Training_notebook_autoencoder_DB_norm - Copy.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Extracting spectrograms:   0%|          | 0/10000 [00:00<?, ?it/s]C:\\Users\\admin\\AppData\\Local\\Temp\\ipykernel_19268\\3870497425.py:10: DeprecationWarning: ANTIALIAS is deprecated and will be removed in Pillow 10 (2023-07-01). Use LANCZOS or Resampling.LANCZOS instead.\n",
            "  img = img.resize((128, 128), Image.ANTIALIAS)\n",
            "Extracting spectrograms: 100%|██████████| 10000/10000 [07:26<00:00, 22.38it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Clean Train spectrograms shape: (10000, 128, 128)\n",
            "Noisy Train spectrograms shape: (10000, 128, 128)\n"
          ]
        }
      ],
      "source": [
        "import os\n",
        "import librosa\n",
        "import numpy as np\n",
        "from tqdm import tqdm\n",
        "from PIL import Image\n",
        "\n",
        "# Function to resize spectrogram to 128x128\n",
        "def resize_spectrogram(spectrogram):\n",
        "    img = Image.fromarray(spectrogram)\n",
        "    img = img.resize((128, 128), Image.ANTIALIAS)\n",
        "    return np.array(img)\n",
        "\n",
        "# Function to extract STFT magnitude spectrograms\n",
        "def extract_spectrograms(clean_folder, noisy_folder, window_length=255, fft_length=255, hop_length=63):\n",
        "    clean_signal_stft_magnitude = []\n",
        "    noisy_signal_stft_magnitude = []\n",
        "\n",
        "    clean_files = os.listdir(clean_folder)\n",
        "    noisy_files = os.listdir(noisy_folder)\n",
        "\n",
        "    for clean_file, noisy_file in tqdm(zip(clean_files, noisy_files), desc=\"Extracting spectrograms\", total=len(clean_files)):\n",
        "        if clean_file.endswith(\".wav\") and noisy_file.endswith(\".wav\"):\n",
        "            clean_filepath = os.path.join(clean_folder, clean_file)\n",
        "            noisy_filepath = os.path.join(noisy_folder, noisy_file)\n",
        "\n",
        "            # Load clean audio file\n",
        "            clean_signal, sr = librosa.load(clean_filepath)\n",
        "            # Load noisy audio file\n",
        "            noisy_signal, _ = librosa.load(noisy_filepath, sr=sr)\n",
        "\n",
        "            # Compute STFT for clean signal\n",
        "            clean_stft = librosa.stft(clean_signal, n_fft=fft_length, hop_length=hop_length, win_length=window_length)\n",
        "            clean_stft_mag = np.abs(clean_stft)\n",
        "            # Convert to dB and scale\n",
        "            clean_stft_mag_db = (librosa.amplitude_to_db(clean_stft_mag, ref=np.max) + 80) / 80\n",
        "            clean_stft_mag_db_resized = resize_spectrogram(clean_stft_mag_db)\n",
        "            clean_signal_stft_magnitude.append(clean_stft_mag_db_resized)\n",
        "\n",
        "            # Compute STFT for noisy signal\n",
        "            noisy_stft = librosa.stft(noisy_signal, n_fft=fft_length, hop_length=hop_length, win_length=window_length)\n",
        "            noisy_stft_mag = np.abs(noisy_stft)\n",
        "            # Convert to dB and scale\n",
        "            noisy_stft_mag_db = (librosa.amplitude_to_db(noisy_stft_mag, ref=np.max) + 80) / 80\n",
        "            noisy_stft_mag_db_resized = resize_spectrogram(noisy_stft_mag_db)\n",
        "            noisy_signal_stft_magnitude.append(noisy_stft_mag_db_resized)\n",
        "\n",
        "    # Convert lists to numpy arrays\n",
        "    clean_signal_stft_magnitude = np.array(clean_signal_stft_magnitude)\n",
        "    noisy_signal_stft_magnitude = np.array(noisy_signal_stft_magnitude)\n",
        "\n",
        "    return clean_signal_stft_magnitude, noisy_signal_stft_magnitude\n",
        "\n",
        "\n",
        "# Example usage:\n",
        "clean_folder = r\"C:\\Users\\admin\\Desktop\\ARTH SHAH\\Hearing Aids\\Wind_noise\\Dataset\\Train\\clean\\clean\"\n",
        "noisy_folder = r\"C:\\Users\\admin\\Desktop\\ARTH SHAH\\Hearing Aids\\Wind_noise\\Dataset\\Train\\noisy\\noisy\"\n",
        "clean_spectrograms, noisy_spectrograms = extract_spectrograms(clean_folder, noisy_folder)\n",
        "print(\"Clean Train spectrograms shape:\", clean_spectrograms.shape)\n",
        "print(\"Noisy Train spectrograms shape:\", noisy_spectrograms.shape)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {},
      "outputs": [
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "Extracting spectrograms:   0%|          | 0/2000 [00:00<?, ?it/s]"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "C:\\Users\\admin\\AppData\\Local\\Temp\\ipykernel_19268\\1601907199.py:10: DeprecationWarning: ANTIALIAS is deprecated and will be removed in Pillow 10 (2023-07-01). Use LANCZOS or Resampling.LANCZOS instead.\n",
            "  img = img.resize((128, 128), Image.ANTIALIAS)\n",
            "Extracting spectrograms: 100%|██████████| 2000/2000 [01:07<00:00, 29.84it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Clean Test spectrograms shape: (2000, 128, 128)\n",
            "Noisy Test spectrograms shape: (2000, 128, 128)\n"
          ]
        }
      ],
      "source": [
        "import os\n",
        "import librosa\n",
        "import numpy as np\n",
        "from tqdm import tqdm\n",
        "from PIL import Image\n",
        "\n",
        "# Function to resize spectrogram to 128x128\n",
        "def resize_spectrogram(spectrogram):\n",
        "    img = Image.fromarray(spectrogram)\n",
        "    img = img.resize((128, 128), Image.ANTIALIAS)\n",
        "    return np.array(img)\n",
        "\n",
        "# Function to extract STFT magnitude spectrograms\n",
        "def extract_spectrograms(clean_folder, noisy_folder, window_length=255, fft_length=255, hop_length=63):\n",
        "    clean_signal_stft_magnitude = []\n",
        "    noisy_signal_stft_magnitude = []\n",
        "\n",
        "    clean_files = os.listdir(clean_folder)\n",
        "    noisy_files = os.listdir(noisy_folder)\n",
        "\n",
        "    for clean_file, noisy_file in tqdm(zip(clean_files, noisy_files), desc=\"Extracting spectrograms\", total=len(clean_files)):\n",
        "        if clean_file.endswith(\".wav\") and noisy_file.endswith(\".wav\"):\n",
        "            clean_filepath = os.path.join(clean_folder, clean_file)\n",
        "            noisy_filepath = os.path.join(noisy_folder, noisy_file)\n",
        "\n",
        "            # Load clean audio file\n",
        "            clean_signal, sr = librosa.load(clean_filepath)\n",
        "            # Load noisy audio file\n",
        "            noisy_signal, _ = librosa.load(noisy_filepath, sr=sr)\n",
        "\n",
        "            # Compute STFT for clean signal\n",
        "            clean_stft = librosa.stft(clean_signal, n_fft=fft_length, hop_length=hop_length, win_length=window_length)\n",
        "            clean_stft_mag = np.abs(clean_stft)\n",
        "            # Convert to dB and scale\n",
        "            clean_stft_mag_db = (librosa.amplitude_to_db(clean_stft_mag, ref=np.max) + 80) / 80\n",
        "            clean_stft_mag_db_resized = resize_spectrogram(clean_stft_mag_db)\n",
        "            clean_signal_stft_magnitude.append(clean_stft_mag_db_resized)\n",
        "\n",
        "            # Compute STFT for noisy signal\n",
        "            noisy_stft = librosa.stft(noisy_signal, n_fft=fft_length, hop_length=hop_length, win_length=window_length)\n",
        "            noisy_stft_mag = np.abs(noisy_stft)\n",
        "            # Convert to dB and scale\n",
        "            noisy_stft_mag_db = (librosa.amplitude_to_db(noisy_stft_mag, ref=np.max) + 80) / 80\n",
        "            noisy_stft_mag_db_resized = resize_spectrogram(noisy_stft_mag_db)\n",
        "            noisy_signal_stft_magnitude.append(noisy_stft_mag_db_resized)\n",
        "\n",
        "    # Convert lists to numpy arrays\n",
        "    clean_test_signal_stft_magnitude = np.array(clean_signal_stft_magnitude)\n",
        "    noisy_test_signal_stft_magnitude = np.array(noisy_signal_stft_magnitude)\n",
        "\n",
        "    return clean_test_signal_stft_magnitude, noisy_test_signal_stft_magnitude\n",
        "\n",
        "\n",
        "# Example usage:\n",
        "clean_folder = r\"C:\\Users\\admin\\Desktop\\ARTH SHAH\\Hearing Aids\\Wind_noise\\Dataset\\Test\\clean\\clean\"\n",
        "noisy_folder = r\"C:\\Users\\admin\\Desktop\\ARTH SHAH\\Hearing Aids\\Wind_noise\\Dataset\\Test\\noisy\\noisy\"\n",
        "clean_test_spectrograms, noisy_test_spectrograms = extract_spectrograms(clean_folder, noisy_folder)\n",
        "print(\"Clean Test spectrograms shape:\", clean_test_spectrograms.shape)\n",
        "print(\"Noisy Test spectrograms shape:\", noisy_test_spectrograms.shape)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "ShEotzYZGGu_"
      },
      "outputs": [],
      "source": [
        "from keras.datasets import mnist\n",
        "import numpy as np\n",
        "import keras\n",
        "from keras import layers\n",
        "from keras.callbacks import TensorBoard\n",
        "import librosa\n",
        "import librosa.display\n",
        "import os\n",
        "import glob\n",
        "import scipy\n",
        "import matplotlib.pyplot as plt\n",
        "import soundfile as sf\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "import tensorflow as tf\n",
        "from tqdm import tqdm\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 36,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "PD-EkAC_VhpO",
        "outputId": "d56877a4-923a-4bec-f2dc-a01017037b95"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(10000, 128, 128)\n",
            "(10000, 128, 128)\n",
            "(2000, 128, 128)\n",
            "(2000, 128, 128)\n"
          ]
        }
      ],
      "source": [
        "# train_count = 10500 #Split is 900+108 for background\n",
        "train_count = 20000\n",
        "test_count = 4000\n",
        "\n",
        "noisy_train=noisy_spectrograms[:train_count]\n",
        "noisy_test=noisy_test_spectrograms[:test_count]\n",
        "clean_train=clean_spectrograms[:train_count]\n",
        "clean_test=clean_test_spectrograms[:test_count]\n",
        "print(noisy_train.shape)\n",
        "print(clean_train.shape)\n",
        "print(noisy_test.shape)\n",
        "print(clean_test.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 37,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nK9DsA3iYeWm",
        "outputId": "80d0f69d-a875-44fc-8213-be9c58802c19"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(10000, 128, 128, 1)\n",
            "(10000, 128, 128, 1)\n",
            "(2000, 128, 128, 1)\n",
            "(2000, 128, 128, 1)\n"
          ]
        }
      ],
      "source": [
        "noisy_train = np.reshape(noisy_train, (noisy_train.shape[0], noisy_train.shape[1], noisy_train.shape[2], 1))\n",
        "noisy_test = np.reshape(noisy_test, (noisy_test.shape[0], noisy_test.shape[1], noisy_test.shape[2], 1))\n",
        "clean_train = np.reshape(clean_train, (clean_train.shape[0], clean_train.shape[1], clean_train.shape[2], 1))\n",
        "clean_test = np.reshape(clean_test, (clean_test.shape[0], clean_test.shape[1], clean_test.shape[2], 1))\n",
        "print(noisy_train.shape)\n",
        "print(clean_train.shape)\n",
        "print(noisy_test.shape)\n",
        "print(clean_test.shape)\n",
        "#Saving RAM\n",
        "noisy_signal_stft_magnitude = 0\n",
        "clean_signal_stft_magnitude = 0\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {},
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": 21,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0TASx3vaoWkk",
        "outputId": "6fe6c374-57bc-4623-aa4b-99dba24bd23f"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"model_1\"\n",
            "__________________________________________________________________________________________________\n",
            " Layer (type)                   Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            " input_2 (InputLayer)           [(None, 128, 128, 1  0           []                               \n",
            "                                )]                                                                \n",
            "                                                                                                  \n",
            " conv2d_24 (Conv2D)             (None, 128, 128, 16  160         ['input_2[0][0]']                \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " leaky_re_lu_23 (LeakyReLU)     (None, 128, 128, 16  0           ['conv2d_24[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_25 (Conv2D)             (None, 128, 128, 16  2320        ['leaky_re_lu_23[0][0]']         \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " leaky_re_lu_24 (LeakyReLU)     (None, 128, 128, 16  0           ['conv2d_25[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " max_pooling2d_4 (MaxPooling2D)  (None, 64, 64, 16)  0           ['leaky_re_lu_24[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_26 (Conv2D)             (None, 64, 64, 32)   4640        ['max_pooling2d_4[0][0]']        \n",
            "                                                                                                  \n",
            " leaky_re_lu_25 (LeakyReLU)     (None, 64, 64, 32)   0           ['conv2d_26[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_27 (Conv2D)             (None, 64, 64, 32)   9248        ['leaky_re_lu_25[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_26 (LeakyReLU)     (None, 64, 64, 32)   0           ['conv2d_27[0][0]']              \n",
            "                                                                                                  \n",
            " max_pooling2d_5 (MaxPooling2D)  (None, 32, 32, 32)  0           ['leaky_re_lu_26[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_28 (Conv2D)             (None, 32, 32, 64)   18496       ['max_pooling2d_5[0][0]']        \n",
            "                                                                                                  \n",
            " leaky_re_lu_27 (LeakyReLU)     (None, 32, 32, 64)   0           ['conv2d_28[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_29 (Conv2D)             (None, 32, 32, 64)   36928       ['leaky_re_lu_27[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_28 (LeakyReLU)     (None, 32, 32, 64)   0           ['conv2d_29[0][0]']              \n",
            "                                                                                                  \n",
            " max_pooling2d_6 (MaxPooling2D)  (None, 16, 16, 64)  0           ['leaky_re_lu_28[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_30 (Conv2D)             (None, 16, 16, 128)  73856       ['max_pooling2d_6[0][0]']        \n",
            "                                                                                                  \n",
            " leaky_re_lu_29 (LeakyReLU)     (None, 16, 16, 128)  0           ['conv2d_30[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_31 (Conv2D)             (None, 16, 16, 128)  147584      ['leaky_re_lu_29[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_30 (LeakyReLU)     (None, 16, 16, 128)  0           ['conv2d_31[0][0]']              \n",
            "                                                                                                  \n",
            " dropout_2 (Dropout)            (None, 16, 16, 128)  0           ['leaky_re_lu_30[0][0]']         \n",
            "                                                                                                  \n",
            " max_pooling2d_7 (MaxPooling2D)  (None, 8, 8, 128)   0           ['dropout_2[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_32 (Conv2D)             (None, 8, 8, 256)    295168      ['max_pooling2d_7[0][0]']        \n",
            "                                                                                                  \n",
            " leaky_re_lu_31 (LeakyReLU)     (None, 8, 8, 256)    0           ['conv2d_32[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_33 (Conv2D)             (None, 8, 8, 256)    590080      ['leaky_re_lu_31[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_32 (LeakyReLU)     (None, 8, 8, 256)    0           ['conv2d_33[0][0]']              \n",
            "                                                                                                  \n",
            " dropout_3 (Dropout)            (None, 8, 8, 256)    0           ['leaky_re_lu_32[0][0]']         \n",
            "                                                                                                  \n",
            " up_sampling2d_4 (UpSampling2D)  (None, 16, 16, 256)  0          ['dropout_3[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_34 (Conv2D)             (None, 16, 16, 128)  131200      ['up_sampling2d_4[0][0]']        \n",
            "                                                                                                  \n",
            " leaky_re_lu_33 (LeakyReLU)     (None, 16, 16, 128)  0           ['conv2d_34[0][0]']              \n",
            "                                                                                                  \n",
            " concatenate_4 (Concatenate)    (None, 16, 16, 256)  0           ['dropout_2[0][0]',              \n",
            "                                                                  'leaky_re_lu_33[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_35 (Conv2D)             (None, 16, 16, 128)  295040      ['concatenate_4[0][0]']          \n",
            "                                                                                                  \n",
            " leaky_re_lu_34 (LeakyReLU)     (None, 16, 16, 128)  0           ['conv2d_35[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_36 (Conv2D)             (None, 16, 16, 128)  147584      ['leaky_re_lu_34[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_35 (LeakyReLU)     (None, 16, 16, 128)  0           ['conv2d_36[0][0]']              \n",
            "                                                                                                  \n",
            " up_sampling2d_5 (UpSampling2D)  (None, 32, 32, 128)  0          ['leaky_re_lu_35[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_37 (Conv2D)             (None, 32, 32, 64)   32832       ['up_sampling2d_5[0][0]']        \n",
            "                                                                                                  \n",
            " leaky_re_lu_36 (LeakyReLU)     (None, 32, 32, 64)   0           ['conv2d_37[0][0]']              \n",
            "                                                                                                  \n",
            " concatenate_5 (Concatenate)    (None, 32, 32, 128)  0           ['leaky_re_lu_28[0][0]',         \n",
            "                                                                  'leaky_re_lu_36[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_38 (Conv2D)             (None, 32, 32, 64)   73792       ['concatenate_5[0][0]']          \n",
            "                                                                                                  \n",
            " leaky_re_lu_37 (LeakyReLU)     (None, 32, 32, 64)   0           ['conv2d_38[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_39 (Conv2D)             (None, 32, 32, 64)   36928       ['leaky_re_lu_37[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_38 (LeakyReLU)     (None, 32, 32, 64)   0           ['conv2d_39[0][0]']              \n",
            "                                                                                                  \n",
            " up_sampling2d_6 (UpSampling2D)  (None, 64, 64, 64)  0           ['leaky_re_lu_38[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_40 (Conv2D)             (None, 64, 64, 32)   8224        ['up_sampling2d_6[0][0]']        \n",
            "                                                                                                  \n",
            " leaky_re_lu_39 (LeakyReLU)     (None, 64, 64, 32)   0           ['conv2d_40[0][0]']              \n",
            "                                                                                                  \n",
            " concatenate_6 (Concatenate)    (None, 64, 64, 64)   0           ['leaky_re_lu_26[0][0]',         \n",
            "                                                                  'leaky_re_lu_39[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_41 (Conv2D)             (None, 64, 64, 32)   18464       ['concatenate_6[0][0]']          \n",
            "                                                                                                  \n",
            " leaky_re_lu_40 (LeakyReLU)     (None, 64, 64, 32)   0           ['conv2d_41[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_42 (Conv2D)             (None, 64, 64, 32)   9248        ['leaky_re_lu_40[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_41 (LeakyReLU)     (None, 64, 64, 32)   0           ['conv2d_42[0][0]']              \n",
            "                                                                                                  \n",
            " up_sampling2d_7 (UpSampling2D)  (None, 128, 128, 32  0          ['leaky_re_lu_41[0][0]']         \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_43 (Conv2D)             (None, 128, 128, 16  2064        ['up_sampling2d_7[0][0]']        \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " leaky_re_lu_42 (LeakyReLU)     (None, 128, 128, 16  0           ['conv2d_43[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " concatenate_7 (Concatenate)    (None, 128, 128, 32  0           ['leaky_re_lu_24[0][0]',         \n",
            "                                )                                 'leaky_re_lu_42[0][0]']         \n",
            "                                                                                                  \n",
            " conv2d_44 (Conv2D)             (None, 128, 128, 16  4624        ['concatenate_7[0][0]']          \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " leaky_re_lu_43 (LeakyReLU)     (None, 128, 128, 16  0           ['conv2d_44[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_45 (Conv2D)             (None, 128, 128, 16  2320        ['leaky_re_lu_43[0][0]']         \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " leaky_re_lu_44 (LeakyReLU)     (None, 128, 128, 16  0           ['conv2d_45[0][0]']              \n",
            "                                )                                                                 \n",
            "                                                                                                  \n",
            " conv2d_46 (Conv2D)             (None, 128, 128, 2)  290         ['leaky_re_lu_44[0][0]']         \n",
            "                                                                                                  \n",
            " leaky_re_lu_45 (LeakyReLU)     (None, 128, 128, 2)  0           ['conv2d_46[0][0]']              \n",
            "                                                                                                  \n",
            " conv2d_47 (Conv2D)             (None, 128, 128, 1)  3           ['leaky_re_lu_45[0][0]']         \n",
            "                                                                                                  \n",
            "==================================================================================================\n",
            "Total params: 1,941,093\n",
            "Trainable params: 1,941,093\n",
            "Non-trainable params: 0\n",
            "__________________________________________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "#IF issues with GPU remove tensorflow.keras.* and only keep keras.*\n",
        "import tensorflow.keras as keras\n",
        "from keras.models import Model\n",
        "from keras.layers import Input, Conv2D, LeakyReLU, MaxPooling2D, Dropout, concatenate, UpSampling2D\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras import backend\n",
        "import tensorflow as tf\n",
        "\n",
        "size_filter_in = 16\n",
        "#normal initialization of weights\n",
        "kernel_init = 'he_normal'\n",
        "#To apply leaky relu after the conv layer \n",
        "activation_layer = None\n",
        "inputs = keras.Input(shape=(128, 128, 1))\n",
        "conv1 = Conv2D(size_filter_in, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(inputs)\n",
        "conv1 = LeakyReLU()(conv1)\n",
        "conv1 = Conv2D(size_filter_in, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv1)\n",
        "conv1 = LeakyReLU()(conv1)\n",
        "pool1 = MaxPooling2D(pool_size=(2, 2))(conv1)\n",
        "conv2 = Conv2D(size_filter_in*2, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(pool1)\n",
        "conv2 = LeakyReLU()(conv2)\n",
        "conv2 = Conv2D(size_filter_in*2, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv2)\n",
        "conv2 = LeakyReLU()(conv2)\n",
        "pool2 = MaxPooling2D(pool_size=(2, 2))(conv2)\n",
        "conv3 = Conv2D(size_filter_in*4, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(pool2)\n",
        "conv3 = LeakyReLU()(conv3)\n",
        "conv3 = Conv2D(size_filter_in*4, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv3)\n",
        "conv3 = LeakyReLU()(conv3)\n",
        "pool3 = MaxPooling2D(pool_size=(2, 2))(conv3)\n",
        "conv4 = Conv2D(size_filter_in*8, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(pool3)\n",
        "conv4 = LeakyReLU()(conv4)\n",
        "conv4 = Conv2D(size_filter_in*8, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv4)\n",
        "conv4 = LeakyReLU()(conv4)\n",
        "drop4 = Dropout(0.5)(conv4)\n",
        "pool4 = MaxPooling2D(pool_size=(2, 2))(drop4)\n",
        "\n",
        "conv5 = Conv2D(size_filter_in*16, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(pool4)\n",
        "conv5 = LeakyReLU()(conv5)\n",
        "conv5 = Conv2D(size_filter_in*16, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv5)\n",
        "conv5 = LeakyReLU()(conv5)\n",
        "drop5 = Dropout(0.5)(conv5)\n",
        "\n",
        "up6 = Conv2D(size_filter_in*8, 2, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(UpSampling2D(size = (2,2))(drop5))\n",
        "up6 = LeakyReLU()(up6)\n",
        "merge6 = concatenate([drop4,up6], axis = 3)\n",
        "conv6 = Conv2D(size_filter_in*8, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(merge6)\n",
        "conv6 = LeakyReLU()(conv6)\n",
        "conv6 = Conv2D(size_filter_in*8, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv6)\n",
        "conv6 = LeakyReLU()(conv6)\n",
        "up7 = Conv2D(size_filter_in*4, 2, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(UpSampling2D(size = (2,2))(conv6))\n",
        "up7 = LeakyReLU()(up7)\n",
        "merge7 = concatenate([conv3,up7], axis = 3)\n",
        "conv7 = Conv2D(size_filter_in*4, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(merge7)\n",
        "conv7 = LeakyReLU()(conv7)\n",
        "conv7 = Conv2D(size_filter_in*4, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv7)\n",
        "conv7 = LeakyReLU()(conv7)\n",
        "up8 = Conv2D(size_filter_in*2, 2, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(UpSampling2D(size = (2,2))(conv7))\n",
        "up8 = LeakyReLU()(up8)\n",
        "merge8 = concatenate([conv2,up8], axis = 3)\n",
        "conv8 = Conv2D(size_filter_in*2, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(merge8)\n",
        "conv8 = LeakyReLU()(conv8)\n",
        "conv8 = Conv2D(size_filter_in*2, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv8)\n",
        "conv8 = LeakyReLU()(conv8)\n",
        "\n",
        "up9 = Conv2D(size_filter_in, 2, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(UpSampling2D(size = (2,2))(conv8))\n",
        "up9 = LeakyReLU()(up9)\n",
        "merge9 = concatenate([conv1,up9], axis = 3)\n",
        "conv9 = Conv2D(size_filter_in, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(merge9)\n",
        "conv9 = LeakyReLU()(conv9)\n",
        "conv9 = Conv2D(size_filter_in, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv9)\n",
        "conv9 = LeakyReLU()(conv9)\n",
        "conv9 = Conv2D(2, 3, activation = activation_layer, padding = 'same', kernel_initializer = kernel_init)(conv9)\n",
        "conv9 = LeakyReLU()(conv9)\n",
        "conv10 = Conv2D(1, 1, activation = 'tanh')(conv9)\n",
        "\n",
        "model = Model(inputs,conv10)\n",
        "\n",
        "model.compile(optimizer = 'adam', loss = tf.keras.losses.Huber(), metrics = ['mae'])\n",
        "#model.compile(optimizer = 'adam', loss = tf.keras.losses.Huber(), metrics = ['mae'])\n",
        "\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LVVsKhtIo8fM"
      },
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": 25,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iCNWhnEUoXZF",
        "outputId": "75978e1c-6ed9-4da0-a7de-18cec730d1fa"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "157/157 [==============================] - 508s 3s/step - loss: 0.0068 - mae: 0.0738\n",
            "Epoch 2/30\n",
            "157/157 [==============================] - 475s 3s/step - loss: 0.0062 - mae: 0.0688\n",
            "Epoch 3/30\n",
            "157/157 [==============================] - 474s 3s/step - loss: 0.0059 - mae: 0.0660\n",
            "Epoch 4/30\n",
            "157/157 [==============================] - 474s 3s/step - loss: 0.0057 - mae: 0.0643\n",
            "Epoch 5/30\n",
            "157/157 [==============================] - 472s 3s/step - loss: 0.0055 - mae: 0.0627\n",
            "Epoch 6/30\n",
            "157/157 [==============================] - 471s 3s/step - loss: 0.0053 - mae: 0.0612\n",
            "Epoch 7/30\n",
            "157/157 [==============================] - 468s 3s/step - loss: 0.0052 - mae: 0.0606\n",
            "Epoch 8/30\n",
            "157/157 [==============================] - 476s 3s/step - loss: 0.0051 - mae: 0.0599\n",
            "Epoch 9/30\n",
            "157/157 [==============================] - 468s 3s/step - loss: 0.0051 - mae: 0.0596\n",
            "Epoch 10/30\n",
            "157/157 [==============================] - 466s 3s/step - loss: 0.0050 - mae: 0.0585\n",
            "Epoch 11/30\n",
            "157/157 [==============================] - 465s 3s/step - loss: 0.0050 - mae: 0.0584\n",
            "Epoch 12/30\n",
            "157/157 [==============================] - 464s 3s/step - loss: 0.0048 - mae: 0.0573\n",
            "Epoch 13/30\n",
            "157/157 [==============================] - 467s 3s/step - loss: 0.0048 - mae: 0.0568\n",
            "Epoch 14/30\n",
            "157/157 [==============================] - 465s 3s/step - loss: 0.0048 - mae: 0.0566\n",
            "Epoch 15/30\n",
            "157/157 [==============================] - 466s 3s/step - loss: 0.0047 - mae: 0.0561\n",
            "Epoch 16/30\n",
            "157/157 [==============================] - 465s 3s/step - loss: 0.0046 - mae: 0.0555\n",
            "Epoch 17/30\n",
            "157/157 [==============================] - 466s 3s/step - loss: 0.0046 - mae: 0.0548\n",
            "Epoch 18/30\n",
            "157/157 [==============================] - 465s 3s/step - loss: 0.0046 - mae: 0.0550\n",
            "Epoch 19/30\n",
            "157/157 [==============================] - 466s 3s/step - loss: 0.0046 - mae: 0.0549\n",
            "Epoch 20/30\n",
            "157/157 [==============================] - 466s 3s/step - loss: 0.0045 - mae: 0.0543\n",
            "Epoch 21/30\n",
            "157/157 [==============================] - 469s 3s/step - loss: 0.0044 - mae: 0.0536\n",
            "Epoch 22/30\n",
            "157/157 [==============================] - 465s 3s/step - loss: 0.0044 - mae: 0.0535\n",
            "Epoch 23/30\n",
            "157/157 [==============================] - 466s 3s/step - loss: 0.0044 - mae: 0.0532\n",
            "Epoch 24/30\n",
            "157/157 [==============================] - 468s 3s/step - loss: 0.0044 - mae: 0.0531\n",
            "Epoch 25/30\n",
            "157/157 [==============================] - 467s 3s/step - loss: 0.0043 - mae: 0.0522\n",
            "Epoch 26/30\n",
            "157/157 [==============================] - 465s 3s/step - loss: 0.0043 - mae: 0.0523\n",
            "Epoch 27/30\n",
            "157/157 [==============================] - 467s 3s/step - loss: 0.0043 - mae: 0.0529\n",
            "Epoch 28/30\n",
            "157/157 [==============================] - 466s 3s/step - loss: 0.0042 - mae: 0.0516\n",
            "Epoch 29/30\n",
            "157/157 [==============================] - 468s 3s/step - loss: 0.0043 - mae: 0.0521\n",
            "Epoch 30/30\n",
            "157/157 [==============================] - 469s 3s/step - loss: 0.0042 - mae: 0.0512\n"
          ]
        }
      ],
      "source": [
        "# model = keras.models.load_model('/content/gdrive/My Drive/Project/New_FFT_Vox_Urban+background_128x128_30_epochs.h5')\n",
        "history = model.fit(noisy_train, clean_train,\n",
        "                epochs=30,\n",
        "                batch_size=64,\n",
        "                shuffle=True,\n",
        "                validation_data=(noisy_test, clean_test),\n",
        "                callbacks=[TensorBoard(log_dir='tensorflow_logs', histogram_freq=0, write_graph=False)])\n",
        "model.save(r\"C:\\Users\\admin\\Desktop\\ARTH SHAH\\Hearing Aids\\Wind_noise\\spectrogram\\New_FFT_Vox_Urban+background_128x128_60_epochs.h5\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 32,
      "metadata": {},
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/30\n",
            "157/157 [==============================] - 491s 3s/step - loss: 0.0042 - mae: 0.0519\n",
            "Epoch 2/30\n",
            "157/157 [==============================] - 484s 3s/step - loss: 0.0042 - mae: 0.0513\n",
            "Epoch 3/30\n",
            "157/157 [==============================] - 486s 3s/step - loss: 0.0041 - mae: 0.0510\n",
            "Epoch 4/30\n",
            "157/157 [==============================] - 509s 3s/step - loss: 0.0041 - mae: 0.0508\n",
            "Epoch 5/30\n",
            "157/157 [==============================] - 504s 3s/step - loss: 0.0042 - mae: 0.0514\n",
            "Epoch 6/30\n",
            "157/157 [==============================] - 481s 3s/step - loss: 0.0041 - mae: 0.0506\n",
            "Epoch 7/30\n",
            "157/157 [==============================] - 484s 3s/step - loss: 0.0040 - mae: 0.0502\n",
            "Epoch 8/30\n",
            "157/157 [==============================] - 481s 3s/step - loss: 0.0041 - mae: 0.0507\n",
            "Epoch 9/30\n",
            "157/157 [==============================] - 481s 3s/step - loss: 0.0040 - mae: 0.0496\n",
            "Epoch 10/30\n",
            "157/157 [==============================] - 495s 3s/step - loss: 0.0040 - mae: 0.0496\n",
            "Epoch 11/30\n",
            "157/157 [==============================] - 518s 3s/step - loss: 0.0040 - mae: 0.0499\n",
            "Epoch 12/30\n",
            "157/157 [==============================] - 551s 4s/step - loss: 0.0040 - mae: 0.0494\n",
            "Epoch 13/30\n",
            "157/157 [==============================] - 531s 3s/step - loss: 0.0039 - mae: 0.0492\n",
            "Epoch 14/30\n",
            "157/157 [==============================] - 494s 3s/step - loss: 0.0040 - mae: 0.0492\n",
            "Epoch 15/30\n",
            "157/157 [==============================] - 484s 3s/step - loss: 0.0040 - mae: 0.0496\n",
            "Epoch 16/30\n",
            "157/157 [==============================] - 476s 3s/step - loss: 0.0039 - mae: 0.0493\n",
            "Epoch 17/30\n",
            "157/157 [==============================] - 472s 3s/step - loss: 0.0040 - mae: 0.0496\n",
            "Epoch 18/30\n",
            "157/157 [==============================] - 472s 3s/step - loss: 0.0039 - mae: 0.0492\n",
            "Epoch 19/30\n",
            "157/157 [==============================] - 470s 3s/step - loss: 0.0039 - mae: 0.0487\n",
            "Epoch 20/30\n",
            "157/157 [==============================] - 471s 3s/step - loss: 0.0038 - mae: 0.0482\n",
            "Epoch 21/30\n",
            "157/157 [==============================] - 470s 3s/step - loss: 0.0038 - mae: 0.0482\n",
            "Epoch 22/30\n",
            "157/157 [==============================] - 474s 3s/step - loss: 0.0039 - mae: 0.0487\n",
            "Epoch 23/30\n",
            "157/157 [==============================] - 474s 3s/step - loss: 0.0038 - mae: 0.0479\n",
            "Epoch 24/30\n",
            "157/157 [==============================] - 471s 3s/step - loss: 0.0037 - mae: 0.0475\n",
            "Epoch 25/30\n",
            "157/157 [==============================] - 492s 3s/step - loss: 0.0037 - mae: 0.0472\n",
            "Epoch 26/30\n",
            "157/157 [==============================] - 477s 3s/step - loss: 0.0037 - mae: 0.0473\n",
            "Epoch 27/30\n",
            "157/157 [==============================] - 476s 3s/step - loss: 0.0037 - mae: 0.0475\n",
            "Epoch 28/30\n",
            "157/157 [==============================] - 476s 3s/step - loss: 0.0038 - mae: 0.0480\n",
            "Epoch 29/30\n",
            "157/157 [==============================] - 477s 3s/step - loss: 0.0037 - mae: 0.0476\n",
            "Epoch 30/30\n",
            "157/157 [==============================] - 476s 3s/step - loss: 0.0036 - mae: 0.0467\n"
          ]
        },
        {
          "ename": "KeyError",
          "evalue": "'val_loss'",
          "output_type": "error",
          "traceback": [
            "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
            "Cell \u001b[1;32mIn[32], line 19\u001b[0m\n\u001b[0;32m     17\u001b[0m \u001b[38;5;66;03m# Extract training and validation losses from the history object\u001b[39;00m\n\u001b[0;32m     18\u001b[0m training_loss \u001b[38;5;241m=\u001b[39m history\u001b[38;5;241m.\u001b[39mhistory[\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mloss\u001b[39m\u001b[38;5;124m'\u001b[39m]\n\u001b[1;32m---> 19\u001b[0m validation_loss \u001b[38;5;241m=\u001b[39m \u001b[43mhistory\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mhistory\u001b[49m\u001b[43m[\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mval_loss\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m]\u001b[49m\n\u001b[0;32m     21\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mTraining Losses:\u001b[39m\u001b[38;5;124m\"\u001b[39m, training_loss)\n\u001b[0;32m     22\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mValidation Losses:\u001b[39m\u001b[38;5;124m\"\u001b[39m, validation_loss)\n",
            "\u001b[1;31mKeyError\u001b[0m: 'val_loss'"
          ]
        }
      ],
      "source": [
        "import keras\n",
        "\n",
        "# Load the model\n",
        "model = keras.models.load_model(r\"C:\\Users\\admin\\Desktop\\ARTH SHAH\\Hearing Aids\\Wind_noise\\spectrogram\\New_FFT_Vox_Urban+background_128x128_60_epochs.h5\")\n",
        "\n",
        "# Train the model\n",
        "history = model.fit(noisy_train, clean_train,\n",
        "                    epochs=30,\n",
        "                    batch_size=64,\n",
        "                    shuffle=True,\n",
        "                    validation_data=(noisy_test, clean_test),\n",
        "                    callbacks=[keras.callbacks.TensorBoard(log_dir='tensorflow_logs', histogram_freq=0, write_graph=False)])\n",
        "\n",
        "# Save the model\n",
        "model.save(r\"C:\\Users\\admin\\Desktop\\ARTH SHAH\\Hearing Aids\\Wind_noise\\spectrogram\\aditional_30_epochs.h5\")\n",
        "\n",
        "# Extract training and validation losses from the history object\n",
        "training_loss = history.history['loss']\n",
        "validation_loss = history.history['val_loss']\n",
        "\n",
        "print(\"Training Losses:\", training_loss)\n",
        "print(\"Validation Losses:\", validation_loss)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 33,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "skjt9dAEnWQI",
        "outputId": "b10fa072-16ff-41d1-836a-02f1dfeee689"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "[0.004233227111399174,\n",
              " 0.004161493852734566,\n",
              " 0.004143517930060625,\n",
              " 0.004123617894947529,\n",
              " 0.004179518669843674,\n",
              " 0.004092097282409668,\n",
              " 0.0040495432913303375,\n",
              " 0.004102342762053013,\n",
              " 0.0039879390969872475,\n",
              " 0.003983957227319479,\n",
              " 0.00402236869558692,\n",
              " 0.0039717513136565685,\n",
              " 0.0039374735206365585,\n",
              " 0.003950375597923994,\n",
              " 0.003972605336457491,\n",
              " 0.00394063163548708,\n",
              " 0.003950472455471754,\n",
              " 0.003930138424038887,\n",
              " 0.0038720928132534027,\n",
              " 0.0038240565918385983,\n",
              " 0.00379776768386364,\n",
              " 0.0038690667133778334,\n",
              " 0.0037949844263494015,\n",
              " 0.0037307427264750004,\n",
              " 0.003710461314767599,\n",
              " 0.003705437760800123,\n",
              " 0.003728243289515376,\n",
              " 0.0037742648273706436,\n",
              " 0.003728339448571205,\n",
              " 0.0036359746009111404]"
            ]
          },
          "execution_count": 33,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "history.history['loss']"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 513
        },
        "id": "0BjPt1bhnyNt",
        "outputId": "fb6982ae-3212-4c16-efcd-4e89885106f8"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1008x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "output_type": "display_data"
        }
      ],
      "source": [
        "loss=training_loss \n",
        "val_loss=validation_loss\n",
        "plt.figure(figsize=(14, 8))\n",
        "# plt.plot(history.history['loss'])\n",
        "# plt.plot(history.history['val_loss'])\n",
        "plt.plot(loss)\n",
        "plt.plot(val_loss)\n",
        "plt.title('Loss graph of UNet Autoencoder model')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['training loss', 'validation loss'], loc='upper right')\n",
        "\n",
        "plt.show()"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [],
      "name": "Training autoencoder DB norm.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.10.7"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}