",
        "\n",
        "    os.remove(soundfont_local_path)\n",
        "\n",
        "    return None"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "TPWTRGQX_vSr"
      },
      "outputs": [],
      "source": [
        "# From audio (.wav) to spectrogram (.png)\n",
        "\n",
        "def from_audio_to_spectrograms(wav_directory = wav_directory, spectrograms_directory = spectrograms_directory):\n",
        "\n",
        "    \"\"\"\n",
        "    Convert an audio file (.wav) into a CQT spectrogram (.png)\n",
        "    \"\"\"\n",
        "\n",
        "    wav_blobs = sorted([blob.name for blob in bucket.list_blobs(prefix = wav_directory) if blob.name.endswith('.wav')])\n",
        "\n",
        "    for i, wav_blob in enumerate(wav_blobs, start = 1):\n",
        "        wav_local_path = download_blob_to_temp_file(wav_blob)\n",
        "        y, sr = librosa.load(wav_local_path, sr = None)\n",
        "\n",
        "        C = librosa.cqt(y, sr = sr)\n",
        "        magnitude = np.abs(C)\n",
        "        C_db = librosa.amplitude_to_db(magnitude, ref = np.max)\n",
        "\n",
        "        plt.figure(figsize=(10, 4))\n",
        "        librosa.display.specshow(C_db, sr = sr)\n",
        "        plt.axis('off')\n",
        "\n",
        "        with tempfile.NamedTemporaryFile(delete = False, suffix = '.png') as spectrogram_temp_file:\n",
        "            spectrogram_local_path = spectrogram_temp_file.name\n",
        "            plt.savefig(spectrogram_local_path, bbox_inches='tight', pad_inches = 0)\n",
        "            plt.close()\n",
        "\n",
        "        spectrogram_blob_name = wav_blob.replace(wav_directory, spectrograms_directory).replace('.wav', '.png')\n",
        "        upload_file_to_bucket(spectrogram_local_path, spectrogram_blob_name)\n",
        "\n",
        "        os.remove(wav_local_path)\n",
        "        os.remove(spectrogram_local_path)\n",
        "\n",
        "        print(f\"Spectrogram {i}/{len(wav_blobs)} saved!\")\n",
        "\n",
        "    return None"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ckUdJzvY_w4a"
      },
      "outputs": [],
      "source": [
        "# Notes one-hot encoding\n",
        "\n",
        "def one_hot_encode_midi(midi_directory = midi_directory):\n",
        "    \"\"\"\n",
        "    One-hot encode the midi file notes (ranging from 21 to 108 for piano).\n",
        "    \"\"\"\n",
        "\n",
        "    one_hot_encoded_midi = []\n",
        "\n",
        "    midi_blobs = sorted([blob.name for blob in bucket.list_blobs(prefix = midi_directory) if blob.name.endswith('.mid')])\n",
        "\n",
        "    for midi_blob in midi_blobs:\n",
        "\n",
        "        midi_local_path = download_blob_to_temp_file(midi_blob)\n",
        "\n",
        "        mid = MidiFile(midi_local_path)\n",
        "        one_hot_vector = [0] * 88\n",
        "\n",
        "        for msg in mid:\n",
        "            if msg.type == 'note_on' and msg.velocity > 0:\n",
        "                note_index = msg.note - 21\n",
        "                one_hot_vector[note_index] = 1\n",
        "\n",
        "        one_hot_encoded_midi.append(one_hot_vector)\n",
        "\n",
        "        os.remove(midi_local_path)\n",
        "\n",
        "    one_hot_encoded_midi = np.array(one_hot_encoded_midi)\n",
        "\n",
        "    return one_hot_encoded_midi"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "eO3vdLm4_ybL"
      },
      "outputs": [],
      "source": [
        "def load_spectrograms_as_array(spectrograms_directory = spectrograms_directory, img_size = (60, 150)):\n",
        "    \"\"\"\n",
        "    Load spectrograms from GCS and convert them to ndarray.\n",
        "    \"\"\"\n",
        "\n",
        "    spectrogram_blobs = sorted([blob.name for blob in bucket.list_blobs(prefix = spectrograms_directory) if blob.name.endswith('.spectrograms') or blob.name.endswith('.png')])\n",
        "\n",
        "    spectrograms = []\n",
        "\n",
        "    for spectrogram_blob in spectrogram_blobs:\n",
        "        spectrogram_local_path = download_blob_to_temp_file(spectrogram_blob)\n",
        "\n",
        "        img = Image.open(spectrogram_local_path)\n",
        "        img = img.convert('L')\n",
        "        img = img.resize(img_size)\n",
        "        img_array = np.array(img) / 255\n",
        "\n",
        "        spectrograms.append(img_array)\n",
        "\n",
        "        os.remove(spectrogram_local_path)\n",
        "\n",
        "    spectrograms_array = np.stack(spectrograms, axis = 0)\n",
        "\n",
        "    return spectrograms_array"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1-DSBkn5_2th"
      },
      "source": [
        "*2*. <u>Create dataset</u>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "uCALmO7h_9W6"
      },
      "outputs": [],
      "source": [
        "# Create X (spectograms)\n",
        "\n",
        "#convert_midi_to_wav()\n",
        "\n",
        "#from_audio_to_spectrograms()\n",
        "\n",
        "img_size = (60, 150)\n",
        "\n",
        "X = load_spectrograms_as_array()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mTFxt4Bh4QaZ",
        "outputId": "fcc173b8-feb8-4d00-b64e-98241fa86871"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "True"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ],
      "source": [
        "# Sanity check\n",
        "\n",
        "midi_blobs = [blob.name for blob in bucket.list_blobs(prefix = midi_directory) if blob.name.endswith('.mid')]\n",
        "wav_blobs = [blob.name for blob in bucket.list_blobs(prefix = wav_directory) if blob.name.endswith('.wav')]\n",
        "spectrogram_blobs = [blob.name for blob in bucket.list_blobs(prefix = spectrograms_directory) if blob.name.endswith('.spectrograms') or blob.name.endswith('.png')]\n",
        "\n",
        "len(midi_blobs) == len(wav_blobs) == len(spectrogram_blobs)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "6uij6bKG_781"
      },
      "outputs": [],
      "source": [
        "# Create y (encoded notes)\n",
        "\n",
        "y = one_hot_encode_midi()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 406
        },
        "id": "GpzalCKK3Fq9",
        "outputId": "68551704-6ce0-489d-cfbb-06898b0ff184"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "# Visualize first spectrogram\n",
        "\n",
        "plt.imshow(X[0], cmap='gray')\n",
        "plt.axis('off')\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "u6cwe2HUTmQ4",
        "outputId": "3a206bd1-b175-4445-c3e5-9cb395612f95"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "File dataset/X uploaded to GCS.\n",
            "File dataset/y uploaded to GCS.\n"
          ]
        }
      ],
      "source": [
        "# Save X and y on GCS\n",
        "\n",
        "save_pkl_to_gcp(X, 'dataset/X')\n",
        "save_pkl_to_gcp(y, 'dataset/y')"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4"
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}