notebooks/07_conditional_dcgan.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "dsT1FPnBguaJ"
   },
   "source": [
    "# Keyword Spotting: GAN-based classification"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0lYQXwCmh402"
   },
   "source": [
    "## Get scripts from GitHub and install wget\n",
    "Only required when running on Colab. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:35:59.125170Z",
     "iopub.status.busy": "2023-01-28T22:35:59.124882Z",
     "iopub.status.idle": "2023-01-28T22:36:01.221938Z",
     "shell.execute_reply": "2023-01-28T22:36:01.221153Z",
     "shell.execute_reply.started": "2023-01-28T22:35:59.125112Z"
    },
    "id": "_V8fO0IpyMLd"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: wget in /usr/local/lib/python3.9/dist-packages (3.2)\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "! pip install wget"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:01.223664Z",
     "iopub.status.busy": "2023-01-28T22:36:01.223443Z",
     "iopub.status.idle": "2023-01-28T22:36:02.404650Z",
     "shell.execute_reply": "2023-01-28T22:36:02.403879Z",
     "shell.execute_reply.started": "2023-01-28T22:36:01.223642Z"
    },
    "id": "J7ZaKB5zhwiz"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File ‘models_utils.py’ already there; not retrieving.\n",
      "\n",
      "File ‘plot_utils.py’ already there; not retrieving.\n",
      "\n",
      "File ‘preprocessing_utils.py’ already there; not retrieving.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "! wget -nc https://raw.githubusercontent.com/NicolaZomer/Keyword_Spotting/main/utils/models_utils.py\n",
    "! wget -nc https://raw.githubusercontent.com/NicolaZomer/Keyword_Spotting/main/utils/plot_utils.py\n",
    "! wget -nc https://raw.githubusercontent.com/NicolaZomer/Keyword_Spotting/main/utils/preprocessing_utils.py"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ZFq83kJvyatv"
   },
   "source": [
    "## Import"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RMYTM7mrgufm"
   },
   "source": [
    "To extract MFCC features from audio signals we use the Python library [python_speech_features](https://github.com/jameslyons/python_speech_features)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:02.407789Z",
     "iopub.status.busy": "2023-01-28T22:36:02.407625Z",
     "iopub.status.idle": "2023-01-28T22:36:04.639807Z",
     "shell.execute_reply": "2023-01-28T22:36:04.639273Z",
     "shell.execute_reply.started": "2023-01-28T22:36:02.407731Z"
    },
    "id": "83mmHptRgufv"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: python_speech_features in /usr/local/lib/python3.9/dist-packages (0.6)\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "! pip install python_speech_features\n",
    "\n",
    "from python_speech_features import logfbank, mfcc, delta"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "7N6hBU0Tgugj"
   },
   "source": [
    "To perform Discrete Wavelet Transform on audio signals we use the Python library [PyWavelets](https://github.com/PyWavelets/pywt)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:04.641814Z",
     "iopub.status.busy": "2023-01-28T22:36:04.641243Z",
     "iopub.status.idle": "2023-01-28T22:36:06.642363Z",
     "shell.execute_reply": "2023-01-28T22:36:06.641814Z",
     "shell.execute_reply.started": "2023-01-28T22:36:04.641794Z"
    },
    "id": "MhiIcU5Xgugy"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: PyWavelets in /usr/local/lib/python3.9/dist-packages (1.3.0)\n",
      "Requirement already satisfied: numpy>=1.17.3 in /usr/local/lib/python3.9/dist-packages (from PyWavelets) (1.23.1)\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "! pip install PyWavelets\n",
    "\n",
    "from pywt import dwt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "4NHIHLP6I10a"
   },
   "source": [
    "To visualize neural network architectures we use the Python library [visualkeras](https://github.com/paulgavrikov/visualkeras)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:06.643209Z",
     "iopub.status.busy": "2023-01-28T22:36:06.643047Z",
     "iopub.status.idle": "2023-01-28T22:36:10.584619Z",
     "shell.execute_reply": "2023-01-28T22:36:10.583982Z",
     "shell.execute_reply.started": "2023-01-28T22:36:06.643192Z"
    },
    "id": "jML1YAgeJETu"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: visualkeras in /usr/local/lib/python3.9/dist-packages (0.0.2)\n",
      "Requirement already satisfied: pillow>=6.2.0 in /usr/local/lib/python3.9/dist-packages (from visualkeras) (9.2.0)\n",
      "Requirement already satisfied: aggdraw>=1.3.11 in /usr/local/lib/python3.9/dist-packages (from visualkeras) (1.3.15)\n",
      "Requirement already satisfied: numpy>=1.18.1 in /usr/local/lib/python3.9/dist-packages (from visualkeras) (1.23.1)\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "! pip install visualkeras\n",
    "\n",
    "import visualkeras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:10.585837Z",
     "iopub.status.busy": "2023-01-28T22:36:10.585472Z",
     "iopub.status.idle": "2023-01-28T22:36:11.183337Z",
     "shell.execute_reply": "2023-01-28T22:36:11.182720Z",
     "shell.execute_reply.started": "2023-01-28T22:36:10.585818Z"
    },
    "id": "Y_4uCFuXgucW"
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import pathlib\n",
    "import requests\n",
    "import wget\n",
    "import time\n",
    "import inspect\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras import Sequential, layers\n",
    "\n",
    "from IPython.display import display, Audio, Image\n",
    "\n",
    "from sklearn.metrics import confusion_matrix, precision_recall_fscore_support, accuracy_score, log_loss, cohen_kappa_score\n",
    "\n",
    "from scipy.io import wavfile\n",
    "\n",
    "from models_utils import available_models, select_model\n",
    "import plot_utils\n",
    "import preprocessing_utils\n",
    "from preprocessing_utils import create_dataset, remove_file_starting_with\n",
    "\n",
    "plt.rcParams['font.size'] = '12'\n",
    "# %matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:11.184543Z",
     "iopub.status.busy": "2023-01-28T22:36:11.184141Z",
     "iopub.status.idle": "2023-01-28T22:36:11.800548Z",
     "shell.execute_reply": "2023-01-28T22:36:11.799644Z",
     "shell.execute_reply.started": "2023-01-28T22:36:11.184507Z"
    },
    "id": "v58f7JtLicT4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /tmp/ipykernel_2078/868953920.py:1: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use `tf.config.list_physical_devices('GPU')` instead.\n",
      "GPU Available: True\n",
      "Version: 2.9.1\n"
     ]
    }
   ],
   "source": [
    "print(\"GPU Available:\", tf.test.is_gpu_available())\n",
    "print(\"Version:\", tf.__version__)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5J6-132RguqS"
   },
   "source": [
    "## 1 - Data loading and preprocessing\n",
    "In this project we use the [Speech Commands dataset](https://www.tensorflow.org/datasets/catalog/speech_commands), which contains short (one-second long) audio clips of English commands, stored as audio files in the WAV format. More in detail, the version 0.02 of the dataset contains 105.829 utterances of 35 short words, by thousands of different people. It was released on April 11th 2018 under [Creative Commons BY 4.0 license](https://creativecommons.org/licenses/by/4.0/) and collected using crowdsourcing, through [AIY](https://aiyprojects.withgoogle.com/) by Google. Some of these words are \"yes\", \"no\", \"up\", \"down\", \"left\", \"right\", \"on\", \"off\", \"stop\" and \"go\".\n",
    "\n",
    "### 1.1 - Download data and create data folder\n",
    "We import both a smaller version of the Speech Commands dataset and the full version. The mini dataset can be used for testing and demo reasons, however the entire work will be focused only on the complete dataset (or a subset of it). "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:11.801631Z",
     "iopub.status.busy": "2023-01-28T22:36:11.801432Z",
     "iopub.status.idle": "2023-01-28T22:36:11.807398Z",
     "shell.execute_reply": "2023-01-28T22:36:11.806565Z",
     "shell.execute_reply.started": "2023-01-28T22:36:11.801614Z"
    },
    "id": "qMEgF_1Oguqb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Working with the full dataset!\n"
     ]
    }
   ],
   "source": [
    "choose_dataset = 2 # 1 for mini, 2 for full\n",
    "download_also_mini = False\n",
    "\n",
    "DATASET_PATH_mini = 'data/mini_speech_commands'\n",
    "DATASET_PATH_full = 'data/full_speech_commands'\n",
    "\n",
    "data_dir_mini = pathlib.Path(DATASET_PATH_mini)\n",
    "if not data_dir_mini.exists() and download_also_mini:\n",
    "  tf.keras.utils.get_file(\n",
    "      'mini_speech_commands.zip',\n",
    "      origin=\"http://storage.googleapis.com/download.tensorflow.org/data/mini_speech_commands.zip\",\n",
    "      extract=True,\n",
    "      cache_dir='.', cache_subdir='data')\n",
    "  print('\\n')\n",
    "  \n",
    "data_dir_full = pathlib.Path(DATASET_PATH_full)\n",
    "if not data_dir_full.exists():\n",
    "  tf.keras.utils.get_file(\n",
    "      'full_speech_commands.zip',\n",
    "      origin=\"http://download.tensorflow.org/data/speech_commands_v0.02.tar.gz\",\n",
    "      extract=True,\n",
    "      cache_dir='.', cache_subdir='data/full_speech_commands')\n",
    "  print('\\n')\n",
    "  \n",
    "if choose_dataset==1:\n",
    "  data_dir = data_dir_mini\n",
    "  data_path = DATASET_PATH_mini\n",
    "  print('Working with the mini dataset!')\n",
    "elif choose_dataset==2:\n",
    "  data_dir = data_dir_full\n",
    "  data_path = DATASET_PATH_full\n",
    "  print('Working with the full dataset!')\n",
    "else:\n",
    "  print(\"Error, 'choose_dataset' must be 1 or 2\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TajS9BZPgurM"
   },
   "source": [
    "### 1.2 - Dataset organization\n",
    "The audio files are organized into folders based on the word they contain. For more details see \\[[Warden18](https://arxiv.org/abs/1804.03209)\\].\n",
    "\n",
    "**Mini dataset** <br>\n",
    "The dataset's audio clips are stored in 8 folders corresponding to each speech command: `down`, `go`, `left`, `no`, `right`, `stop`, `up`, `yes`, and `stop`. \n",
    "\n",
    "**Full dataset** <br>\n",
    "The dataset's audio clips are stored in 35 folders corresponding to each speech command: <br>\n",
    "`backward`, `bed`, `bird`, `cat`, `dog`, `down`, `eight`, `five`, `follow`, `forward`, `four`, `go`, `happy`, `house`, `learn`, `left`, `marvin`, `nine`, `no`, `off`, `on`, `one`, `right`, `seven`, `sheila`, `six`, `stop`, `three`, `tree`, `two`, `up`, `visual`, `wow`, `yes`, `zero`\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:11.808373Z",
     "iopub.status.busy": "2023-01-28T22:36:11.808199Z",
     "iopub.status.idle": "2023-01-28T22:36:11.813424Z",
     "shell.execute_reply": "2023-01-28T22:36:11.812866Z",
     "shell.execute_reply.started": "2023-01-28T22:36:11.808358Z"
    },
    "id": "gE9p410cgur8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Commands: ['backward', 'bed', 'bird', 'cat', 'dog', 'down', 'eight', 'five', 'follow', 'forward', 'four', 'go', 'happy', 'house', 'learn', 'left', 'marvin', 'nine', 'no', 'off', 'on', 'one', 'right', 'seven', 'sheila', 'six', 'stop', 'three', 'tree', 'two', 'up', 'visual', 'wow', 'yes', 'zero']\n",
      "\n",
      "Number of commands: 35\n"
     ]
    }
   ],
   "source": [
    "commands = np.array(tf.io.gfile.listdir(str(data_dir)))\n",
    "\n",
    "if choose_dataset==1:\n",
    "  mask = commands != 'README.md'\n",
    "\n",
    "elif choose_dataset==2:\n",
    "  mask = (\n",
    "      np.array(commands != 'README.md')                 &\n",
    "      np.array(commands != 'LICENSE')                   &\n",
    "      np.array(commands != '.DS_Store')                 &\n",
    "      np.array(commands != 'training_list.txt')         &\n",
    "      np.array(commands != 'validation_list.txt')       &\n",
    "      np.array(commands != 'testing_list.txt')          &\n",
    "      np.array(commands != 'full_speech_commands.zip')  &\n",
    "      np.array(commands != '_background_noise_')        \n",
    "  )\n",
    "  \n",
    "commands = sorted(commands[mask])\n",
    "print('Commands:', commands)\n",
    "\n",
    "print('\\nNumber of commands: %i' %len(commands))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "N_HQ4GIRguvw"
   },
   "source": [
    "### 1.3 - Processing of the original audio files \n",
    "The original audio files were captured in a variety of formats and then converted to .WAV file at a 16000 sample rate. The audio was then trimmed to a one second length to align most utterances, using the [extract_loudest_section](https://github.com/petewarden/extract_loudest_section) tool. The audio files were then screened for silence or incorrect words, and arranged into folders by label.\n",
    "\n",
    "### 1.4 - Training, testing and validation sets\n",
    "The text files `validation_list.txt` and `testing_list.txt` available in the [Speech Commands dataset](https://www.tensorflow.org/datasets/catalog/speech_commands) contain the paths to all the files in each set, with each path on a new line. Any files that aren't in either of these lists can be considered to be part of the training set. The validation and test set sizes thus obtained are approximatly 10% of the size of the complete set. For more information on how the partition is made, see the README.md file of the dataset. \n",
    "\n",
    "The following code performs these operations:\n",
    "- Load the reference validation and testing files as separated Pandas dataframe objects. \n",
    "- Store the path of all audio files in a dictionary. \n",
    "- Create the reference file for training data, by storing in a Pandas dataframe object the paths of all files that aren't in the validation or testing list. If it has been already created, we simply import it as we did for the reference validation and testing files. \n",
    "\n",
    "To each file we associate the corresponding label (i.e. the name of the folder it is in), so that we get a proper labeling for each file to be used with our loss function. Of course, we still need to load the actual data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:11.814159Z",
     "iopub.status.busy": "2023-01-28T22:36:11.814000Z",
     "iopub.status.idle": "2023-01-28T22:36:12.931668Z",
     "shell.execute_reply": "2023-01-28T22:36:12.931002Z",
     "shell.execute_reply.started": "2023-01-28T22:36:11.814144Z"
    },
    "id": "_fygRaSrguv0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are 105829 audio files in the dataset\n"
     ]
    }
   ],
   "source": [
    "training_list_path = data_path + \"/training_list.txt\"\n",
    "\n",
    "reference_df_val = pd.read_csv(data_path+\"/validation_list.txt\", sep=\"/\", header=None, names=['label', 'file'])\n",
    "reference_df_test = pd.read_csv(data_path+\"/testing_list.txt\", sep=\"/\", header=None, names=['label', 'file'])\n",
    "\n",
    "# all audio files\n",
    "all_files = {}\n",
    "for command in commands:\n",
    "    all_files[command] = os.listdir(data_path+'/'+command)\n",
    "    \n",
    "num_files = 0\n",
    "for x_ in all_files.values():\n",
    "    num_files += len(x_)\n",
    "        \n",
    "print('There are %i audio files in the dataset' %num_files)\n",
    "\n",
    "# reference file for training set\n",
    "# if already exists import it\n",
    "if pathlib.Path(training_list_path).exists(): \n",
    "    reference_df_train =  pd.read_csv(data_path+\"/training_list.txt\", sep=\"/\", header=None, names=['label', 'file'])\n",
    "    \n",
    "# else search if it can be dowloaded\n",
    "else: \n",
    "  training_list_url = 'https://raw.githubusercontent.com/NicolaZomer/Keyword_Spotting/main/training_list.txt'\n",
    "  response = requests.get(training_list_url)\n",
    "\n",
    "  if response.status_code == 200:\n",
    "      wget.download(training_list_url, out=str(data_path))\n",
    "      print('File downloaded from GitHub')\n",
    "\n",
    "      reference_df_train =  pd.read_csv(data_path+\"/training_list.txt\", sep=\"/\", header=None, names=['label', 'file'])\n",
    "\n",
    "  else: # create it and store the dataframe as a .txt file; creating the pd object can take some time\n",
    "      print('File not found, error '+str(response.status_code))\n",
    "\n",
    "      reference_df_train = pd.DataFrame(columns=['label', 'file'])\n",
    "      for i, command in enumerate(commands):\n",
    "          print(\"Processing command '%s', %i commands missing...\" %(command, len(commands)-i-1))\n",
    "          for file in all_files[command]:\n",
    "              if file in reference_df_val['file'].values:\n",
    "                  continue\n",
    "              if file in reference_df_test['file'].values:\n",
    "                  continue\n",
    "\n",
    "              reference_df_train.loc[len(reference_df_train)] = [command, file]\n",
    "\n",
    "      print('Done!')\n",
    "      print('\\n')\n",
    "\n",
    "      # store reference_df_train in .txt file\n",
    "      reference_df_train.to_csv(data_path+'/training_list.txt', header=None, index=None, sep='/', mode='a')\n",
    "\n",
    "# print the result\n",
    "# print('----------------------------')\n",
    "# print('TRAINING FILES:')\n",
    "# print(reference_df_train)\n",
    "# print('----------------------------')\n",
    "# print('VALIDATION FILES:')\n",
    "# print(reference_df_val)\n",
    "# print('----------------------------')\n",
    "# print('TESTING FILES:')\n",
    "# print(reference_df_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2naK4mEeguv5"
   },
   "source": [
    "Encode target labels with value between 0 and n_commands-1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:12.932727Z",
     "iopub.status.busy": "2023-01-28T22:36:12.932529Z",
     "iopub.status.idle": "2023-01-28T22:36:12.936198Z",
     "shell.execute_reply": "2023-01-28T22:36:12.935427Z",
     "shell.execute_reply.started": "2023-01-28T22:36:12.932710Z"
    },
    "id": "tyTfKJQbguv6"
   },
   "outputs": [],
   "source": [
    "label_to_class = {commands[i]:i for i in range(len(commands))}\n",
    "# label_to_class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:12.937097Z",
     "iopub.status.busy": "2023-01-28T22:36:12.936929Z",
     "iopub.status.idle": "2023-01-28T22:36:12.940225Z",
     "shell.execute_reply": "2023-01-28T22:36:12.939594Z",
     "shell.execute_reply.started": "2023-01-28T22:36:12.937081Z"
    },
    "id": "9bEaJKJkguv7"
   },
   "outputs": [],
   "source": [
    "class_to_label = {i:commands[i] for i in range(len(commands))}\n",
    "# class_to_label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:12.942841Z",
     "iopub.status.busy": "2023-01-28T22:36:12.942683Z",
     "iopub.status.idle": "2023-01-28T22:36:12.960183Z",
     "shell.execute_reply": "2023-01-28T22:36:12.959531Z",
     "shell.execute_reply.started": "2023-01-28T22:36:12.942828Z"
    },
    "id": "8vgwIDfSguv8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TRAINING FILES:\n",
      "          label                   file  class\n",
      "0      backward  0165e0e8_nohash_0.wav      0\n",
      "1      backward  017c4098_nohash_0.wav      0\n",
      "2      backward  017c4098_nohash_1.wav      0\n",
      "3      backward  017c4098_nohash_2.wav      0\n",
      "4      backward  017c4098_nohash_3.wav      0\n",
      "...         ...                    ...    ...\n",
      "84838      zero  ffd2ba2f_nohash_1.wav     34\n",
      "84839      zero  ffd2ba2f_nohash_2.wav     34\n",
      "84840      zero  ffd2ba2f_nohash_3.wav     34\n",
      "84841      zero  ffd2ba2f_nohash_4.wav     34\n",
      "84842      zero  fffcabd1_nohash_0.wav     34\n",
      "\n",
      "[84843 rows x 3 columns]\n",
      "----------------------------\n",
      "VALIDATION FILES:\n",
      "      label                   file  class\n",
      "0     right  a69b9b3e_nohash_0.wav     22\n",
      "1     right  439c84f4_nohash_1.wav     22\n",
      "2     right  409c962a_nohash_1.wav     22\n",
      "3     right  dbaf8fc6_nohash_2.wav     22\n",
      "4     right  a6d586b7_nohash_1.wav     22\n",
      "...     ...                    ...    ...\n",
      "9976   four  d107dc42_nohash_0.wav     10\n",
      "9977   four  ad63d93c_nohash_0.wav     10\n",
      "9978   four  d3831f6a_nohash_1.wav     10\n",
      "9979   four  3c4aa5ef_nohash_3.wav     10\n",
      "9980   four  c6389ab0_nohash_0.wav     10\n",
      "\n",
      "[9981 rows x 3 columns]\n",
      "----------------------------\n",
      "TESTING FILES:\n",
      "       label                   file  class\n",
      "0      right  bb05582b_nohash_3.wav     22\n",
      "1      right  97f4c236_nohash_2.wav     22\n",
      "2      right  f2e59fea_nohash_3.wav     22\n",
      "3      right  fdb5155e_nohash_2.wav     22\n",
      "4      right  dc75148d_nohash_0.wav     22\n",
      "...      ...                    ...    ...\n",
      "11000   four  189cbabe_nohash_2.wav     10\n",
      "11001   four  8fe67225_nohash_3.wav     10\n",
      "11002   four  c9e251d2_nohash_1.wav     10\n",
      "11003   four  0c40e715_nohash_0.wav     10\n",
      "11004   four  8769c34c_nohash_3.wav     10\n",
      "\n",
      "[11005 rows x 3 columns]\n"
     ]
    }
   ],
   "source": [
    "reference_df_train['class'] = reference_df_train['label'].map(label_to_class)\n",
    "reference_df_val['class']   = reference_df_val['label'].map(label_to_class)\n",
    "reference_df_test['class']  = reference_df_test['label'].map(label_to_class)\n",
    "\n",
    "# print the result\n",
    "print('TRAINING FILES:')\n",
    "print(reference_df_train)\n",
    "print('----------------------------')\n",
    "print('VALIDATION FILES:')\n",
    "print(reference_df_val)\n",
    "print('----------------------------')\n",
    "print('TESTING FILES:')\n",
    "print(reference_df_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YWHq02qWguwV"
   },
   "source": [
    "#### 1.3.1 - Smoke test\n",
    "We enable the possibility to perform some tests without loading the entire dataset by setting `smoke_test=True`. With this option the train, validation and test set are reduced to a fraction `smoke_test_frac` of elements. This can be useful to test if everything is working as intended before the full training, or if the current hardware is not powerful enough. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:12.961056Z",
     "iopub.status.busy": "2023-01-28T22:36:12.960882Z",
     "iopub.status.idle": "2023-01-28T22:36:12.965939Z",
     "shell.execute_reply": "2023-01-28T22:36:12.965199Z",
     "shell.execute_reply.started": "2023-01-28T22:36:12.961041Z"
    },
    "id": "kWb3R_eQguwW"
   },
   "outputs": [],
   "source": [
    "smoke_test = False\n",
    "smoke_test_frac = 0.15\n",
    "smoke_test_threshold = 1000 # minimum length of reduced datasets\n",
    "\n",
    "if smoke_test:\n",
    "    len_min = smoke_test_frac * min(len(reference_df_train), len(reference_df_val), len(reference_df_test))\n",
    "    if len_min < smoke_test_threshold:\n",
    "        raise Exception(f'One of the reduced datasets would have a number of samples < {smoke_test_threshold}! Select a larger value for \"smoke_test_frac\".')\n",
    "    else: \n",
    "        reference_df_train = reference_df_train.sample(frac=smoke_test_frac, replace=False)\n",
    "        reference_df_val   = reference_df_val.sample(frac=smoke_test_frac, replace=False)\n",
    "        reference_df_test  = reference_df_test.sample(frac=smoke_test_frac, replace=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:12.966737Z",
     "iopub.status.busy": "2023-01-28T22:36:12.966582Z",
     "iopub.status.idle": "2023-01-28T22:36:12.971255Z",
     "shell.execute_reply": "2023-01-28T22:36:12.970643Z",
     "shell.execute_reply.started": "2023-01-28T22:36:12.966723Z"
    },
    "id": "QdOpKmpwguwi"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TRAINING SET\n",
      "\tnumber of audio files: 84843\n",
      "\tpercentage of audio files: 80.17%\n",
      "VALIDATION SET\n",
      "\tnumber of audio files: 9981\n",
      "\tpercentage of audio files: 9.43%\n",
      "TEST SET\n",
      "\tnumber of audio files: 11005\n",
      "\tpercentage of audio files: 10.40%\n"
     ]
    }
   ],
   "source": [
    "print('TRAINING SET')\n",
    "print('\\tnumber of audio files: %i' %len(reference_df_train))\n",
    "print('\\tpercentage of audio files: %.2f%%' %(len(reference_df_train)/num_files*100))\n",
    "\n",
    "print('VALIDATION SET')\n",
    "print('\\tnumber of audio files: %i' %len(reference_df_val))\n",
    "print('\\tpercentage of audio files: %.2f%%' %(len(reference_df_val)/num_files*100))\n",
    "\n",
    "print('TEST SET')\n",
    "print('\\tnumber of audio files: %i' %len(reference_df_test))\n",
    "print('\\tpercentage of audio files: %.2f%%' %(len(reference_df_test)/num_files*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Qpc9ckEFguwm"
   },
   "source": [
    "#### 1.3.2 - Dataset distribution\n",
    "Let's now plot the percentage of elements of each class in each set. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:12.972377Z",
     "iopub.status.busy": "2023-01-28T22:36:12.972218Z",
     "iopub.status.idle": "2023-01-28T22:36:13.709453Z",
     "shell.execute_reply": "2023-01-28T22:36:13.708844Z",
     "shell.execute_reply.started": "2023-01-28T22:36:12.972364Z"
    },
    "id": "YhKiDxNuguwo"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1008x1512 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(3, 1, figsize=(14, 21))\n",
    "\n",
    "plot_utils.hist_class_distribution(reference_df_train, ax[0], commands)\n",
    "plot_utils.hist_class_distribution(reference_df_val, ax[1], commands)\n",
    "plot_utils.hist_class_distribution(reference_df_test, ax[2], commands)\n",
    "\n",
    "for i in range(3):\n",
    "    ax[i].set_ylabel('Normalized counts')\n",
    "    ax[i].grid(alpha=0.5)\n",
    "    ax[i].tick_params(axis='x', labelrotation=90)\n",
    "    \n",
    "ax[0].set_title('Training set distribution', fontsize=18)\n",
    "ax[1].set_title('Validation set distribution', fontsize=18)\n",
    "ax[2].set_title('Testing set distribution', fontsize=18)\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "KLmeFdAExBOD"
   },
   "source": [
    "### 1.4 Noise loading \n",
    "We load here the available noisy audio signals and we defined the object `noise_dict`, which is required by the function `load_and_preprocessing_data` (see next section)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:13.710675Z",
     "iopub.status.busy": "2023-01-28T22:36:13.710487Z",
     "iopub.status.idle": "2023-01-28T22:36:13.726865Z",
     "shell.execute_reply": "2023-01-28T22:36:13.726244Z",
     "shell.execute_reply.started": "2023-01-28T22:36:13.710657Z"
    },
    "id": "jDxNRt3Rw64J"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/notebooks/preprocessing_utils.py:70: WavFileWarning: Chunk (non-data) not understood, skipping it.\n",
      "  _, data = wavfile.read(file_path)\n"
     ]
    }
   ],
   "source": [
    "noise_path = '_background_noise_'\n",
    "\n",
    "# load all noises \n",
    "noise_1 = preprocessing_utils.load_data('doing_the_dishes.wav', noise_path)\n",
    "noise_2 = preprocessing_utils.load_data('dude_miaowing.wav', noise_path)\n",
    "noise_3 = preprocessing_utils.load_data('exercise_bike.wav', noise_path)\n",
    "noise_4 = preprocessing_utils.load_data('pink_noise.wav', noise_path)\n",
    "noise_5 = preprocessing_utils.load_data('running_tap.wav', noise_path)\n",
    "noise_6 = preprocessing_utils.load_data('white_noise.wav', noise_path)\n",
    "\n",
    "noise_dict = {\n",
    "    '1': noise_1, \n",
    "    '2': noise_2, \n",
    "    '3': noise_3, \n",
    "    '4': noise_4, \n",
    "    '5': noise_5, \n",
    "    '6': noise_6\n",
    "}"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "id": "af53Vbe4gu07"
   },
   "source": [
    "## 2 - Data Loading using the Dataset API\n",
    "As the dataset is quite large (>5GiB), there is the risk that it will not entirely fit in the memory, plus the preprocessing pipeline would be computationally intensive. For these reasons, we want to implement a proper and optimized data-loading pipeline, as reading the data can be the main bottleneck of the entire training process. This can be done by exploit the [Tensorflow Dataset API](https://www.tensorflow.org/guide/datasets).\n",
    "\n",
    "The main goal of this phase is to create a `tf.data.Dataset` object to efficiently load and preprocess your data. \n",
    "\n",
    "### 2.1 Loading and preprocessing\n",
    "First, to load and preprocess the data, we use the function `load_and_preprocess_data`, defined in the file `preprocessing_utils.py`. This function puts together all loading and preprocessing methods that can be explored in the notebook `B1_data_analysis_and_preprocessing_inspection.ipynb`. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:13.727856Z",
     "iopub.status.busy": "2023-01-28T22:36:13.727661Z",
     "iopub.status.idle": "2023-01-28T22:36:13.732443Z",
     "shell.execute_reply": "2023-01-28T22:36:13.731877Z",
     "shell.execute_reply.started": "2023-01-28T22:36:13.727839Z"
    },
    "id": "_X9BLSnXxZ3m"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "def load_and_preprocess_data(file_name, file_label, data_path_=data_path):\n",
      "\n",
      "    # load data\n",
      "    data = load_data(file_name, file_label, data_path_=data_path_)\n",
      "    \n",
      "    # padding/trimming\n",
      "    data = padding_trimming(data)\n",
      "        \n",
      "    # TensorFlow takes as input 32-bit floating point data\n",
      "    return data.astype(np.float32)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(inspect.getsource(preprocessing_utils.load_and_preprocess_data))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "id": "2fIv5VEdgu1a"
   },
   "source": [
    "### 2.2 - Dataset definition\n",
    "We now use the function `create_dataset`, defined in the file `preprocessing_utils.py`. This function initializes a `tf.data.Dataset` object and properly maps all the required processing. More details on this procedure can be found [here](https://www.tensorflow.org/guide/performance/datasets), along with additional tips and tricks for performance optimization."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:13.733286Z",
     "iopub.status.busy": "2023-01-28T22:36:13.733125Z",
     "iopub.status.idle": "2023-01-28T22:36:13.736516Z",
     "shell.execute_reply": "2023-01-28T22:36:13.736018Z",
     "shell.execute_reply.started": "2023-01-28T22:36:13.733272Z"
    },
    "id": "Az6lLEjixidk"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "def create_dataset(df, is_train=True, data_path_=data_path, cache_file=None, shuffle=True, apply_random_shift=False, apply_background_noise=False, noise_dict=None, noise_reduction=0.5, features=1, batch_size=32):\n",
      "    '''\n",
      "    features:\n",
      "    - 1 for MFCC features [delta_order=2] (default)\n",
      "    - 2 for log Mel-filterbank energy features\n",
      "    - 3 for spectrogram\n",
      "    - 4 for Discrete Wavelet Transform + MFCC features\n",
      "    - 5 for MFCC features [delta_order=0]\n",
      "    - 6 for log Mel-filterbank energy features [winlen=32, winstep=15.5, nfilt=64]\n",
      "    - 7 for log Mel-filterbank energy features [winlen=25, winstep=8,    nfilt=80]\n",
      "    '''\n",
      "\n",
      "    # Convert DataFrame to lists\n",
      "    file_names  = df['file'].tolist()\n",
      "    file_labels = df['class'].tolist()\n",
      "\n",
      "    # Create a Dataset object\n",
      "    dataset = tf.data.Dataset.from_tensor_slices((file_names, file_labels))\n",
      "\n",
      "    # Map the \"load_and_preprocess_data\" function\n",
      "    dataset = dataset.map(lambda file_name, file_label: (tf.numpy_function(load_and_preprocess_data, [file_name, file_label, data_path_], tf.float32), file_label), num_parallel_calls=os.cpu_count())\n",
      "\n",
      "    if is_train:\n",
      "\n",
      "        # Cache\n",
      "        if cache_file:\n",
      "            dataset = dataset.cache(filename=cache_file)\n",
      "\n",
      "        # Shuffle\n",
      "        if shuffle:\n",
      "            dataset = dataset.shuffle(buffer_size=len(df))\n",
      "\n",
      "        # Repeat the dataset indefinitely\n",
      "        dataset = dataset.repeat()\n",
      "\n",
      "    # Map the \"random_time_shift\" function\n",
      "    if apply_random_shift:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(random_time_shift, [data], tf.float32), label))\n",
      "\n",
      "    # Map the \"background_noise\" function\n",
      "    if apply_background_noise and np.random.uniform() < 0.8:\n",
      "        noise_list = [tf.convert_to_tensor(noise, dtype=tf.float32) for noise in noise_dict.values()]\n",
      "        dataset    = dataset.map(lambda data, label: (tf.numpy_function(background_noise, [data, -1, -1, noise_reduction, *noise_list], tf.float32), label))\n",
      "\n",
      "    # Extract features\n",
      "    if   features == 1:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_mfcc,        [data], tf.float32), label))\n",
      "    elif features == 2:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_logfbank,    [data], tf.float32), label))\n",
      "    elif features == 3:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_spectrogram, [data], tf.float32), label))\n",
      "    elif features == 4:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_dwt,         [data], tf.float32), label))\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_mfcc,        [data], tf.float32), label))\n",
      "    elif features == 5:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_mfcc,        [data, 0], tf.float32), label))\n",
      "    elif features == 6:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_logfbank,    [data, 16000, 32, 15.5, 64], tf.float32), label))\n",
      "    elif features == 7:\n",
      "        dataset = dataset.map(lambda data, label: (tf.numpy_function(get_logfbank,    [data, 16000, 25, 8, 80], tf.float32), label))\n",
      "\n",
      "    if not is_train:\n",
      "\n",
      "        # Cache\n",
      "        if cache_file:\n",
      "            dataset = dataset.cache(filename=cache_file)\n",
      "\n",
      "        # Shuffle\n",
      "        if shuffle:\n",
      "            dataset = dataset.shuffle(buffer_size=len(df))\n",
      "\n",
      "        # Repeat the dataset indefinitely\n",
      "        dataset = dataset.repeat()\n",
      "\n",
      "    # Batch\n",
      "    dataset = dataset.batch(batch_size=batch_size)\n",
      "\n",
      "    # Prefetch\n",
      "    dataset = dataset.prefetch(buffer_size=1)\n",
      "\n",
      "    # Steps\n",
      "    steps = int(np.ceil(len(df) / batch_size))\n",
      "\n",
      "    return dataset, steps\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(inspect.getsource(preprocessing_utils.create_dataset))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "id": "g7wcCw0Kgu1b"
   },
   "source": [
    "### 2.3 - Dataset initialization\n",
    "Finally, we define the training, validation and testing dataset, applying the above function to each of the corresponding reference dataframes. We also evaluate the number of steps (*train_steps*, *val_steps* and *test_steps*) required to load and process the entire dataset (num. of samples / batch_size).\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:13.743050Z",
     "iopub.status.busy": "2023-01-28T22:36:13.742888Z",
     "iopub.status.idle": "2023-01-28T22:36:14.258348Z",
     "shell.execute_reply": "2023-01-28T22:36:14.257739Z",
     "shell.execute_reply.started": "2023-01-28T22:36:13.743036Z"
    },
    "id": "i2c5zxUVgu1b"
   },
   "outputs": [],
   "source": [
    "'''\n",
    "features:\n",
    "- 1 for MFCC features [delta_order=2] (default)\n",
    "- 2 for log Mel-filterbank energy features\n",
    "- 3 for spectrogram\n",
    "- 4 for Discrete Wavelet Transform + MFCC features\n",
    "- 5 for MFCC features [delta_order=0]\n",
    "- 6 for log Mel-filterbank energy features [winlen=32, winstep=15.5, nfilt=64]\n",
    "- 7 for log Mel-filterbank energy features [winlen=25, winstep=8,    nfilt=80]\n",
    "'''\n",
    "\n",
    "apply_bn = True\n",
    "features_to_extract = 6\n",
    "batch_size = 32\n",
    "\n",
    "train_dataset, train_steps = create_dataset(\n",
    "    reference_df_train,\n",
    "    is_train=True,\n",
    "    cache_file='train_cache',\n",
    "    shuffle=True,\n",
    "    apply_background_noise=apply_bn,\n",
    "    noise_dict=noise_dict,\n",
    "    features=features_to_extract,\n",
    "    batch_size=batch_size\n",
    ")\n",
    "\n",
    "val_dataset, val_steps = create_dataset(\n",
    "    reference_df_val,\n",
    "    is_train=False,\n",
    "    cache_file='val_cache',\n",
    "    shuffle=True,\n",
    "    apply_background_noise=False,\n",
    "    features=features_to_extract,\n",
    "    batch_size=batch_size\n",
    ")\n",
    "\n",
    "# important! shuffle=False\n",
    "test_dataset, test_steps = create_dataset(\n",
    "    reference_df_test,\n",
    "    is_train=False,\n",
    "    cache_file='test_cache',\n",
    "    shuffle=False,\n",
    "    apply_background_noise=False,\n",
    "    features=features_to_extract,\n",
    "    batch_size=batch_size\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1dp116A5gu1c"
   },
   "source": [
    "## 3 - GAN-based classifier model\n",
    "References:\n",
    "- [Deep Convolutional Generative Adversarial Network | TensorFlow](https://www.tensorflow.org/tutorials/generative/dcgan)\n",
    "- [GAN-based Data Generation for Speech Emotion Recognition | Microsoft](https://isca-speech.org/archive_v0/Interspeech_2020/pdfs/2898.pdf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:14.259448Z",
     "iopub.status.busy": "2023-01-28T22:36:14.259249Z",
     "iopub.status.idle": "2023-01-28T22:36:16.435453Z",
     "shell.execute_reply": "2023-01-28T22:36:16.434707Z",
     "shell.execute_reply.started": "2023-01-28T22:36:14.259430Z"
    },
    "id": "R8FZJmRkjFXf"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: tensorflow-addons in /usr/local/lib/python3.9/dist-packages (0.19.0)\n",
      "Requirement already satisfied: packaging in /usr/local/lib/python3.9/dist-packages (from tensorflow-addons) (21.3)\n",
      "Requirement already satisfied: typeguard>=2.7 in /usr/local/lib/python3.9/dist-packages (from tensorflow-addons) (2.13.3)\n",
      "Requirement already satisfied: pyparsing!=3.0.5,>=2.0.2 in /usr/local/lib/python3.9/dist-packages (from packaging->tensorflow-addons) (3.0.9)\n",
      "\u001b[33mWARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv\u001b[0m\u001b[33m\n",
      "\u001b[0m"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras import Sequential\n",
    "from tensorflow.keras.layers import Dense, Reshape, Conv2D, UpSampling2D, LeakyReLU, Flatten, Activation, BatchNormalization\n",
    "\n",
    "! pip install tensorflow-addons\n",
    "from tensorflow_addons.layers import SpectralNormalization\n",
    "\n",
    "from tensorflow.keras.losses import BinaryCrossentropy, CategoricalCrossentropy\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau\n",
    "\n",
    "from tensorflow.keras.models import Model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6ehk_BTxgu1d"
   },
   "source": [
    "### 3.1 - Create the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.436630Z",
     "iopub.status.busy": "2023-01-28T22:36:16.436413Z",
     "iopub.status.idle": "2023-01-28T22:36:16.440549Z",
     "shell.execute_reply": "2023-01-28T22:36:16.439664Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.436580Z"
    },
    "id": "vp4J8PSRM1j9"
   },
   "outputs": [],
   "source": [
    "input_shape = (64, 64)\n",
    "\n",
    "N_z = 128 # noise dimension\n",
    "N_c = 35  # number of classes\n",
    "\n",
    "use_large_models = False"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-Iok2-rPJBn5"
   },
   "source": [
    "#### 3.1.1 - Generator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.441326Z",
     "iopub.status.busy": "2023-01-28T22:36:16.441195Z",
     "iopub.status.idle": "2023-01-28T22:36:16.447563Z",
     "shell.execute_reply": "2023-01-28T22:36:16.447059Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.441326Z"
    },
    "id": "2XvcwRIKJ3jL"
   },
   "outputs": [],
   "source": [
    "def make_generator_model(name='generator', large=False):\n",
    "\n",
    "    model = Sequential(name=name)\n",
    "\n",
    "    if large:\n",
    "        \n",
    "        # 1st linear layer\n",
    "        model.add(SpectralNormalization(Dense(4096, input_shape=(N_z + N_c,))))\n",
    "        model.add(LeakyReLU())\n",
    "        \n",
    "        # 2nd linear layer\n",
    "        model.add(SpectralNormalization(Dense(32768)))\n",
    "        model.add(LeakyReLU())\n",
    "\n",
    "        # intermediate representation reshaped into 8x8 images\n",
    "        model.add(Reshape((8, 8, 512)))\n",
    "        \n",
    "        filter_sizes = [512, 256, 128]\n",
    "    \n",
    "    else:\n",
    "        \n",
    "        # 1st linear layer\n",
    "        model.add(SpectralNormalization(Dense(512, input_shape=(N_z + N_c,))))\n",
    "        model.add(LeakyReLU())\n",
    "        \n",
    "        # 2nd linear layer\n",
    "        model.add(SpectralNormalization(Dense(4096)))\n",
    "        model.add(LeakyReLU())\n",
    "        \n",
    "        # intermediate representation reshaped into 8x8 images\n",
    "        model.add(Reshape((8, 8, 64)))\n",
    "        \n",
    "        filter_sizes = [64, 32, 16]\n",
    "\n",
    "    for filter_size in filter_sizes:\n",
    "\n",
    "        # 1st convolutional layer\n",
    "        model.add(SpectralNormalization(Conv2D(filter_size, (3, 3), strides=(1, 1), padding='same')))\n",
    "        model.add(LeakyReLU())\n",
    "\n",
    "        # 2nd convolutional layer\n",
    "        model.add(SpectralNormalization(Conv2D(filter_size, (3, 3), strides=(2, 2), padding='same')))\n",
    "        model.add(LeakyReLU())\n",
    "        \n",
    "        # upsampling (nearest-neighbor interpolation)\n",
    "        model.add(UpSampling2D(size=(4, 4), interpolation='nearest'))\n",
    "\n",
    "    # last convolutional layer\n",
    "    model.add(SpectralNormalization(Conv2D(1, (3, 3), strides=(1, 1), padding='same')))\n",
    "    #model.add(Activation('tanh'))\n",
    "\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.448273Z",
     "iopub.status.busy": "2023-01-28T22:36:16.448084Z",
     "iopub.status.idle": "2023-01-28T22:36:16.613757Z",
     "shell.execute_reply": "2023-01-28T22:36:16.613077Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.448248Z"
    },
    "id": "bMUrunH2KCZL"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"pt_generator\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " spectral_normalization (Spe  (None, 512)              84480     \n",
      " ctralNormalization)                                             \n",
      "                                                                 \n",
      " leaky_re_lu (LeakyReLU)     (None, 512)               0         \n",
      "                                                                 \n",
      " spectral_normalization_1 (S  (None, 4096)             2105344   \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_1 (LeakyReLU)   (None, 4096)              0         \n",
      "                                                                 \n",
      " reshape (Reshape)           (None, 8, 8, 64)          0         \n",
      "                                                                 \n",
      " spectral_normalization_2 (S  (None, 8, 8, 64)         36992     \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_2 (LeakyReLU)   (None, 8, 8, 64)          0         \n",
      "                                                                 \n",
      " spectral_normalization_3 (S  (None, 4, 4, 64)         36992     \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_3 (LeakyReLU)   (None, 4, 4, 64)          0         \n",
      "                                                                 \n",
      " up_sampling2d (UpSampling2D  (None, 16, 16, 64)       0         \n",
      " )                                                               \n",
      "                                                                 \n",
      " spectral_normalization_4 (S  (None, 16, 16, 32)       18496     \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_4 (LeakyReLU)   (None, 16, 16, 32)        0         \n",
      "                                                                 \n",
      " spectral_normalization_5 (S  (None, 8, 8, 32)         9280      \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_5 (LeakyReLU)   (None, 8, 8, 32)          0         \n",
      "                                                                 \n",
      " up_sampling2d_1 (UpSampling  (None, 32, 32, 32)       0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " spectral_normalization_6 (S  (None, 32, 32, 16)       4640      \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_6 (LeakyReLU)   (None, 32, 32, 16)        0         \n",
      "                                                                 \n",
      " spectral_normalization_7 (S  (None, 16, 16, 16)       2336      \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_7 (LeakyReLU)   (None, 16, 16, 16)        0         \n",
      "                                                                 \n",
      " up_sampling2d_2 (UpSampling  (None, 64, 64, 16)       0         \n",
      " 2D)                                                             \n",
      "                                                                 \n",
      " spectral_normalization_8 (S  (None, 64, 64, 1)        146       \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 2,298,706\n",
      "Trainable params: 2,293,873\n",
      "Non-trainable params: 4,833\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "pt_generator = make_generator_model(name='pt_generator', large=use_large_models)\n",
    "\n",
    "pt_generator.build(input_shape=(None, N_z + N_c,))\n",
    "pt_generator.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.614642Z",
     "iopub.status.busy": "2023-01-28T22:36:16.614486Z",
     "iopub.status.idle": "2023-01-28T22:36:16.647902Z",
     "shell.execute_reply": "2023-01-28T22:36:16.647322Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.614642Z"
    },
    "id": "hv_nZJcBxy0Y"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<PIL.Image.Image image mode=RGBA size=1509x342>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "visualkeras.layered_view(pt_generator)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "upLnJxCIfKH9"
   },
   "source": [
    "#### 3.1.2 - Discriminator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.648832Z",
     "iopub.status.busy": "2023-01-28T22:36:16.648659Z",
     "iopub.status.idle": "2023-01-28T22:36:16.654617Z",
     "shell.execute_reply": "2023-01-28T22:36:16.653956Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.648790Z"
    },
    "id": "yFp8XUvgfaij"
   },
   "outputs": [],
   "source": [
    "def make_discriminator_model(name='discriminator', large=False):\n",
    "\n",
    "    model = Sequential(name=name)\n",
    "\n",
    "    model.add(Reshape(input_shape=input_shape, target_shape=(input_shape[0], input_shape[1], 1)))\n",
    "    #model.add(BatchNormalization())\n",
    "    \n",
    "    if large:\n",
    "\n",
    "        filter_sizes = [128, 256, 512]\n",
    "        \n",
    "    else:\n",
    "        \n",
    "        filter_sizes = [16, 32, 64]\n",
    "\n",
    "    for filter_size in filter_sizes:\n",
    "\n",
    "        # 1st convolutional layer\n",
    "        model.add(SpectralNormalization(Conv2D(filter_size, (3, 3), strides=(1, 1), padding='same')))\n",
    "        model.add(LeakyReLU())\n",
    "\n",
    "        # 2nd convolutional layer\n",
    "        model.add(SpectralNormalization(Conv2D(filter_size, (3, 3), strides=(2, 2), padding='same')))\n",
    "        model.add(LeakyReLU())\n",
    "\n",
    "    # intermediate representation flattened\n",
    "    model.add(Flatten())\n",
    "    \n",
    "    if large:\n",
    "\n",
    "        # 1st linear layer\n",
    "        model.add(SpectralNormalization(Dense(4096)))\n",
    "        model.add(LeakyReLU())\n",
    "        \n",
    "    else:\n",
    "        \n",
    "        # 1st linear layer\n",
    "        model.add(SpectralNormalization(Dense(512)))\n",
    "        model.add(LeakyReLU())\n",
    "\n",
    "    # 2nd linear layer\n",
    "    model.add(SpectralNormalization(Dense(1)))\n",
    "\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.655524Z",
     "iopub.status.busy": "2023-01-28T22:36:16.655355Z",
     "iopub.status.idle": "2023-01-28T22:36:16.752052Z",
     "shell.execute_reply": "2023-01-28T22:36:16.748520Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.655520Z"
    },
    "id": "lGI34t49gJXV"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"pt_discriminator\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " reshape_1 (Reshape)         (None, 64, 64, 1)         0         \n",
      "                                                                 \n",
      " spectral_normalization_9 (S  (None, 64, 64, 16)       176       \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_8 (LeakyReLU)   (None, 64, 64, 16)        0         \n",
      "                                                                 \n",
      " spectral_normalization_10 (  (None, 32, 32, 16)       2336      \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_9 (LeakyReLU)   (None, 32, 32, 16)        0         \n",
      "                                                                 \n",
      " spectral_normalization_11 (  (None, 32, 32, 32)       4672      \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_10 (LeakyReLU)  (None, 32, 32, 32)        0         \n",
      "                                                                 \n",
      " spectral_normalization_12 (  (None, 16, 16, 32)       9280      \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_11 (LeakyReLU)  (None, 16, 16, 32)        0         \n",
      "                                                                 \n",
      " spectral_normalization_13 (  (None, 16, 16, 64)       18560     \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_12 (LeakyReLU)  (None, 16, 16, 64)        0         \n",
      "                                                                 \n",
      " spectral_normalization_14 (  (None, 8, 8, 64)         36992     \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_13 (LeakyReLU)  (None, 8, 8, 64)          0         \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 4096)              0         \n",
      "                                                                 \n",
      " spectral_normalization_15 (  (None, 512)              2098176   \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_14 (LeakyReLU)  (None, 512)               0         \n",
      "                                                                 \n",
      " spectral_normalization_16 (  (None, 1)                514       \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 2,170,706\n",
      "Trainable params: 2,169,969\n",
      "Non-trainable params: 737\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "pt_discriminator = make_discriminator_model(name='pt_discriminator', large=use_large_models)\n",
    "    \n",
    "pt_discriminator.build(input_shape=(None, input_shape[0], input_shape[1],))\n",
    "pt_discriminator.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.753658Z",
     "iopub.status.busy": "2023-01-28T22:36:16.753101Z",
     "iopub.status.idle": "2023-01-28T22:36:16.775587Z",
     "shell.execute_reply": "2023-01-28T22:36:16.775026Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.753642Z"
    },
    "id": "fVF3qf7exuri"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<PIL.Image.Image image mode=RGBA size=1009x342>"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "visualkeras.layered_view(pt_discriminator)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 - Define the training loop"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.2.1 - Build subclassed model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.776942Z",
     "iopub.status.busy": "2023-01-28T22:36:16.776421Z",
     "iopub.status.idle": "2023-01-28T22:36:16.787357Z",
     "shell.execute_reply": "2023-01-28T22:36:16.786709Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.776925Z"
    },
    "id": "TZ31_1TNcYLE"
   },
   "outputs": [],
   "source": [
    "class GAN(Model):\n",
    "    \n",
    "    def __init__(self, generator, discriminator, fine_tuning, *args, **kwargs):\n",
    "        \n",
    "        # pass through args and kwargs to base class\n",
    "        super().__init__(*args, **kwargs)\n",
    "        \n",
    "        # create attributes for generator and discriminator\n",
    "        self.generator     = generator\n",
    "        self.discriminator = discriminator\n",
    "        \n",
    "        # create attribute for training type\n",
    "        self.fine_tuning = fine_tuning\n",
    "        \n",
    "    def compile(self, g_opt, d_opt, g_loss, d_loss, *args, **kwargs):\n",
    "        \n",
    "        # compile with base class\n",
    "        super().compile(*args, **kwargs)\n",
    "        \n",
    "        # create attributes for losses\n",
    "        self.g_loss = g_loss\n",
    "        self.d_loss = d_loss \n",
    "        \n",
    "        # create attributes for optimizers\n",
    "        self.g_opt = g_opt\n",
    "        self.d_opt = d_opt\n",
    "\n",
    "    def train_step(self, batch, *args, **kwargs):\n",
    "        \n",
    "        # noise\n",
    "        batch_size = tf.shape(batch[0])[0]\n",
    "        noise      = tf.random.normal([batch_size, N_z])\n",
    "        \n",
    "        # condition vector\n",
    "        if self.fine_tuning:\n",
    "            code = tf.one_hot(batch[1], depth=N_c)\n",
    "        else:\n",
    "            code = tf.zeros([batch_size, N_c])\n",
    "\n",
    "        # generator input\n",
    "        g_input = tf.concat(values=[noise, code], axis=1)\n",
    "        \n",
    "        # get the data \n",
    "        real_images = batch[0]\n",
    "        fake_images = self.generator(g_input, training=False)\n",
    "        \n",
    "        # train the discriminator\n",
    "        with tf.GradientTape() as d_tape:\n",
    "            \n",
    "            # get the predicted labels\n",
    "            yhat_real = self.discriminator(real_images, training=True) \n",
    "            yhat_fake = self.discriminator(fake_images, training=True)\n",
    "            yhat      = tf.concat([yhat_real, yhat_fake], axis=0)\n",
    "            \n",
    "            # create the true labels\n",
    "            if self.fine_tuning:\n",
    "                y = tf.concat([tf.one_hot(batch[1], depth=N_c+1), tf.one_hot(tf.repeat(N_c, batch_size), depth=N_c+1)], axis=0)\n",
    "            else:\n",
    "                y = tf.concat([tf.zeros_like(yhat_real), tf.ones_like(yhat_fake)], axis=0)\n",
    "            \n",
    "            # add noise to the true labels\n",
    "            noise_real = 0.15*tf.random.uniform(tf.shape(yhat_real))\n",
    "            noise_fake = -0.15*tf.random.uniform(tf.shape(yhat_fake))\n",
    "            y         += tf.concat([noise_real, noise_fake], axis=0)\n",
    "            \n",
    "            # calculate loss\n",
    "            total_d_loss = self.d_loss(y, yhat)\n",
    "            \n",
    "        # apply backpropagation\n",
    "        d_grad = d_tape.gradient(total_d_loss, self.discriminator.trainable_variables) \n",
    "        self.d_opt.apply_gradients(zip(d_grad, self.discriminator.trainable_variables))\n",
    "        \n",
    "        # train the generator \n",
    "        with tf.GradientTape() as g_tape:\n",
    "            \n",
    "            # generate new images\n",
    "            gen_images = self.generator(g_input, training=True)\n",
    "                                        \n",
    "            # get the predicted labels\n",
    "            predicted_labels = self.discriminator(gen_images, training=False)\n",
    "                                        \n",
    "            # calculate loss\n",
    "            if self.fine_tuning:\n",
    "                total_g_loss = self.g_loss(tf.one_hot(batch[1], depth=N_c+1), predicted_labels)\n",
    "            else:\n",
    "                total_g_loss = self.g_loss(tf.zeros_like(predicted_labels), predicted_labels)\n",
    "            \n",
    "        # apply backpropagation\n",
    "        g_grad = g_tape.gradient(total_g_loss, self.generator.trainable_variables)\n",
    "        self.g_opt.apply_gradients(zip(g_grad, self.generator.trainable_variables))\n",
    "        \n",
    "        return {\"generator_loss\":total_g_loss, \"discriminator_loss\":total_d_loss}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ki8qVV1XSZjE"
   },
   "source": [
    "### 3.3 - Pre-training"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "N2K61H2hg1KA"
   },
   "source": [
    "#### 3.3.1 - Define losses and optimizers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.788231Z",
     "iopub.status.busy": "2023-01-28T22:36:16.788092Z",
     "iopub.status.idle": "2023-01-28T22:36:16.791604Z",
     "shell.execute_reply": "2023-01-28T22:36:16.790910Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.788227Z"
    },
    "id": "m5FrP20Nhtws"
   },
   "outputs": [],
   "source": [
    "g_loss = BinaryCrossentropy(from_logits=True)\n",
    "g_opt  = Adam(learning_rate=1e-4)#beta_1=0.5, beta_2=0.99, learning_rate=2e-5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.792589Z",
     "iopub.status.busy": "2023-01-28T22:36:16.792434Z",
     "iopub.status.idle": "2023-01-28T22:36:16.795473Z",
     "shell.execute_reply": "2023-01-28T22:36:16.794810Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.792578Z"
    },
    "id": "80jXMt70hZRd"
   },
   "outputs": [],
   "source": [
    "d_loss = BinaryCrossentropy(from_logits=True)\n",
    "d_opt  = Adam(learning_rate=1e-4)#beta_1=0.5, beta_2=0.99, learning_rate=2e-4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YAkQh6CQjF29"
   },
   "source": [
    "#### 3.3.2 - Pre-train the GAN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.796212Z",
     "iopub.status.busy": "2023-01-28T22:36:16.796096Z",
     "iopub.status.idle": "2023-01-28T22:36:16.808131Z",
     "shell.execute_reply": "2023-01-28T22:36:16.807457Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.796198Z"
    }
   },
   "outputs": [],
   "source": [
    "# create instance of subclassed model\n",
    "gan = GAN(pt_generator, pt_discriminator, fine_tuning=False)\n",
    "\n",
    "# compile the model\n",
    "gan.compile(g_opt, d_opt, g_loss, d_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.809225Z",
     "iopub.status.busy": "2023-01-28T22:36:16.809058Z",
     "iopub.status.idle": "2023-01-28T22:36:16.812033Z",
     "shell.execute_reply": "2023-01-28T22:36:16.811411Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.809218Z"
    },
    "id": "tZX4_jrUezS6"
   },
   "outputs": [],
   "source": [
    "# uncomment to remove cache files\n",
    "#remove_file_starting_with('train_cache')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:36:16.812986Z",
     "iopub.status.busy": "2023-01-28T22:36:16.812826Z",
     "iopub.status.idle": "2023-01-28T22:44:29.461399Z",
     "shell.execute_reply": "2023-01-28T22:44:29.460921Z",
     "shell.execute_reply.started": "2023-01-28T22:36:16.812986Z"
    },
    "id": "eg7d25OCjS1L"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "2652/2652 [==============================] - 43s 14ms/step - generator_loss: 2.0391 - discriminator_loss: 0.3764\n",
      "Epoch 2/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 2.1058 - discriminator_loss: 0.3802\n",
      "Epoch 3/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 2.3643 - discriminator_loss: 0.3278\n",
      "Epoch 4/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 2.5895 - discriminator_loss: 0.2865\n",
      "Epoch 5/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.6211 - discriminator_loss: 0.2766\n",
      "Epoch 6/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.6615 - discriminator_loss: 0.2739\n",
      "Epoch 7/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.5963 - discriminator_loss: 0.2872\n",
      "Epoch 8/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 2.6424 - discriminator_loss: 0.2719\n",
      "Epoch 9/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.6355 - discriminator_loss: 0.2690\n",
      "Epoch 10/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.6263 - discriminator_loss: 0.2754\n",
      "Epoch 11/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.6346 - discriminator_loss: 0.2682\n",
      "Epoch 12/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.6342 - discriminator_loss: 0.2682\n",
      "Epoch 13/100\n",
      "2652/2652 [==============================] - 37s 14ms/step - generator_loss: 2.6356 - discriminator_loss: 0.2685\n"
     ]
    }
   ],
   "source": [
    "num_epochs = 100\n",
    "\n",
    "g_early_stopping = EarlyStopping(monitor='generator_loss', patience=10)\n",
    "#d_early_stopping = EarlyStopping(monitor='discriminator_loss', patience=10)\n",
    "\n",
    "hist = gan.fit(\n",
    "            train_dataset,\n",
    "            epochs=num_epochs,\n",
    "            steps_per_epoch=train_steps,\n",
    "            callbacks=[g_early_stopping]\n",
    "            )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:29.462814Z",
     "iopub.status.busy": "2023-01-28T22:44:29.462193Z",
     "iopub.status.idle": "2023-01-28T22:44:41.205341Z",
     "shell.execute_reply": "2023-01-28T22:44:41.204867Z",
     "shell.execute_reply.started": "2023-01-28T22:44:29.462793Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "_, ax = plt.subplots(figsize=(8, 5))\n",
    "\n",
    "ax.plot(hist.history['generator_loss'], label='generator', color='blue', linewidth=1.5)\n",
    "ax.plot(hist.history['discriminator_loss'], label='discriminator', color='firebrick', linewidth=1.5)\n",
    "ax.set_xticks(range(len(hist.history['generator_loss'])))\n",
    "ax.set_xticklabels(range(1, len(hist.history['generator_loss'])+1))\n",
    "ax.set_xlabel('epoch')\n",
    "ax.set_ylabel('loss')\n",
    "ax.grid(alpha=0.5)\n",
    "ax.legend(edgecolor='black', facecolor='linen', fontsize=12, loc ='best')\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Uncomment the cell below to save the weights of the trained models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:41.206674Z",
     "iopub.status.busy": "2023-01-28T22:44:41.206168Z",
     "iopub.status.idle": "2023-01-28T22:44:42.083524Z",
     "shell.execute_reply": "2023-01-28T22:44:42.082879Z",
     "shell.execute_reply.started": "2023-01-28T22:44:41.206654Z"
    },
    "id": "MZPp_BCt0kSN"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.\n",
      "WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.\n"
     ]
    }
   ],
   "source": [
    "pt_generator.save('pt_generator.h5')\n",
    "pt_discriminator.save('pt_discriminator.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Uncomment the cell below to load pretrained models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.084459Z",
     "iopub.status.busy": "2023-01-28T22:44:42.084259Z",
     "iopub.status.idle": "2023-01-28T22:44:42.087259Z",
     "shell.execute_reply": "2023-01-28T22:44:42.086794Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.084442Z"
    }
   },
   "outputs": [],
   "source": [
    "#pt_generator     = tf.keras.models.load_model('pt_generator.h5')\n",
    "#pt_discriminator = tf.keras.models.load_model('pt_discriminator.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "QaZzdBSjXv-e"
   },
   "source": [
    "### 3.4 - Fine-tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3.4.1 - Create the final discriminator model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.088032Z",
     "iopub.status.busy": "2023-01-28T22:44:42.087859Z",
     "iopub.status.idle": "2023-01-28T22:44:42.166980Z",
     "shell.execute_reply": "2023-01-28T22:44:42.166363Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.088018Z"
    },
    "id": "FIalscZwXr4z"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"ft_discriminator\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " reshape_1 (Reshape)         (None, 64, 64, 1)         0         \n",
      "                                                                 \n",
      " spectral_normalization_9 (S  (None, 64, 64, 16)       176       \n",
      " pectralNormalization)                                           \n",
      "                                                                 \n",
      " leaky_re_lu_8 (LeakyReLU)   (None, 64, 64, 16)        0         \n",
      "                                                                 \n",
      " spectral_normalization_10 (  (None, 32, 32, 16)       2336      \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_9 (LeakyReLU)   (None, 32, 32, 16)        0         \n",
      "                                                                 \n",
      " spectral_normalization_11 (  (None, 32, 32, 32)       4672      \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_10 (LeakyReLU)  (None, 32, 32, 32)        0         \n",
      "                                                                 \n",
      " spectral_normalization_12 (  (None, 16, 16, 32)       9280      \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_11 (LeakyReLU)  (None, 16, 16, 32)        0         \n",
      "                                                                 \n",
      " spectral_normalization_13 (  (None, 16, 16, 64)       18560     \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_12 (LeakyReLU)  (None, 16, 16, 64)        0         \n",
      "                                                                 \n",
      " spectral_normalization_14 (  (None, 8, 8, 64)         36992     \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_13 (LeakyReLU)  (None, 8, 8, 64)          0         \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 4096)              0         \n",
      "                                                                 \n",
      " spectral_normalization_15 (  (None, 512)              2098176   \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " leaky_re_lu_14 (LeakyReLU)  (None, 512)               0         \n",
      "                                                                 \n",
      " spectral_normalization_17 (  (None, 36)               18504     \n",
      " SpectralNormalization)                                          \n",
      "                                                                 \n",
      " activation (Activation)     (None, 36)                0         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 2,188,696\n",
      "Trainable params: 2,187,924\n",
      "Non-trainable params: 772\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "ft_discriminator = Sequential(name='ft_discriminator')\n",
    "\n",
    "for i in range(len(pt_discriminator.layers) - 1):\n",
    "    ft_discriminator.add(pt_discriminator.layers[i])\n",
    "\n",
    "# unfreeze all layers\n",
    "for layer in ft_discriminator.layers:\n",
    "    layer.trainable = True\n",
    "\n",
    "ft_discriminator.add(SpectralNormalization(Dense(N_c + 1)))\n",
    "ft_discriminator.add(Activation('softmax'))\n",
    "\n",
    "ft_discriminator.build(input_shape=(None, input_shape[0], input_shape[1],))\n",
    "ft_discriminator.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.170520Z",
     "iopub.status.busy": "2023-01-28T22:44:42.170335Z",
     "iopub.status.idle": "2023-01-28T22:44:42.193329Z",
     "shell.execute_reply": "2023-01-28T22:44:42.192741Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.170505Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<PIL.Image.Image image mode=RGBA size=1039x342>"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "visualkeras.layered_view(ft_discriminator)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "E7ZN_CEUaMFa"
   },
   "source": [
    "#### 3.4.2 - Define losses and optimizers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.194298Z",
     "iopub.status.busy": "2023-01-28T22:44:42.194109Z",
     "iopub.status.idle": "2023-01-28T22:44:42.197535Z",
     "shell.execute_reply": "2023-01-28T22:44:42.197028Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.194283Z"
    },
    "id": "nVMQUZ4maUpi"
   },
   "outputs": [],
   "source": [
    "g_loss = CategoricalCrossentropy(from_logits=False)\n",
    "g_opt  = Adam(learning_rate=1e-4)#beta_1=0.5, beta_2=0.99, learning_rate=2e-5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.198223Z",
     "iopub.status.busy": "2023-01-28T22:44:42.198070Z",
     "iopub.status.idle": "2023-01-28T22:44:42.201267Z",
     "shell.execute_reply": "2023-01-28T22:44:42.200786Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.198210Z"
    },
    "id": "p48bxT2yaUpf"
   },
   "outputs": [],
   "source": [
    "d_loss = CategoricalCrossentropy(from_logits=False)\n",
    "d_opt  = Adam(learning_rate=1e-4)#beta_1=0.5, beta_2=0.99, learning_rate=2e-4)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0-ejnJeOaUpk"
   },
   "source": [
    "#### 3.4.3 - Fine-tune the GAN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.202079Z",
     "iopub.status.busy": "2023-01-28T22:44:42.201924Z",
     "iopub.status.idle": "2023-01-28T22:44:42.211378Z",
     "shell.execute_reply": "2023-01-28T22:44:42.210852Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.202066Z"
    }
   },
   "outputs": [],
   "source": [
    "# create instance of subclassed model\n",
    "gan = GAN(pt_generator, ft_discriminator, fine_tuning=True)\n",
    "\n",
    "# compile the model\n",
    "gan.compile(g_opt, d_opt, g_loss, d_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.212222Z",
     "iopub.status.busy": "2023-01-28T22:44:42.212052Z",
     "iopub.status.idle": "2023-01-28T22:44:42.214699Z",
     "shell.execute_reply": "2023-01-28T22:44:42.214226Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.212208Z"
    },
    "id": "t9coR2ScaUpl"
   },
   "outputs": [],
   "source": [
    "# uncomment to remove cache files\n",
    "#remove_file_starting_with('train_cache')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:44:42.215328Z",
     "iopub.status.busy": "2023-01-28T22:44:42.215180Z",
     "iopub.status.idle": "2023-01-28T22:53:50.171399Z",
     "shell.execute_reply": "2023-01-28T22:53:50.170804Z",
     "shell.execute_reply.started": "2023-01-28T22:44:42.215316Z"
    },
    "id": "TPTY-R8VaUpl"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "2652/2652 [==============================] - 41s 14ms/step - generator_loss: 16.0149 - discriminator_loss: -10.3240\n",
      "Epoch 2/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 17.8526 - discriminator_loss: -13.8225\n",
      "Epoch 3/100\n",
      "2652/2652 [==============================] - 39s 15ms/step - generator_loss: 18.0925 - discriminator_loss: -14.0788\n",
      "Epoch 4/100\n",
      "2652/2652 [==============================] - 39s 15ms/step - generator_loss: 18.0552 - discriminator_loss: -14.0610\n",
      "Epoch 5/100\n",
      "2652/2652 [==============================] - 39s 15ms/step - generator_loss: 18.0402 - discriminator_loss: -13.9815\n",
      "Epoch 6/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 18.0326 - discriminator_loss: -13.5985\n",
      "Epoch 7/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 17.0985 - discriminator_loss: -10.5802\n",
      "Epoch 8/100\n",
      "2652/2652 [==============================] - 45s 17ms/step - generator_loss: 16.5495 - discriminator_loss: -9.2100\n",
      "Epoch 9/100\n",
      "2652/2652 [==============================] - 39s 15ms/step - generator_loss: 17.5657 - discriminator_loss: -11.0107\n",
      "Epoch 10/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 17.3996 - discriminator_loss: -12.7852\n",
      "Epoch 11/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 17.1862 - discriminator_loss: -13.5547\n",
      "Epoch 12/100\n",
      "2652/2652 [==============================] - 38s 14ms/step - generator_loss: 17.6047 - discriminator_loss: -14.0604\n",
      "Epoch 13/100\n",
      "2652/2652 [==============================] - 39s 15ms/step - generator_loss: 17.8694 - discriminator_loss: -14.0956\n",
      "Epoch 14/100\n",
      "2652/2652 [==============================] - 39s 15ms/step - generator_loss: 17.8325 - discriminator_loss: -14.0257\n"
     ]
    }
   ],
   "source": [
    "num_epochs = 100\n",
    "\n",
    "#g_early_stopping = EarlyStopping(monitor='generator_loss', patience=10)\n",
    "d_early_stopping = EarlyStopping(monitor='discriminator_loss', patience=10)\n",
    "\n",
    "hist = gan.fit(\n",
    "            train_dataset,\n",
    "            epochs=num_epochs,\n",
    "            steps_per_epoch=train_steps,\n",
    "            callbacks=[d_early_stopping]\n",
    "            )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:50.172366Z",
     "iopub.status.busy": "2023-01-28T22:53:50.172188Z",
     "iopub.status.idle": "2023-01-28T22:53:50.437669Z",
     "shell.execute_reply": "2023-01-28T22:53:50.437049Z",
     "shell.execute_reply.started": "2023-01-28T22:53:50.172351Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "_, ax = plt.subplots(figsize=(8, 5))\n",
    "\n",
    "ax.plot(hist.history['generator_loss'], label='generator', color='blue', linewidth=1.5)\n",
    "ax.plot(hist.history['discriminator_loss'], label='discriminator', color='firebrick', linewidth=1.5)\n",
    "ax.set_xticks(range(len(hist.history['generator_loss'])))\n",
    "ax.set_xticklabels(range(1, len(hist.history['generator_loss'])+1))\n",
    "ax.set_xlabel('epoch')\n",
    "ax.set_ylabel('loss')\n",
    "ax.grid(alpha=0.5)\n",
    "ax.legend(edgecolor='black', facecolor='linen', fontsize=12, loc ='best')\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Uncomment the cell below to save the weights of the trained models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:50.438644Z",
     "iopub.status.busy": "2023-01-28T22:53:50.438463Z",
     "iopub.status.idle": "2023-01-28T22:53:51.567818Z",
     "shell.execute_reply": "2023-01-28T22:53:51.567160Z",
     "shell.execute_reply.started": "2023-01-28T22:53:50.438627Z"
    },
    "id": "GwMvrTJt1J2e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.\n",
      "WARNING:tensorflow:Compiled the loaded model, but the compiled metrics have yet to be built. `model.compile_metrics` will be empty until you train or evaluate the model.\n"
     ]
    }
   ],
   "source": [
    "pt_generator.save('ft_generator.h5')\n",
    "ft_discriminator.save('ft_discriminator.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Uncomment the cell below to load pretrained models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:51.568958Z",
     "iopub.status.busy": "2023-01-28T22:53:51.568739Z",
     "iopub.status.idle": "2023-01-28T22:53:51.571833Z",
     "shell.execute_reply": "2023-01-28T22:53:51.571311Z",
     "shell.execute_reply.started": "2023-01-28T22:53:51.568941Z"
    }
   },
   "outputs": [],
   "source": [
    "#pt_generator     = tf.keras.models.load_model('ft_generator.h5')\n",
    "#ft_discriminator = tf.keras.models.load_model('ft_discriminator.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:51.572644Z",
     "iopub.status.busy": "2023-01-28T22:53:51.572479Z",
     "iopub.status.idle": "2023-01-28T22:53:51.575562Z",
     "shell.execute_reply": "2023-01-28T22:53:51.575115Z",
     "shell.execute_reply.started": "2023-01-28T22:53:51.572630Z"
    },
    "id": "e-Fw7BRPXr9p"
   },
   "outputs": [],
   "source": [
    "model = ft_discriminator"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-khCC16dgu2K"
   },
   "source": [
    "## 4 - Model testing"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "id": "GqXQhhpvgu5P"
   },
   "source": [
    "### 4.2 - Validation metrics\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:51.576371Z",
     "iopub.status.busy": "2023-01-28T22:53:51.576201Z",
     "iopub.status.idle": "2023-01-28T22:53:52.347256Z",
     "shell.execute_reply": "2023-01-28T22:53:52.346516Z",
     "shell.execute_reply.started": "2023-01-28T22:53:51.576357Z"
    },
    "id": "d9qSfMZlgu5U"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "344/344 [==============================] - 1s 2ms/step\n"
     ]
    }
   ],
   "source": [
    "# true labels\n",
    "y_true = reference_df_test['class'].tolist()\n",
    "\n",
    "# get predictions for test set\n",
    "test_pred = model.predict(test_dataset, steps=test_steps, verbose=1)[:len(y_true)].squeeze()\n",
    "\n",
    "# predicted labels\n",
    "y_pred = np.argmax(test_pred, axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jinyZOKNgu7V"
   },
   "source": [
    "Then, to visualize the performance of the algorithm we plot the confusion matrix:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:52.348594Z",
     "iopub.status.busy": "2023-01-28T22:53:52.348397Z",
     "iopub.status.idle": "2023-01-28T22:53:53.120211Z",
     "shell.execute_reply": "2023-01-28T22:53:53.119609Z",
     "shell.execute_reply.started": "2023-01-28T22:53:52.348577Z"
    },
    "id": "cHIhE3c3gu7e"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/notebooks/plot_utils.py:56: RuntimeWarning: invalid value encountered in divide\n",
      "  cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1512x1296 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cm = confusion_matrix(y_pred, y_true)\n",
    "plot_utils.plot_confusion_matrix(cm, labels=commands+['FAKE'], normalize=True, saveit=False, model_name='')\n",
    "# plt.savefig(f'figures/cm_{model_name}.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:53.121456Z",
     "iopub.status.busy": "2023-01-28T22:53:53.121280Z",
     "iopub.status.idle": "2023-01-28T22:53:53.158482Z",
     "shell.execute_reply": "2023-01-28T22:53:53.157999Z",
     "shell.execute_reply.started": "2023-01-28T22:53:53.121439Z"
    },
    "id": "NPE9Z35Xgu70"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy:      2.67%\n",
      "Error rate:    97.33%\n",
      "Cross-entropy: 3.7170\n",
      "Cohen's Kappa: -0.0010\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.9/dist-packages/sklearn/metrics/_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.9/dist-packages/sklearn/metrics/_classification.py:1327: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.9/dist-packages/sklearn/metrics/_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.9/dist-packages/sklearn/metrics/_classification.py:1327: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "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>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>macro</th>\n",
       "      <td>0.0053</td>\n",
       "      <td>0.0200</td>\n",
       "      <td>0.0072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>micro</th>\n",
       "      <td>0.0267</td>\n",
       "      <td>0.0267</td>\n",
       "      <td>0.0267</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>weighted</th>\n",
       "      <td>0.0072</td>\n",
       "      <td>0.0267</td>\n",
       "      <td>0.0097</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          precision  recall  f1-score\n",
       "macro        0.0053  0.0200    0.0072\n",
       "micro        0.0267  0.0267    0.0267\n",
       "weighted     0.0072  0.0267    0.0097"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "accuracy = accuracy_score(y_true=y_true, y_pred=y_pred)\n",
    "logloss  = log_loss(y_true, test_pred, labels=np.arange(36))\n",
    "cohen_k  = cohen_kappa_score(y_true, y_pred)\n",
    "\n",
    "print('Accuracy:      {:.2f}%'.format(accuracy*100))\n",
    "print('Error rate:    {:.2f}%'.format((1-accuracy)*100))\n",
    "print('Cross-entropy: {:.4f}'.format(logloss))\n",
    "print(\"Cohen's Kappa: {:.4f}\".format(cohen_k))\n",
    "\n",
    "def class_report(y_true, y_pred):\n",
    "    precision_macro,    recall_macro,    fscore_macro,    _ = precision_recall_fscore_support(y_true=y_true, y_pred=y_pred, average='macro')\n",
    "    precision_micro,    recall_micro,    fscore_micro,    _ = precision_recall_fscore_support(y_true=y_true, y_pred=y_pred, average='micro')\n",
    "    precision_weighted, recall_weighted, fscore_weighted, _ = precision_recall_fscore_support(y_true=y_true, y_pred=y_pred, average='weighted')\n",
    "    \n",
    "    df_report = pd.DataFrame(\n",
    "        {\n",
    "            'precision': [precision_macro, precision_micro, precision_weighted],\n",
    "            'recall': [recall_macro, recall_micro, recall_weighted],\n",
    "            'f1-score': [fscore_macro, fscore_micro, fscore_weighted]\n",
    "        }, \n",
    "        index=['macro', 'micro', 'weighted']\n",
    "    )\n",
    "    \n",
    "    return df_report\n",
    "\n",
    "cr = class_report(y_true, y_pred)\n",
    "display(cr.round(4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "execution": {
     "iopub.execute_input": "2023-01-28T22:53:53.159649Z",
     "iopub.status.busy": "2023-01-28T22:53:53.159476Z",
     "iopub.status.idle": "2023-01-28T22:53:53.168152Z",
     "shell.execute_reply": "2023-01-28T22:53:53.167389Z",
     "shell.execute_reply.started": "2023-01-28T22:53:53.159634Z"
    },
    "id": "SaXtDPNwgu73"
   },
   "outputs": [
    {
     "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>precision</th>\n",
       "      <th>recall</th>\n",
       "      <th>f1-score</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>cross-entropy</th>\n",
       "      <th>Cohen-Kappa</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0072</td>\n",
       "      <td>0.0267</td>\n",
       "      <td>0.0097</td>\n",
       "      <td>0.0267</td>\n",
       "      <td>3.717</td>\n",
       "      <td>-0.001</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   precision  recall  f1-score  accuracy  cross-entropy  Cohen-Kappa\n",
       "0     0.0072  0.0267    0.0097    0.0267          3.717       -0.001"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "weighted_metrics = cr.iloc[-1].tolist()\n",
    "\n",
    "df_metrics = pd.DataFrame(\n",
    "    [weighted_metrics + [accuracy, logloss, cohen_k]], \n",
    "    columns = ['precision', 'recall', 'f1-score', 'accuracy', 'cross-entropy', 'Cohen-Kappa'], \n",
    ")\n",
    "\n",
    "display(df_metrics.round(4))\n",
    "\n",
    "# Uncomment next line to store the final metrics \n",
    "# df_metrics.to_csv(f'metrics/metr_{model_name}.csv')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {
    "id": "aqEwef54gu74"
   },
   "source": [
    "# References\n",
    "**[Sainath15]** <br>\n",
    " Tara N. Sainath, Carolina Parada, _Convolutional Neural Networks for Small-footprint Keyword Spotting_, INTERSPEECH, Dresden, Germany, September 2015.\n",
    "\n",
    "**[Warden18]** <br>\n",
    " Pete Warden, _Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition_, arXiv:1804.03209, April 2018. \n",
    " \n",
    "**[Lyons14]** <br>\n",
    " James Lyons et al. (2020, January 14). jameslyons/python_speech_features: release v0.6.1 (Version 0.6.1). Zenodo.\n",
    "\n",
    "**[Eskimez20]** <br>\n",
    "Eskimez, Sefik & Dimitriadis, Dimitrios & Gmyr, Robert & Kumanati, Kenichi. (2020). GAN-Based Data Generation for Speech Emotion Recognition. 3446-3450. 10.21437/Interspeech.2020-2898.\n"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "provenance": []
  },
  "gpuClass": "premium",
  "kernelspec": {
   "display_name": "hda",
   "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": "3.10.8"
  },
  "vscode": {
   "interpreter": {
    "hash": "2dbd5f52e176f26a59a7f95f14a81615dde9e9336e1642592a7a0db563ccdd35"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}