SnapSense.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Setup and Get Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.1 Install Dependencies and Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: labelme in c:\\users\\computer\\anaconda3\\lib\\site-packages (5.4.1)\n",
      "Requirement already satisfied: tensorflow in c:\\users\\computer\\anaconda3\\lib\\site-packages (2.15.0)\n",
      "Collecting tensorflow-gpu\n",
      "  Using cached tensorflow-gpu-2.12.0.tar.gz (2.6 kB)\n",
      "  Preparing metadata (setup.py): started\n",
      "  Preparing metadata (setup.py): finished with status 'done'\n",
      "Requirement already satisfied: opencv-python in c:\\users\\computer\\anaconda3\\lib\\site-packages (4.9.0.80)\n",
      "Requirement already satisfied: matplotlib in c:\\users\\computer\\anaconda3\\lib\\site-packages (3.5.2)\n",
      "Requirement already satisfied: albumentations in c:\\users\\computer\\anaconda3\\lib\\site-packages (1.3.1)\n",
      "Requirement already satisfied: Pillow>=2.8 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (9.2.0)\n",
      "Requirement already satisfied: termcolor in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (2.4.0)\n",
      "Requirement already satisfied: onnxruntime!=1.16.0,>=1.14.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (1.16.3)\n",
      "Requirement already satisfied: qtpy!=1.11.2 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (2.2.0)\n",
      "Requirement already satisfied: gdown in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (5.0.1)\n",
      "Requirement already satisfied: colorama in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (0.4.5)\n",
      "Requirement already satisfied: numpy in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (1.24.4)\n",
      "Requirement already satisfied: natsort>=7.1.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (8.4.0)\n",
      "Requirement already satisfied: PyYAML in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (6.0)\n",
      "Requirement already satisfied: scikit-image in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (0.19.2)\n",
      "Requirement already satisfied: imgviz>=1.7.5 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from labelme) (1.7.5)\n",
      "Requirement already satisfied: tensorflow-intel==2.15.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow) (2.15.0)\n",
      "Requirement already satisfied: setuptools in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (63.4.1)\n",
      "Requirement already satisfied: tensorflow-io-gcs-filesystem>=0.23.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (0.31.0)\n",
      "Requirement already satisfied: typing-extensions>=3.6.6 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (4.11.0)\n",
      "Requirement already satisfied: grpcio<2.0,>=1.24.3 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (1.60.0)\n",
      "Requirement already satisfied: protobuf!=4.21.0,!=4.21.1,!=4.21.2,!=4.21.3,!=4.21.4,!=4.21.5,<5.0.0dev,>=3.20.3 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (4.23.4)\n",
      "Requirement already satisfied: keras<2.16,>=2.15.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (2.15.0)\n",
      "Requirement already satisfied: tensorboard<2.16,>=2.15 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (2.15.1)\n",
      "Requirement already satisfied: google-pasta>=0.1.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (0.2.0)\n",
      "Requirement already satisfied: six>=1.12.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (1.16.0)\n",
      "Requirement already satisfied: packaging in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (21.3)\n",
      "Requirement already satisfied: gast!=0.5.0,!=0.5.1,!=0.5.2,>=0.2.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (0.5.4)\n",
      "Requirement already satisfied: libclang>=13.0.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (16.0.6)\n",
      "Requirement already satisfied: opt-einsum>=2.3.2 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (3.3.0)\n",
      "Requirement already satisfied: flatbuffers>=23.5.26 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (23.5.26)\n",
      "Requirement already satisfied: ml-dtypes~=0.2.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (0.2.0)\n",
      "Requirement already satisfied: h5py>=2.9.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (3.7.0)\n",
      "Requirement already satisfied: absl-py>=1.0.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (2.1.0)\n",
      "Requirement already satisfied: wrapt<1.15,>=1.11.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (1.14.1)\n",
      "Requirement already satisfied: astunparse>=1.6.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (1.6.3)\n",
      "Requirement already satisfied: tensorflow-estimator<2.16,>=2.15.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-intel==2.15.0->tensorflow) (2.15.0)\n",
      "Requirement already satisfied: python_version>\"3.7\" in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorflow-gpu) (0.0.2)\n",
      "Requirement already satisfied: fonttools>=4.22.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (4.25.0)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (1.4.2)\n",
      "Requirement already satisfied: cycler>=0.10 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (0.11.0)\n",
      "Requirement already satisfied: python-dateutil>=2.7 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (2.8.2)\n",
      "Requirement already satisfied: pyparsing>=2.2.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (3.0.9)\n",
      "Requirement already satisfied: qudida>=0.0.4 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (0.0.4)\n",
      "Requirement already satisfied: scipy>=1.1.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (1.9.1)\n",
      "Requirement already satisfied: opencv-python-headless>=4.1.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (4.9.0.80)\n",
      "Requirement already satisfied: coloredlogs in c:\\users\\computer\\anaconda3\\lib\\site-packages (from onnxruntime!=1.16.0,>=1.14.1->labelme) (15.0.1)\n",
      "Requirement already satisfied: sympy in c:\\users\\computer\\anaconda3\\lib\\site-packages (from onnxruntime!=1.16.0,>=1.14.1->labelme) (1.10.1)\n",
      "Requirement already satisfied: scikit-learn>=0.19.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from qudida>=0.0.4->albumentations) (1.0.2)\n",
      "Requirement already satisfied: imageio>=2.4.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image->labelme) (2.19.3)\n",
      "Requirement already satisfied: PyWavelets>=1.1.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image->labelme) (1.3.0)\n",
      "Requirement already satisfied: tifffile>=2019.7.26 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image->labelme) (2021.7.2)\n",
      "Requirement already satisfied: networkx>=2.2 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image->labelme) (2.8.4)\n",
      "Requirement already satisfied: tqdm in c:\\users\\computer\\anaconda3\\lib\\site-packages (from gdown->labelme) (4.64.1)\n",
      "Requirement already satisfied: beautifulsoup4 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from gdown->labelme) (4.11.1)\n",
      "Requirement already satisfied: filelock in c:\\users\\computer\\anaconda3\\lib\\site-packages (from gdown->labelme) (3.6.0)\n",
      "Requirement already satisfied: requests[socks] in c:\\users\\computer\\anaconda3\\lib\\site-packages (from gdown->labelme) (2.28.1)\n",
      "Requirement already satisfied: wheel<1.0,>=0.23.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from astunparse>=1.6.0->tensorflow-intel==2.15.0->tensorflow) (0.37.1)\n",
      "Requirement already satisfied: joblib>=0.11 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-learn>=0.19.1->qudida>=0.0.4->albumentations) (1.1.0)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-learn>=0.19.1->qudida>=0.0.4->albumentations) (2.2.0)\n",
      "Requirement already satisfied: markdown>=2.6.8 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (3.3.4)\n",
      "Requirement already satisfied: tensorboard-data-server<0.8.0,>=0.7.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (0.7.2)\n",
      "Requirement already satisfied: google-auth-oauthlib<2,>=0.5 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (1.2.0)\n",
      "Requirement already satisfied: werkzeug>=1.0.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (2.0.3)\n",
      "Requirement already satisfied: google-auth<3,>=1.6.3 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (2.27.0)\n",
      "Requirement already satisfied: soupsieve>1.2 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from beautifulsoup4->gdown->labelme) (2.3.1)\n",
      "Requirement already satisfied: humanfriendly>=9.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from coloredlogs->onnxruntime!=1.16.0,>=1.14.1->labelme) (10.0)\n",
      "Requirement already satisfied: charset-normalizer<3,>=2 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from requests[socks]->gdown->labelme) (2.0.4)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from requests[socks]->gdown->labelme) (2022.9.14)\n",
      "Requirement already satisfied: idna<4,>=2.5 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from requests[socks]->gdown->labelme) (3.3)\n",
      "Requirement already satisfied: urllib3<1.27,>=1.21.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from requests[socks]->gdown->labelme) (1.26.11)\n",
      "Requirement already satisfied: PySocks!=1.5.7,>=1.5.6 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from requests[socks]->gdown->labelme) (1.7.1)\n",
      "Requirement already satisfied: mpmath>=0.19 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from sympy->onnxruntime!=1.16.0,>=1.14.1->labelme) (1.2.1)\n",
      "Requirement already satisfied: rsa<5,>=3.1.4 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from google-auth<3,>=1.6.3->tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (4.9)\n",
      "Requirement already satisfied: pyasn1-modules>=0.2.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from google-auth<3,>=1.6.3->tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (0.2.8)\n",
      "Requirement already satisfied: cachetools<6.0,>=2.0.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from google-auth<3,>=1.6.3->tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (5.3.2)\n",
      "Requirement already satisfied: requests-oauthlib>=0.7.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from google-auth-oauthlib<2,>=0.5->tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (1.3.1)\n",
      "Requirement already satisfied: pyreadline3 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from humanfriendly>=9.1->coloredlogs->onnxruntime!=1.16.0,>=1.14.1->labelme) (3.4.1)\n",
      "Requirement already satisfied: pyasn1<0.5.0,>=0.4.6 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from pyasn1-modules>=0.2.1->google-auth<3,>=1.6.3->tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (0.4.8)\n",
      "Requirement already satisfied: oauthlib>=3.0.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<2,>=0.5->tensorboard<2.16,>=2.15->tensorflow-intel==2.15.0->tensorflow) (3.2.2)\n",
      "Building wheels for collected packages: tensorflow-gpu\n",
      "  Building wheel for tensorflow-gpu (setup.py): started\n",
      "  Building wheel for tensorflow-gpu (setup.py): finished with status 'error'\n",
      "  Running setup.py clean for tensorflow-gpu\n",
      "Failed to build tensorflow-gpu\n",
      "Installing collected packages: tensorflow-gpu\n",
      "  Running setup.py install for tensorflow-gpu: started\n",
      "  Running setup.py install for tensorflow-gpu: finished with status 'error'\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  error: subprocess-exited-with-error\n",
      "  \n",
      "  python setup.py bdist_wheel did not run successfully.\n",
      "  exit code: 1\n",
      "  \n",
      "  [18 lines of output]\n",
      "  Traceback (most recent call last):\n",
      "    File \"<string>\", line 2, in <module>\n",
      "    File \"<pip-setuptools-caller>\", line 34, in <module>\n",
      "    File \"C:\\Users\\Computer\\AppData\\Local\\Temp\\pip-install-d7om_717\\tensorflow-gpu_4de45126a0f044e884edde6f010e33a7\\setup.py\", line 37, in <module>\n",
      "      raise Exception(TF_REMOVAL_WARNING)\n",
      "  Exception:\n",
      "  \n",
      "  =========================================================\n",
      "  The \"tensorflow-gpu\" package has been removed!\n",
      "  \n",
      "  Please install \"tensorflow\" instead.\n",
      "  \n",
      "  Other than the name, the two packages have been identical\n",
      "  since TensorFlow 2.1, or roughly since Sep 2019. For more\n",
      "  information, see: pypi.org/project/tensorflow-gpu\n",
      "  =========================================================\n",
      "  \n",
      "  \n",
      "  [end of output]\n",
      "  \n",
      "  note: This error originates from a subprocess, and is likely not a problem with pip.\n",
      "  ERROR: Failed building wheel for tensorflow-gpu\n",
      "  error: subprocess-exited-with-error\n",
      "  \n",
      "  Running setup.py install for tensorflow-gpu did not run successfully.\n",
      "  exit code: 1\n",
      "  \n",
      "  [18 lines of output]\n",
      "  Traceback (most recent call last):\n",
      "    File \"<string>\", line 2, in <module>\n",
      "    File \"<pip-setuptools-caller>\", line 34, in <module>\n",
      "    File \"C:\\Users\\Computer\\AppData\\Local\\Temp\\pip-install-d7om_717\\tensorflow-gpu_4de45126a0f044e884edde6f010e33a7\\setup.py\", line 37, in <module>\n",
      "      raise Exception(TF_REMOVAL_WARNING)\n",
      "  Exception:\n",
      "  \n",
      "  =========================================================\n",
      "  The \"tensorflow-gpu\" package has been removed!\n",
      "  \n",
      "  Please install \"tensorflow\" instead.\n",
      "  \n",
      "  Other than the name, the two packages have been identical\n",
      "  since TensorFlow 2.1, or roughly since Sep 2019. For more\n",
      "  information, see: pypi.org/project/tensorflow-gpu\n",
      "  =========================================================\n",
      "  \n",
      "  \n",
      "  [end of output]\n",
      "  \n",
      "  note: This error originates from a subprocess, and is likely not a problem with pip.\n",
      "error: legacy-install-failure\n",
      "\n",
      "Encountered error while trying to install package.\n",
      "\n",
      "tensorflow-gpu\n",
      "\n",
      "note: This is an issue with the package mentioned above, not pip.\n",
      "hint: See above for output from the failure.\n"
     ]
    }
   ],
   "source": [
    "!pip install labelme tensorflow tensorflow-gpu opencv-python matplotlib albumentations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: opencv-python in c:\\users\\computer\\anaconda3\\lib\\site-packages (4.9.0.80)\n",
      "Requirement already satisfied: numpy>=1.17.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from opencv-python) (1.24.4)\n"
     ]
    }
   ],
   "source": [
    "!pip install  opencv-python"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: matplotlib in c:\\users\\computer\\anaconda3\\lib\\site-packages (3.5.2)\n",
      "Requirement already satisfied: albumentations in c:\\users\\computer\\anaconda3\\lib\\site-packages (1.3.1)\n",
      "Requirement already satisfied: numpy>=1.17 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (1.24.4)\n",
      "Requirement already satisfied: cycler>=0.10 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (0.11.0)\n",
      "Requirement already satisfied: pyparsing>=2.2.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (3.0.9)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (1.4.2)\n",
      "Requirement already satisfied: fonttools>=4.22.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (4.25.0)\n",
      "Requirement already satisfied: python-dateutil>=2.7 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (2.8.2)\n",
      "Requirement already satisfied: pillow>=6.2.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (9.2.0)\n",
      "Requirement already satisfied: packaging>=20.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from matplotlib) (21.3)\n",
      "Requirement already satisfied: scikit-image>=0.16.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (0.19.2)\n",
      "Requirement already satisfied: qudida>=0.0.4 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (0.0.4)\n",
      "Requirement already satisfied: PyYAML in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (6.0)\n",
      "Requirement already satisfied: scipy>=1.1.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (1.9.1)\n",
      "Requirement already satisfied: opencv-python-headless>=4.1.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from albumentations) (4.9.0.80)\n",
      "Requirement already satisfied: six>=1.5 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from python-dateutil>=2.7->matplotlib) (1.16.0)\n",
      "Requirement already satisfied: scikit-learn>=0.19.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from qudida>=0.0.4->albumentations) (1.0.2)\n",
      "Requirement already satisfied: typing-extensions in c:\\users\\computer\\anaconda3\\lib\\site-packages (from qudida>=0.0.4->albumentations) (4.11.0)\n",
      "Requirement already satisfied: networkx>=2.2 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image>=0.16.1->albumentations) (2.8.4)\n",
      "Requirement already satisfied: imageio>=2.4.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image>=0.16.1->albumentations) (2.19.3)\n",
      "Requirement already satisfied: tifffile>=2019.7.26 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image>=0.16.1->albumentations) (2021.7.2)\n",
      "Requirement already satisfied: PyWavelets>=1.1.1 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-image>=0.16.1->albumentations) (1.3.0)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-learn>=0.19.1->qudida>=0.0.4->albumentations) (2.2.0)\n",
      "Requirement already satisfied: joblib>=0.11 in c:\\users\\computer\\anaconda3\\lib\\site-packages (from scikit-learn>=0.19.1->qudida>=0.0.4->albumentations) (1.1.0)\n"
     ]
    }
   ],
   "source": [
    "!pip install  matplotlib albumentations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 Collect Images Using OpenCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import time\n",
    "import uuid\n",
    "import cv2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "IMAGES_PATH = os.path.join('data','images')\n",
    "number_images = 30"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting image 0\n",
      "Collecting image 1\n",
      "Collecting image 2\n",
      "Collecting image 3\n",
      "Collecting image 4\n",
      "Collecting image 5\n",
      "Collecting image 6\n",
      "Collecting image 7\n",
      "Collecting image 8\n",
      "Collecting image 9\n",
      "Collecting image 10\n",
      "Collecting image 11\n",
      "Collecting image 12\n",
      "Collecting image 13\n",
      "Collecting image 14\n",
      "Collecting image 15\n",
      "Collecting image 16\n",
      "Collecting image 17\n",
      "Collecting image 18\n",
      "Collecting image 19\n",
      "Collecting image 20\n",
      "Collecting image 21\n",
      "Collecting image 22\n",
      "Collecting image 23\n",
      "Collecting image 24\n",
      "Collecting image 25\n",
      "Collecting image 26\n",
      "Collecting image 27\n",
      "Collecting image 28\n",
      "Collecting image 29\n",
      "Collecting image 30\n",
      "Collecting image 31\n",
      "Collecting image 32\n",
      "Collecting image 33\n",
      "Collecting image 34\n",
      "Collecting image 35\n",
      "Collecting image 36\n",
      "Collecting image 37\n",
      "Collecting image 38\n",
      "Collecting image 39\n",
      "Collecting image 40\n",
      "Collecting image 41\n",
      "Collecting image 42\n",
      "Collecting image 43\n",
      "Collecting image 44\n",
      "Collecting image 45\n",
      "Collecting image 46\n",
      "Collecting image 47\n",
      "Collecting image 48\n",
      "Collecting image 49\n",
      "Collecting image 50\n",
      "Collecting image 51\n",
      "Collecting image 52\n",
      "Collecting image 53\n",
      "Collecting image 54\n",
      "Collecting image 55\n",
      "Collecting image 56\n",
      "Collecting image 57\n",
      "Collecting image 58\n",
      "Collecting image 59\n"
     ]
    }
   ],
   "source": [
    "import cv2\n",
    "import os\n",
    "import uuid\n",
    "import time\n",
    "\n",
    "# Set the number of frames to capture\n",
    "number_images = 60\n",
    "\n",
    "# Create a directory to store the images\n",
    "IMAGES_PATH = \"images\"\n",
    "os.makedirs(IMAGES_PATH, exist_ok=True)\n",
    "\n",
    "# Initialize video capture\n",
    "cap = cv2.VideoCapture(0)  # Use 0 instead of 1 for the default camera\n",
    "\n",
    "for imgnum in range(number_images):\n",
    "    print('Collecting image {}'.format(imgnum))\n",
    "\n",
    "    # Capture frame\n",
    "    ret, frame = cap.read()\n",
    "\n",
    "    # Check if frame capture was successful\n",
    "    if ret:\n",
    "        imgname = os.path.join(IMAGES_PATH, f'{str(uuid.uuid1())}.jpg')\n",
    "        cv2.imwrite(imgname, frame)\n",
    "        cv2.imshow('frame', frame)\n",
    "        cv2.waitKey(1)  # Display the frame for 1 millisecond\n",
    "        time.sleep(0.5)  # Adjust delay as needed\n",
    "    else:\n",
    "        print(f'Failed to capture frame {imgnum}. Error code: {ret}')\n",
    "\n",
    "# Release resources\n",
    "cap.release()\n",
    "cv2.destroyAllWindows()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.3 Annotate Images with LabelMe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-23 13:15:44,397 [INFO   ] __init__:get_config:67- Loading config file from: C:\\Users\\Computer\\.labelmerc\n"
     ]
    }
   ],
   "source": [
    "!labelme"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Review Dataset and Build Image Loading Function"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.1 Import TF and Deps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\losses.py:2976: The name tf.losses.sparse_softmax_cross_entropy is deprecated. Please use tf.compat.v1.losses.sparse_softmax_cross_entropy instead.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import json\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.2 Limit GPU Memory Growth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Avoid OOM errors by setting GPU Memory Consumption Growth\n",
    "gpus = tf.config.experimental.list_physical_devices('GPU')\n",
    "for gpu in gpus: \n",
    "    tf.config.experimental.set_memory_growth(gpu, True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.config.list_physical_devices('GPU')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.3 Load Image into TF Data Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "images = tf.data.Dataset.list_files('data\\\\images\\\\*.jpg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "images.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def load_image(x): \n",
    "    byte_img = tf.io.read_file(x)\n",
    "    img = tf.io.decode_jpeg(byte_img)\n",
    "    return img"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "images = images.map(load_image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "images.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "type(images)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2.4 View Raw Images with Matplotlib"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "image_generator = images.batch(4).as_numpy_iterator()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "plot_images = image_generator.next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "fig, ax = plt.subplots(ncols=4, figsize=(20,20))\n",
    "for idx, image in enumerate(plot_images):\n",
    "    ax[idx].imshow(image) \n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3. Partition Unaugmented Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.1 MANUALLY SPLT DATA INTO TRAIN TEST AND VAL"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "125.99999999999999"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "180*.7 # 125 to train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "27.0"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "180*.15 # 27 and 28 to test and val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "180"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "125+55"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3.2 Move the Matching Labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "for folder in ['train','test','val']:\n",
    "    for file in os.listdir(os.path.join('data', folder, 'images')):\n",
    "        \n",
    "        filename = file.split('.')[0]+'.json'\n",
    "        existing_filepath = os.path.join('data','labels', filename)\n",
    "        if os.path.exists(existing_filepath): \n",
    "            new_filepath = os.path.join('data',folder,'labels',filename)\n",
    "            os.replace(existing_filepath, new_filepath)      "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 4. Apply Image Augmentation on Images and Labels using Albumentations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.1 Setup Albumentations Transform Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import albumentations as alb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "augmentor = alb.Compose([alb.RandomCrop(width=450, height=450), \n",
    "                         alb.HorizontalFlip(p=0.5), \n",
    "                         alb.RandomBrightnessContrast(p=0.2),\n",
    "                         alb.RandomGamma(p=0.2), \n",
    "                         alb.RGBShift(p=0.2), \n",
    "                         alb.VerticalFlip(p=0.5)], \n",
    "                       bbox_params=alb.BboxParams(format='albumentations', \n",
    "                                                  label_fields=['class_labels']))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.2 Load a Test Image and Annotation with OpenCV and JSON"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "img = cv2.imread(os.path.join('data','train', 'images','8320b4db-00d3-11ef-8754-0028f88277aa.jpg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(480, 640, 3)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "img.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "with open(os.path.join('data', 'train', 'labels', '8320b4db-00d3-11ef-8754-0028f88277aa.json'), 'r') as f:\n",
    "    label = json.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[[219.73958333333337, 60.31249999999998], [516.09375, 408.22916666666663]]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "label['shapes'][0]['points']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.3 Extract Coordinates and Rescale to Match Image Resolution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "coords = [0,0,0,0]\n",
    "coords[0] = label['shapes'][0]['points'][0][0]\n",
    "coords[1] = label['shapes'][0]['points'][0][1]\n",
    "coords[2] = label['shapes'][0]['points'][1][0]\n",
    "coords[3] = label['shapes'][0]['points'][1][1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[219.73958333333337, 60.31249999999998, 516.09375, 408.22916666666663]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coords"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "coords = list(np.divide(coords, [640,480,640,480]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.34334309895833337, 0.12565104166666663, 0.806396484375, 0.8504774305555555]"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coords"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4.4 Apply Augmentations and View Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "augmented = augmentor(image=img, bboxes=[coords], class_labels=['face'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.785023148148148, 0.8904166666666667)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "augmented['bboxes'][0][2:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(0.12645833333333334,\n",
       "  0.1172685185185186,\n",
       "  0.785023148148148,\n",
       "  0.8904166666666667)]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "augmented['bboxes']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "cv2.rectangle(augmented['image'], \n",
    "              tuple(np.multiply(augmented['bboxes'][0][:2], [450,450]).astype(int)),\n",
    "              tuple(np.multiply(augmented['bboxes'][0][2:], [450,450]).astype(int)), \n",
    "                    (255,0,0), 2)\n",
    "\n",
    "plt.imshow(augmented['image'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 5. Build and Run Augmentation Pipeline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.1 Run Augmentation Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "image must be numpy array type\n",
      "x_max is less than or equal to x_min for bbox [0.3758951822916667, 0.802734375, 0.011311848958333325, 0.1104600694444444, 'face'].\n",
      "x_max is less than or equal to x_min for bbox [0.2635904947916667, 0.6888020833333334, 0.004801432291666652, 0.21028645833333331, 'face'].\n",
      "x_max is less than or equal to x_min for bbox [0.8299967447916667, 0.8949652777777778, 0.2562662760416667, 0.17990451388888884, 'face'].\n",
      "x_max is less than or equal to x_min for bbox [0.908935546875, 0.9752604166666666, 0.31892903645833337, 0.22981770833333331, 'face'].\n",
      "x_max is less than or equal to x_min for bbox [0.28230794270833337, 0.8602430555555555, 0.0, 0.33682232057701467, 'face'].\n",
      "x_max is less than or equal to x_min for bbox [0.899169921875, 0.9437934027777778, 0.266845703125, 0.2276475694444444, 'face'].\n"
     ]
    }
   ],
   "source": [
    "for partition in ['train','test','val']: \n",
    "    for image in os.listdir(os.path.join('data', partition, 'images')):\n",
    "        img = cv2.imread(os.path.join('data', partition, 'images', image))\n",
    "\n",
    "        coords = [0,0,0.00001,0.00001]\n",
    "        label_path = os.path.join('data', partition, 'labels', f'{image.split(\".\")[0]}.json')\n",
    "        if os.path.exists(label_path):\n",
    "            with open(label_path, 'r') as f:\n",
    "                label = json.load(f)\n",
    "\n",
    "            coords[0] = label['shapes'][0]['points'][0][0]\n",
    "            coords[1] = label['shapes'][0]['points'][0][1]\n",
    "            coords[2] = label['shapes'][0]['points'][1][0]\n",
    "            coords[3] = label['shapes'][0]['points'][1][1]\n",
    "            coords = list(np.divide(coords, [640,480,640,480]))\n",
    "\n",
    "        try: \n",
    "            for x in range(60):\n",
    "                augmented = augmentor(image=img, bboxes=[coords], class_labels=['face'])\n",
    "                cv2.imwrite(os.path.join('aug_data', partition, 'images', f'{image.split(\".\")[0]}.{x}.jpg'), augmented['image'])\n",
    "\n",
    "                annotation = {}\n",
    "                annotation['image'] = image\n",
    "\n",
    "                if os.path.exists(label_path):\n",
    "                    if len(augmented['bboxes']) == 0: \n",
    "                        annotation['bbox'] = [0,0,0,0]\n",
    "                        annotation['class'] = 0 \n",
    "                    else: \n",
    "                        annotation['bbox'] = augmented['bboxes'][0]\n",
    "                        annotation['class'] = 1\n",
    "                else: \n",
    "                    annotation['bbox'] = [0,0,0,0]\n",
    "                    annotation['class'] = 0 \n",
    "\n",
    "\n",
    "                with open(os.path.join('aug_data', partition, 'labels', f'{image.split(\".\")[0]}.{x}.json'), 'w') as f:\n",
    "                    json.dump(annotation, f)\n",
    "\n",
    "        except Exception as e:\n",
    "            print(e)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5.2 Load Augmented Images to Tensorflow Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "train_images = tf.data.Dataset.list_files('aug_data\\\\train\\\\images\\\\*.jpg', shuffle=False)\n",
    "train_images = train_images.map(load_image)\n",
    "train_images = train_images.map(lambda x: tf.image.resize(x, (120,120)))\n",
    "train_images = train_images.map(lambda x: x/255)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "test_images = tf.data.Dataset.list_files('aug_data\\\\test\\\\images\\\\*.jpg', shuffle=False)\n",
    "test_images = test_images.map(load_image)\n",
    "test_images = test_images.map(lambda x: tf.image.resize(x, (120,120)))\n",
    "test_images = test_images.map(lambda x: x/255)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "val_images = tf.data.Dataset.list_files('aug_data\\\\val\\\\images\\\\*.jpg', shuffle=False)\n",
    "val_images = val_images.map(load_image)\n",
    "val_images = val_images.map(lambda x: tf.image.resize(x, (120,120)))\n",
    "val_images = val_images.map(lambda x: x/255)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[[0.24136029, 0.16691177, 0.13008578],\n",
       "        [0.27150735, 0.18345588, 0.16678922],\n",
       "        [0.29099265, 0.19295344, 0.20814951],\n",
       "        ...,\n",
       "        [0.72414213, 0.7751838 , 0.78020835],\n",
       "        [0.71574754, 0.7853554 , 0.79221815],\n",
       "        [0.71066177, 0.78566176, 0.8166054 ]],\n",
       "\n",
       "       [[0.29681373, 0.21280637, 0.14613971],\n",
       "        [0.30661765, 0.2129902 , 0.1747549 ],\n",
       "        [0.30496323, 0.21770833, 0.22457108],\n",
       "        ...,\n",
       "        [0.7500613 , 0.78927696, 0.7936887 ],\n",
       "        [0.7343137 , 0.7868872 , 0.79172796],\n",
       "        [0.7434436 , 0.7996324 , 0.8061887 ]],\n",
       "\n",
       "       [[0.2564951 , 0.21875   , 0.12990196],\n",
       "        [0.297549  , 0.22640932, 0.17254902],\n",
       "        [0.3011642 , 0.2291054 , 0.21390931],\n",
       "        ...,\n",
       "        [0.74313724, 0.80085784, 0.8126838 ],\n",
       "        [0.7632353 , 0.7985294 , 0.82598037],\n",
       "        [0.7384191 , 0.7889093 , 0.79724264]],\n",
       "\n",
       "       ...,\n",
       "\n",
       "       [[0.56985295, 0.6155025 , 0.6819853 ],\n",
       "        [0.59993875, 0.61580884, 0.65974265],\n",
       "        [0.60808825, 0.6403799 , 0.71292895],\n",
       "        ...,\n",
       "        [0.51709557, 0.21268383, 0.12040441],\n",
       "        [0.55539215, 0.24571079, 0.12463235],\n",
       "        [0.5688113 , 0.23192401, 0.12015931]],\n",
       "\n",
       "       [[0.5745711 , 0.60428923, 0.64724267],\n",
       "        [0.58327204, 0.63278186, 0.668076  ],\n",
       "        [0.5912377 , 0.6461397 , 0.6961397 ],\n",
       "        ...,\n",
       "        [0.4735294 , 0.15919118, 0.07352941],\n",
       "        [0.47696078, 0.16813725, 0.0504902 ],\n",
       "        [0.49417892, 0.16531862, 0.05943627]],\n",
       "\n",
       "       [[0.5960172 , 0.62555146, 0.6321691 ],\n",
       "        [0.56764704, 0.6240196 , 0.6387255 ],\n",
       "        [0.57297796, 0.62677693, 0.6408088 ],\n",
       "        ...,\n",
       "        [0.47291666, 0.15251225, 0.08321078],\n",
       "        [0.4466912 , 0.13143383, 0.04289216],\n",
       "        [0.49767157, 0.17506127, 0.09852941]]], dtype=float32)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_images.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 6. Prepare Labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6.1 Build Label Loading Function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def load_labels(label_path):\n",
    "    with open(label_path.numpy(), 'r', encoding = \"utf-8\") as f:\n",
    "        label = json.load(f)\n",
    "        \n",
    "    return [label['class']], label['bbox']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6.2 Load Labels to Tensorflow Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "train_labels = tf.data.Dataset.list_files('aug_data\\\\train\\\\labels\\\\*.json', shuffle=False)\n",
    "train_labels = train_labels.map(lambda x: tf.py_function(load_labels, [x], [tf.uint8, tf.float16]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "test_labels = tf.data.Dataset.list_files('aug_data\\\\test\\\\labels\\\\*.json', shuffle=False)\n",
    "test_labels = test_labels.map(lambda x: tf.py_function(load_labels, [x], [tf.uint8, tf.float16]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "val_labels = tf.data.Dataset.list_files('aug_data\\\\val\\\\labels\\\\*.json', shuffle=False)\n",
    "val_labels = val_labels.map(lambda x: tf.py_function(load_labels, [x], [tf.uint8, tf.float16]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([1], dtype=uint8),\n",
       " array([0.2399, 0.3447, 0.787 , 0.9336], dtype=float16))"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_labels.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 7. Combine Label and Image Samples"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7.1 Check Partition Lengths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(7248, 7140, 1745, 1620, 1680, 1680)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_images), len(train_labels), len(test_images), len(test_labels), len(val_images), len(val_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7.2 Create Final Datasets (Images/Labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "train = tf.data.Dataset.zip((train_images, train_labels))\n",
    "train = train.shuffle(5000)\n",
    "train = train.batch(8)\n",
    "train = train.prefetch(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "test = tf.data.Dataset.zip((test_images, test_labels))\n",
    "test = test.shuffle(1300)\n",
    "test = test.batch(8)\n",
    "test = test.prefetch(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "val = tf.data.Dataset.zip((val_images, val_labels))\n",
    "val = val.shuffle(1000)\n",
    "val = val.batch(8)\n",
    "val = val.prefetch(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(8, 120, 120, 3)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.as_numpy_iterator().next()[0].shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7.3 View Images and Annotations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "data_samples = train.as_numpy_iterator()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.data.ops.dataset_ops.NumpyIterator at 0x209ea71f740>"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "res = data_samples.next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import cv2\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "fig, ax = plt.subplots(ncols=2, figsize=(2,2))\n",
    "for idx in range(2): \n",
    "    sample_image = res[0][idx]\n",
    "    sample_coords = res[1][1][idx]\n",
    "    \n",
    "    sample_image_umat = cv2.UMat(sample_image)  # Convert to UMat\n",
    "    \n",
    "    cv2.rectangle(sample_image_umat, \n",
    "                  tuple(np.multiply(sample_coords[:2], [120,120]).astype(int)),\n",
    "                  tuple(np.multiply(sample_coords[2:], [120,120]).astype(int)), \n",
    "                  (255,0,0), 2)\n",
    "\n",
    "    ax[idx].imshow(sample_image_umat.get())  # Convert back to NumPy array for visualization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 8. Build Deep Learning using the Functional API"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.1 Import Layers and Base Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import Model\n",
    "from tensorflow.keras.layers import Input, Conv2D, Dense, GlobalMaxPooling2D\n",
    "from tensorflow.keras.applications import VGG16"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.2 Download VGG16"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\backend.py:1398: The name tf.executing_eagerly_outside_functions is deprecated. Please use tf.compat.v1.executing_eagerly_outside_functions instead.\n",
      "\n",
      "WARNING:tensorflow:From C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\layers\\pooling\\max_pooling2d.py:161: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "vgg = VGG16(include_top=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"vgg16\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_1 (InputLayer)        [(None, None, None, 3)]   0         \n",
      "                                                                 \n",
      " block1_conv1 (Conv2D)       (None, None, None, 64)    1792      \n",
      "                                                                 \n",
      " block1_conv2 (Conv2D)       (None, None, None, 64)    36928     \n",
      "                                                                 \n",
      " block1_pool (MaxPooling2D)  (None, None, None, 64)    0         \n",
      "                                                                 \n",
      " block2_conv1 (Conv2D)       (None, None, None, 128)   73856     \n",
      "                                                                 \n",
      " block2_conv2 (Conv2D)       (None, None, None, 128)   147584    \n",
      "                                                                 \n",
      " block2_pool (MaxPooling2D)  (None, None, None, 128)   0         \n",
      "                                                                 \n",
      " block3_conv1 (Conv2D)       (None, None, None, 256)   295168    \n",
      "                                                                 \n",
      " block3_conv2 (Conv2D)       (None, None, None, 256)   590080    \n",
      "                                                                 \n",
      " block3_conv3 (Conv2D)       (None, None, None, 256)   590080    \n",
      "                                                                 \n",
      " block3_pool (MaxPooling2D)  (None, None, None, 256)   0         \n",
      "                                                                 \n",
      " block4_conv1 (Conv2D)       (None, None, None, 512)   1180160   \n",
      "                                                                 \n",
      " block4_conv2 (Conv2D)       (None, None, None, 512)   2359808   \n",
      "                                                                 \n",
      " block4_conv3 (Conv2D)       (None, None, None, 512)   2359808   \n",
      "                                                                 \n",
      " block4_pool (MaxPooling2D)  (None, None, None, 512)   0         \n",
      "                                                                 \n",
      " block5_conv1 (Conv2D)       (None, None, None, 512)   2359808   \n",
      "                                                                 \n",
      " block5_conv2 (Conv2D)       (None, None, None, 512)   2359808   \n",
      "                                                                 \n",
      " block5_conv3 (Conv2D)       (None, None, None, 512)   2359808   \n",
      "                                                                 \n",
      " block5_pool (MaxPooling2D)  (None, None, None, 512)   0         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 14714688 (56.13 MB)\n",
      "Trainable params: 14714688 (56.13 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "vgg.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.3 Build instance of Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def build_model(): \n",
    "    input_layer = Input(shape=(120,120,3))\n",
    "    \n",
    "    vgg = VGG16(include_top=False)(input_layer)\n",
    "\n",
    "    # Classification Model  \n",
    "    f1 = GlobalMaxPooling2D()(vgg)\n",
    "    class1 = Dense(2048, activation='relu')(f1)\n",
    "    class2 = Dense(1, activation='sigmoid')(class1)\n",
    "    \n",
    "    # Bounding box model\n",
    "    f2 = GlobalMaxPooling2D()(vgg)\n",
    "    regress1 = Dense(2048, activation='relu')(f2)\n",
    "    regress2 = Dense(4, activation='sigmoid')(regress1)\n",
    "    \n",
    "    facetracker = Model(inputs=input_layer, outputs=[class2, regress2])\n",
    "    return facetracker"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8.4 Test out Neural Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "facetracker = build_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                Output Shape                 Param #   Connected to                  \n",
      "==================================================================================================\n",
      " input_2 (InputLayer)        [(None, 120, 120, 3)]        0         []                            \n",
      "                                                                                                  \n",
      " vgg16 (Functional)          (None, None, None, 512)      1471468   ['input_2[0][0]']             \n",
      "                                                          8                                       \n",
      "                                                                                                  \n",
      " global_max_pooling2d (Glob  (None, 512)                  0         ['vgg16[0][0]']               \n",
      " alMaxPooling2D)                                                                                  \n",
      "                                                                                                  \n",
      " global_max_pooling2d_1 (Gl  (None, 512)                  0         ['vgg16[0][0]']               \n",
      " obalMaxPooling2D)                                                                                \n",
      "                                                                                                  \n",
      " dense (Dense)               (None, 2048)                 1050624   ['global_max_pooling2d[0][0]']\n",
      "                                                                                                  \n",
      " dense_2 (Dense)             (None, 2048)                 1050624   ['global_max_pooling2d_1[0][0]\n",
      "                                                                    ']                            \n",
      "                                                                                                  \n",
      " dense_1 (Dense)             (None, 1)                    2049      ['dense[0][0]']               \n",
      "                                                                                                  \n",
      " dense_3 (Dense)             (None, 4)                    8196      ['dense_2[0][0]']             \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 16826181 (64.19 MB)\n",
      "Trainable params: 16826181 (64.19 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "facetracker.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "X, y = train.as_numpy_iterator().next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(8, 120, 120, 3)"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 1s 1s/step\n"
     ]
    }
   ],
   "source": [
    "classes, coords = facetracker.predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([[0.6230478 ],\n",
       "        [0.5226348 ],\n",
       "        [0.62090576],\n",
       "        [0.5656733 ],\n",
       "        [0.5649991 ],\n",
       "        [0.54261225],\n",
       "        [0.631094  ],\n",
       "        [0.5590043 ]], dtype=float32),\n",
       " array([[0.41650143, 0.6059952 , 0.6639341 , 0.56947696],\n",
       "        [0.48584506, 0.6662046 , 0.5753207 , 0.4930787 ],\n",
       "        [0.4968375 , 0.62846446, 0.68508327, 0.46709794],\n",
       "        [0.4935765 , 0.5994388 , 0.7072449 , 0.61054295],\n",
       "        [0.4043766 , 0.5967258 , 0.5924022 , 0.5302819 ],\n",
       "        [0.46386018, 0.6212354 , 0.7206175 , 0.5196013 ],\n",
       "        [0.42874393, 0.67712736, 0.68027997, 0.5016522 ],\n",
       "        [0.40322375, 0.6950601 , 0.6394331 , 0.5302619 ]], dtype=float32))"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "classes, coords"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 9. Define Losses and Optimizers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.1 Define Optimizer and LR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "893"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "batches_per_epoch = 893\n",
    "lr_decay = (1./0.75 -1)/batches_per_epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0003732736095558043"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr_decay"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "opt = tf.keras.optimizers.legacy.Adam(learning_rate=0.0001, decay=lr_decay)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.2 Create Localization Loss and Classification Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def localization_loss(y_true, yhat):            \n",
    "    delta_coord = tf.reduce_sum(tf.square(y_true[:,:2] - yhat[:,:2]))\n",
    "                  \n",
    "    h_true = y_true[:,3] - y_true[:,1] \n",
    "    w_true = y_true[:,2] - y_true[:,0] \n",
    "\n",
    "    h_pred = yhat[:,3] - yhat[:,1] \n",
    "    w_pred = yhat[:,2] - yhat[:,0] \n",
    "    \n",
    "    delta_size = tf.reduce_sum(tf.square(w_true - w_pred) + tf.square(h_true-h_pred))\n",
    "    \n",
    "    return delta_coord + delta_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "classloss = tf.keras.losses.BinaryCrossentropy()\n",
    "regressloss = localization_loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9.3 Test out Loss Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: shape=(), dtype=float32, numpy=8.539465>"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "localization_loss(y[1], coords)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: shape=(), dtype=float32, numpy=0.54906344>"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "classloss(y[0], classes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<tf.Tensor: shape=(), dtype=float32, numpy=8.539465>"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "regressloss(y[1], coords)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 10. Train Neural Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 10.1 Create Custom Model Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "class FaceTracker(Model): \n",
    "    def __init__(self, eyetracker,  **kwargs): \n",
    "        super().__init__(**kwargs)\n",
    "        self.model = eyetracker\n",
    "\n",
    "    def compile(self, opt, classloss, localizationloss, **kwargs):\n",
    "        super().compile(**kwargs)\n",
    "        self.closs = classloss\n",
    "        self.lloss = localizationloss\n",
    "        self.opt = opt\n",
    "    \n",
    "    def train_step(self, batch, **kwargs): \n",
    "        \n",
    "        X, y = batch\n",
    "        \n",
    "        with tf.GradientTape() as tape: \n",
    "            classes, coords = self.model(X, training=True)\n",
    "            \n",
    "            batch_classloss = self.closs(y[0], classes)\n",
    "            batch_localizationloss = self.lloss(tf.cast(y[1], tf.float32), coords)\n",
    "            \n",
    "            total_loss = batch_localizationloss+0.5*batch_classloss\n",
    "            \n",
    "            grad = tape.gradient(total_loss, self.model.trainable_variables)\n",
    "        \n",
    "        opt.apply_gradients(zip(grad, self.model.trainable_variables))\n",
    "        \n",
    "        return {\"total_loss\":total_loss, \"class_loss\":batch_classloss, \"regress_loss\":batch_localizationloss}\n",
    "    \n",
    "    def test_step(self, batch, **kwargs): \n",
    "        X, y = batch\n",
    "        \n",
    "        classes, coords = self.model(X, training=False)\n",
    "        \n",
    "        batch_classloss = self.closs(y[0], classes)\n",
    "        batch_localizationloss = self.lloss(tf.cast(y[1], tf.float32), coords)\n",
    "        total_loss = batch_localizationloss+0.5*batch_classloss\n",
    "        \n",
    "        return {\"total_loss\":total_loss, \"class_loss\":batch_classloss, \"regress_loss\":batch_localizationloss}\n",
    "        \n",
    "    def call(self, X, **kwargs): \n",
    "        return self.model(X, **kwargs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "model = FaceTracker(facetracker)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\optimizers\\__init__.py:309: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "model.compile(opt, classloss, regressloss)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 10.2 Train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "logdir='logs'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=logdir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "893/893 [==============================] - 3434s 4s/step - total_loss: 1.0274 - class_loss: 0.2516 - regress_loss: 0.9016 - val_total_loss: 3.2477 - val_class_loss: 1.1869 - val_regress_loss: 2.6543\n",
      "Epoch 2/10\n",
      "893/893 [==============================] - 3604s 4s/step - total_loss: 0.7596 - class_loss: 0.1582 - regress_loss: 0.6805 - val_total_loss: 3.2890 - val_class_loss: 1.0801 - val_regress_loss: 2.7489\n",
      "Epoch 3/10\n",
      "893/893 [==============================] - 3337s 4s/step - total_loss: 0.6570 - class_loss: 0.1237 - regress_loss: 0.5952 - val_total_loss: 3.7399 - val_class_loss: 1.3759 - val_regress_loss: 3.0519\n",
      "Epoch 4/10\n",
      "893/893 [==============================] - 3187s 4s/step - total_loss: 0.5768 - class_loss: 0.0975 - regress_loss: 0.5281 - val_total_loss: 3.9104 - val_class_loss: 1.4422 - val_regress_loss: 3.1893\n",
      "Epoch 5/10\n",
      "812/893 [==========================>...] - ETA: 4:02 - total_loss: 0.5315 - class_loss: 0.0863 - regress_loss: 0.4884"
     ]
    }
   ],
   "source": [
    "hist = model.fit(train, epochs=10, validation_data=val, callbacks=[tensorboard_callback])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/4\n",
      "893/893 [==============================] - 2161s 2s/step - total_loss: 1.0611 - class_loss: 0.2611 - regress_loss: 0.9306 - val_total_loss: 2.9681 - val_class_loss: 0.9012 - val_regress_loss: 2.5175\n",
      "Epoch 2/4\n",
      "488/893 [===============>..............] - ETA: 14:53 - total_loss: 0.8055 - class_loss: 0.1741 - regress_loss: 0.7184"
     ]
    }
   ],
   "source": [
    "hist = model.fit(train, epochs=4, validation_data=val, callbacks=[tensorboard_callback])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    " final training on following parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/6\n",
      "893/893 [==============================] - 2461s 3s/step - total_loss: 1.0690 - class_loss: 0.2690 - regress_loss: 0.9346 - val_total_loss: 3.1664 - val_class_loss: 1.2630 - val_regress_loss: 2.5349\n",
      "Epoch 2/6\n",
      "893/893 [==============================] - 2530s 3s/step - total_loss: 0.7964 - class_loss: 0.1762 - regress_loss: 0.7083 - val_total_loss: 3.0112 - val_class_loss: 0.9904 - val_regress_loss: 2.5160\n",
      "Epoch 3/6\n",
      "893/893 [==============================] - 2442s 3s/step - total_loss: 0.6529 - class_loss: 0.1229 - regress_loss: 0.5915 - val_total_loss: 4.9322 - val_class_loss: 1.8040 - val_regress_loss: 4.0302\n",
      "Epoch 4/6\n",
      "893/893 [==============================] - 2419s 3s/step - total_loss: 0.5880 - class_loss: 0.1038 - regress_loss: 0.5361 - val_total_loss: 3.6207 - val_class_loss: 1.2245 - val_regress_loss: 3.0085\n",
      "Epoch 5/6\n",
      "893/893 [==============================] - 2358s 3s/step - total_loss: 0.5464 - class_loss: 0.0914 - regress_loss: 0.5007 - val_total_loss: 3.2408 - val_class_loss: 1.3026 - val_regress_loss: 2.5895\n",
      "Epoch 6/6\n",
      "893/893 [==============================] - 2141s 2s/step - total_loss: 0.5074 - class_loss: 0.0795 - regress_loss: 0.4677 - val_total_loss: 4.3729 - val_class_loss: 2.1030 - val_regress_loss: 3.3214\n"
     ]
    }
   ],
   "source": [
    "hist = model.fit(train, epochs=6, validation_data=val, callbacks=[tensorboard_callback])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 10.3 Plot Performance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'total_loss': [0.27117621898651123,\n",
       "  0.43171826004981995,\n",
       "  0.4246556758880615,\n",
       "  0.0678386315703392,\n",
       "  0.10874461382627487,\n",
       "  0.18816711008548737],\n",
       " 'class_loss': [0.036681946367025375,\n",
       "  0.1341761350631714,\n",
       "  0.17337539792060852,\n",
       "  0.0001947887212736532,\n",
       "  0.00013150667655281723,\n",
       "  0.047165293246507645],\n",
       " 'regress_loss': [0.2528352439403534,\n",
       "  0.36463019251823425,\n",
       "  0.33796799182891846,\n",
       "  0.06774123758077621,\n",
       "  0.10867886245250702,\n",
       "  0.1645844578742981],\n",
       " 'val_total_loss': [3.1663973331451416,\n",
       "  3.0111796855926514,\n",
       "  4.932160377502441,\n",
       "  3.6206915378570557,\n",
       "  3.240795135498047,\n",
       "  4.372920989990234],\n",
       " 'val_class_loss': [1.2629976272583008,\n",
       "  0.9903843998908997,\n",
       "  1.8039813041687012,\n",
       "  1.22446870803833,\n",
       "  1.3026294708251953,\n",
       "  2.1029558181762695],\n",
       " 'val_regress_loss': [2.534898519515991,\n",
       "  2.5159873962402344,\n",
       "  4.030169486999512,\n",
       "  3.0084571838378906,\n",
       "  2.589480400085449,\n",
       "  3.3214433193206787]}"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hist.history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x500 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(ncols=3, figsize=(20,5))\n",
    "\n",
    "ax[0].plot(hist.history['total_loss'], color='teal', label='loss')\n",
    "ax[0].plot(hist.history['val_total_loss'], color='orange', label='val loss')\n",
    "ax[0].title.set_text('Loss')\n",
    "ax[0].legend()\n",
    "\n",
    "ax[1].plot(hist.history['class_loss'], color='teal', label='class loss')\n",
    "ax[1].plot(hist.history['val_class_loss'], color='orange', label='val class loss')\n",
    "ax[1].title.set_text('Classification Loss')\n",
    "ax[1].legend()\n",
    "\n",
    "ax[2].plot(hist.history['regress_loss'], color='teal', label='regress loss')\n",
    "ax[2].plot(hist.history['val_regress_loss'], color='orange', label='val regress loss')\n",
    "ax[2].title.set_text('Regression Loss')\n",
    "ax[2].legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "val loss had more better bet fit line than the loss's best fit line, in loss, classification loss and regression loss. (all three)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 11. Make Predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 11.1 Make Predictions on Test Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "test_data = test.as_numpy_iterator()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 279,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "test_sample = test_data.next()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 280,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1/1 [==============================] - 1s 1s/step\n"
     ]
    }
   ],
   "source": [
    "yhat = facetracker.predict(test_sample[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "fig, ax = plt.subplots(ncols=2, figsize=(4, 4))\n",
    "for idx in range(2): \n",
    "    sample_image = test_sample[0][idx]\n",
    "    sample_coords = yhat[1][idx]\n",
    "    \n",
    "    if yhat[0][idx] > 0.9:\n",
    "        sample_image_copy = sample_image.copy()  # Create a copy of the image\n",
    "        cv2.rectangle(sample_image_copy, \n",
    "                      tuple(np.multiply(sample_coords[:2], [120, 120]).astype(int)),\n",
    "                      tuple(np.multiply(sample_coords[2:], [120, 120]).astype(int)), \n",
    "                      (255, 0, 0), 2)\n",
    "        ax[idx].imshow(sample_image_copy)\n",
    "    else:\n",
    "        ax[idx].imshow(sample_image)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 11.2 Save the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\losses.py:2976: The name tf.losses.sparse_softmax_cross_entropy is deprecated. Please use tf.compat.v1.losses.sparse_softmax_cross_entropy instead.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.models import load_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {
    "tags": []
   },
   "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"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\engine\\training.py:3103: UserWarning: You are saving your model as an HDF5 file via `model.save()`. This file format is considered legacy. We recommend using instead the native Keras format, e.g. `model.save('my_model.keras')`.\n",
      "  saving_api.save_model(\n"
     ]
    }
   ],
   "source": [
    "facetracker.save('facetracker.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\backend.py:1398: The name tf.executing_eagerly_outside_functions is deprecated. Please use tf.compat.v1.executing_eagerly_outside_functions instead.\n",
      "\n",
      "WARNING:tensorflow:From C:\\Users\\Computer\\anaconda3\\lib\\site-packages\\keras\\src\\layers\\pooling\\max_pooling2d.py:161: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.\n",
      "\n",
      "WARNING:tensorflow:No training configuration found in the save file, so the model was *not* compiled. Compile it manually.\n"
     ]
    }
   ],
   "source": [
    "facetracker = load_model('facetracker.h5')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 11.3 Real Time Detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import cv2\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "\n",
    "# Load the face tracking model (facetracker) and other necessary configurations\n",
    "\n",
    "cap = cv2.VideoCapture(0)  # Use default camera (index 0)\n",
    "\n",
    "while cap.isOpened():\n",
    "    ret, frame = cap.read()\n",
    "    \n",
    "    if not ret:\n",
    "        break\n",
    "    \n",
    "    frame = frame[50:500, 50:500, :]\n",
    "    \n",
    "    rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)\n",
    "    resized = tf.image.resize(rgb, (120, 120))\n",
    "    \n",
    "    yhat = facetracker.predict(np.expand_dims(resized / 255, 0))\n",
    "    sample_coords = yhat[1][0]\n",
    "    \n",
    "    if yhat[0] > 0.5:\n",
    "        # Controls the main rectangle\n",
    "        cv2.rectangle(frame,\n",
    "                      tuple(np.multiply(sample_coords[:2], [450, 450]).astype(int)),\n",
    "                      tuple(np.multiply(sample_coords[2:], [450, 450]).astype(int)),\n",
    "                      (255, 0, 0), 2)\n",
    "        # Controls the label rectangle\n",
    "        cv2.rectangle(frame,\n",
    "                      tuple(np.add(np.multiply(sample_coords[:2], [450, 450]).astype(int),\n",
    "                                    [0, -30])),\n",
    "                      tuple(np.add(np.multiply(sample_coords[:2], [450, 450]).astype(int),\n",
    "                                    [80, 0])),\n",
    "                      (255, 0, 0), -1)\n",
    "\n",
    "        # Controls the text rendered\n",
    "        cv2.putText(frame, 'face', tuple(np.add(np.multiply(sample_coords[:2], [450, 450]).astype(int),\n",
    "                                                 [0, -5])),\n",
    "                    cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2, cv2.LINE_AA)\n",
    "\n",
    "    cv2.imshow('EyeTrack', frame)\n",
    "\n",
    "    if cv2.waitKey(1) & 0xFF == ord('q'):\n",
    "        break\n",
    "\n",
    "cap.release()\n",
    "cv2.destroyAllWindows()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}