ASL_Model_Training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Isolated American Sign Language Recognition Model\n",
    "- Dataset used: GISLR"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Importing Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import mediapipe as mp\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import sklearn\n",
    "import tensorflow as tf\n",
    "import tensorflow_addons as tfa\n",
    "import os"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "N_ROWS = 543\n",
    "N_DIMS = 3\n",
    "DIM_NAMES = ['x', 'y', 'z']\n",
    "SEED = 42\n",
    "NUM_CLASSES = 250\n",
    "VERBOSE = 1\n",
    "\n",
    "INPUT_SIZE = 64\n",
    "\n",
    "BATCH_ALL_SIGNS_N = 4\n",
    "BATCH_SIZE = 256\n",
    "N_EPOCHS = 100\n",
    "LR_MAX = 1e-3\n",
    "N_WARMUP_EPOCHS = 0\n",
    "WD_RATIO = 0.05\n",
    "MASK_VAL = 423"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('train.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "train['sign_ord'] = train['sign'].astype('category').cat.codes\n",
    "\n",
    "SIGN2ORD = train[['sign', 'sign_ord']].set_index('sign').squeeze().to_dict()\n",
    "ORD2SIGN = train[['sign_ord', 'sign']].set_index('sign_ord').squeeze().to_dict()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "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>path</th>\n",
       "      <th>participant_id</th>\n",
       "      <th>sequence_id</th>\n",
       "      <th>sign</th>\n",
       "      <th>sign_ord</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>train_landmark_files/26734/1000035562.parquet</td>\n",
       "      <td>26734</td>\n",
       "      <td>1000035562</td>\n",
       "      <td>blow</td>\n",
       "      <td>25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>train_landmark_files/28656/1000106739.parquet</td>\n",
       "      <td>28656</td>\n",
       "      <td>1000106739</td>\n",
       "      <td>wait</td>\n",
       "      <td>232</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>train_landmark_files/16069/100015657.parquet</td>\n",
       "      <td>16069</td>\n",
       "      <td>100015657</td>\n",
       "      <td>cloud</td>\n",
       "      <td>48</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>train_landmark_files/25571/1000210073.parquet</td>\n",
       "      <td>25571</td>\n",
       "      <td>1000210073</td>\n",
       "      <td>bird</td>\n",
       "      <td>23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>train_landmark_files/62590/1000240708.parquet</td>\n",
       "      <td>62590</td>\n",
       "      <td>1000240708</td>\n",
       "      <td>owie</td>\n",
       "      <td>164</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            path  participant_id  sequence_id  \\\n",
       "0  train_landmark_files/26734/1000035562.parquet           26734   1000035562   \n",
       "1  train_landmark_files/28656/1000106739.parquet           28656   1000106739   \n",
       "2   train_landmark_files/16069/100015657.parquet           16069    100015657   \n",
       "3  train_landmark_files/25571/1000210073.parquet           25571   1000210073   \n",
       "4  train_landmark_files/62590/1000240708.parquet           62590   1000240708   \n",
       "\n",
       "    sign  sign_ord  \n",
       "0   blow        25  \n",
       "1   wait       232  \n",
       "2  cloud        48  \n",
       "3   bird        23  \n",
       "4   owie       164  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 94477 entries, 0 to 94476\n",
      "Data columns (total 5 columns):\n",
      " #   Column          Non-Null Count  Dtype \n",
      "---  ------          --------------  ----- \n",
      " 0   path            94477 non-null  object\n",
      " 1   participant_id  94477 non-null  int64 \n",
      " 2   sequence_id     94477 non-null  int64 \n",
      " 3   sign            94477 non-null  object\n",
      " 4   sign_ord        94477 non-null  int16 \n",
      "dtypes: int16(1), int64(2), object(2)\n",
      "memory usage: 3.1+ MB\n"
     ]
    }
   ],
   "source": [
    "train.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "ROWS_PER_FRAME = 543  # number of landmarks per frame\n",
    "\n",
    "def load_relevant_data_subset(pq_path):\n",
    "    data_columns = ['x', 'y', 'z']\n",
    "    data = pd.read_parquet(pq_path, columns=data_columns)\n",
    "    n_frames = int(len(data) / ROWS_PER_FRAME)\n",
    "    data = data.values.reshape(n_frames, ROWS_PER_FRAME, len(data_columns))\n",
    "    return data.astype(np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# HAND_IDXS: 21, N_COLS: 66\n"
     ]
    }
   ],
   "source": [
    "USE_TYPES = ['left_hand', 'pose', 'right_hand']\n",
    "START_IDX = 468\n",
    "LIPS_IDXS0 = np.array([\n",
    "        61, 185, 40, 39, 37, 0, 267, 269, 270, 409,\n",
    "        291, 146, 91, 181, 84, 17, 314, 405, 321, 375,\n",
    "        78, 191, 80, 81, 82, 13, 312, 311, 310, 415,\n",
    "        95, 88, 178, 87, 14, 317, 402, 318, 324, 308,\n",
    "    ])\n",
    "# Landmark indices in original data\n",
    "LEFT_HAND_IDXS0 = np.arange(468,489)\n",
    "RIGHT_HAND_IDXS0 = np.arange(522,543)\n",
    "LEFT_POSE_IDXS0 = np.array([502, 504, 506, 508, 510])\n",
    "RIGHT_POSE_IDXS0 = np.array([503, 505, 507, 509, 511])\n",
    "LANDMARK_IDXS_LEFT_DOMINANT0 = np.concatenate((LIPS_IDXS0, LEFT_HAND_IDXS0, LEFT_POSE_IDXS0))\n",
    "LANDMARK_IDXS_RIGHT_DOMINANT0 = np.concatenate((LIPS_IDXS0, RIGHT_HAND_IDXS0, RIGHT_POSE_IDXS0))\n",
    "HAND_IDXS0 = np.concatenate((LEFT_HAND_IDXS0, RIGHT_HAND_IDXS0), axis=0)\n",
    "N_COLS = LANDMARK_IDXS_LEFT_DOMINANT0.size\n",
    "# Landmark indices in processed data\n",
    "LIPS_IDXS = np.argwhere(np.isin(LANDMARK_IDXS_LEFT_DOMINANT0, LIPS_IDXS0)).squeeze()\n",
    "LEFT_HAND_IDXS = np.argwhere(np.isin(LANDMARK_IDXS_LEFT_DOMINANT0, LEFT_HAND_IDXS0)).squeeze()\n",
    "RIGHT_HAND_IDXS = np.argwhere(np.isin(LANDMARK_IDXS_LEFT_DOMINANT0, RIGHT_HAND_IDXS0)).squeeze()\n",
    "HAND_IDXS = np.argwhere(np.isin(LANDMARK_IDXS_LEFT_DOMINANT0, HAND_IDXS0)).squeeze()\n",
    "POSE_IDXS = np.argwhere(np.isin(LANDMARK_IDXS_LEFT_DOMINANT0, LEFT_POSE_IDXS0)).squeeze()\n",
    "\n",
    "print(f'# HAND_IDXS: {len(HAND_IDXS)}, N_COLS: {N_COLS}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LIPS_START: 0, LEFT_HAND_START: 40, RIGHT_HAND_START: 61, POSE_START: 61\n"
     ]
    }
   ],
   "source": [
    "LIPS_START = 0\n",
    "LEFT_HAND_START = LIPS_IDXS.size\n",
    "RIGHT_HAND_START = LEFT_HAND_START + LEFT_HAND_IDXS.size\n",
    "POSE_START = RIGHT_HAND_START + RIGHT_HAND_IDXS.size\n",
    "\n",
    "print(f'LIPS_START: {LIPS_START}, LEFT_HAND_START: {LEFT_HAND_START}, RIGHT_HAND_START: {RIGHT_HAND_START}, POSE_START: {POSE_START}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-05-06 21:48:59.876115: I metal_plugin/src/device/metal_device.cc:1154] Metal device set to: Apple M1 Pro\n",
      "2024-05-06 21:48:59.876145: I metal_plugin/src/device/metal_device.cc:296] systemMemory: 16.00 GB\n",
      "2024-05-06 21:48:59.876150: I metal_plugin/src/device/metal_device.cc:313] maxCacheSize: 5.33 GB\n",
      "2024-05-06 21:48:59.876376: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:306] Could not identify NUMA node of platform GPU ID 0, defaulting to 0. Your kernel may not have been built with NUMA support.\n",
      "2024-05-06 21:48:59.876711: I tensorflow/core/common_runtime/pluggable_device/pluggable_device_factory.cc:272] Created TensorFlow device (/job:localhost/replica:0/task:0/device:GPU:0 with 0 MB memory) -> physical PluggableDevice (device: 0, name: METAL, pci bus id: <undefined>)\n"
     ]
    }
   ],
   "source": [
    "class PreprocessLayer(tf.keras.layers.Layer):\n",
    "    def __init__(self):\n",
    "        super(PreprocessLayer, self).__init__()\n",
    "        normalisation_correction = tf.constant([\n",
    "                    # Add 0.50 to left hand (original right hand) and substract 0.50 of right hand (original left hand)\n",
    "                    [0] * len(LIPS_IDXS) + [0.50] * len(LEFT_HAND_IDXS) + [0.50] * len(POSE_IDXS),\n",
    "                    # Y coordinates stay intact\n",
    "                    [0] * len(LANDMARK_IDXS_LEFT_DOMINANT0),\n",
    "                    # Z coordinates stay intact\n",
    "                    [0] * len(LANDMARK_IDXS_LEFT_DOMINANT0),\n",
    "                ],\n",
    "                dtype=tf.float32,\n",
    "            )\n",
    "        self.normalisation_correction = tf.transpose(normalisation_correction, [1,0])\n",
    "        \n",
    "    def pad_edge(self, t, repeats, side):\n",
    "        if side == 'LEFT':\n",
    "            return tf.concat((tf.repeat(t[:1], repeats=repeats, axis=0), t), axis=0)\n",
    "        elif side == 'RIGHT':\n",
    "            return tf.concat((t, tf.repeat(t[-1:], repeats=repeats, axis=0)), axis=0)\n",
    "    \n",
    "    @tf.function(\n",
    "        input_signature=(tf.TensorSpec(shape=[None,N_ROWS,N_DIMS], dtype=tf.float32),),\n",
    "    )\n",
    "    def call(self, data0):\n",
    "        # Number of Frames in Video\n",
    "        N_FRAMES0 = tf.shape(data0)[0]\n",
    "        \n",
    "        # Find dominant hand by comparing summed absolute coordinates\n",
    "        left_hand_sum = tf.math.reduce_sum(tf.where(tf.math.is_nan(tf.gather(data0, LEFT_HAND_IDXS0, axis=1)), 0, 1))\n",
    "        right_hand_sum = tf.math.reduce_sum(tf.where(tf.math.is_nan(tf.gather(data0, RIGHT_HAND_IDXS0, axis=1)), 0, 1))\n",
    "        left_dominant = left_hand_sum >= right_hand_sum\n",
    "        \n",
    "        # Count non NaN Hand values in each frame for the dominant hand\n",
    "        if left_dominant:\n",
    "            frames_hands_non_nan_sum = tf.math.reduce_sum(\n",
    "                    tf.where(tf.math.is_nan(tf.gather(data0, LEFT_HAND_IDXS0, axis=1)), 0, 1),\n",
    "                    axis=[1, 2],\n",
    "                )\n",
    "        else:\n",
    "            frames_hands_non_nan_sum = tf.math.reduce_sum(\n",
    "                    tf.where(tf.math.is_nan(tf.gather(data0, RIGHT_HAND_IDXS0, axis=1)), 0, 1),\n",
    "                    axis=[1, 2],\n",
    "                )\n",
    "        \n",
    "        # Find frames indices with coordinates of dominant hand\n",
    "        non_empty_frames_idxs = tf.where(frames_hands_non_nan_sum > 0)\n",
    "        non_empty_frames_idxs = tf.squeeze(non_empty_frames_idxs, axis=1)\n",
    "        # Filter frames\n",
    "        data = tf.gather(data0, non_empty_frames_idxs, axis=0)\n",
    "        \n",
    "        # Cast Indices in float32 to be compatible with Tensorflow Lite\n",
    "        non_empty_frames_idxs = tf.cast(non_empty_frames_idxs, tf.float32)\n",
    "        # Normalize to start with 0\n",
    "        non_empty_frames_idxs -= tf.reduce_min(non_empty_frames_idxs)\n",
    "        \n",
    "        # Number of Frames in Filtered Video\n",
    "        N_FRAMES = tf.shape(data)[0]\n",
    "        \n",
    "        # Gather Relevant Landmark Columns\n",
    "        if left_dominant:\n",
    "            data = tf.gather(data, LANDMARK_IDXS_LEFT_DOMINANT0, axis=1)\n",
    "        else:\n",
    "            data = tf.gather(data, LANDMARK_IDXS_RIGHT_DOMINANT0, axis=1)\n",
    "            data = (\n",
    "                    self.normalisation_correction + (\n",
    "                        (data - self.normalisation_correction) * tf.where(self.normalisation_correction != 0, -1.0, 1.0))\n",
    "                )\n",
    "        \n",
    "        # Video fits in INPUT_SIZE\n",
    "        if N_FRAMES < INPUT_SIZE:\n",
    "            # Pad With -1 to indicate padding\n",
    "            non_empty_frames_idxs = tf.pad(non_empty_frames_idxs, [[0, INPUT_SIZE-N_FRAMES]], constant_values=-1)\n",
    "            # Pad Data With Zeros\n",
    "            data = tf.pad(data, [[0, INPUT_SIZE-N_FRAMES], [0,0], [0,0]], constant_values=0)\n",
    "            # Fill NaN Values With 0\n",
    "            data = tf.where(tf.math.is_nan(data), 0.0, data)\n",
    "            return data, non_empty_frames_idxs\n",
    "        # Video needs to be downsampled to INPUT_SIZE\n",
    "        else:\n",
    "            # Repeat\n",
    "            if N_FRAMES < INPUT_SIZE**2:\n",
    "                repeats = tf.math.floordiv(INPUT_SIZE * INPUT_SIZE, N_FRAMES0)\n",
    "                data = tf.repeat(data, repeats=repeats, axis=0)\n",
    "                non_empty_frames_idxs = tf.repeat(non_empty_frames_idxs, repeats=repeats, axis=0)\n",
    "\n",
    "            # Pad To Multiple Of Input Size\n",
    "            pool_size = tf.math.floordiv(len(data), INPUT_SIZE)\n",
    "            if tf.math.mod(len(data), INPUT_SIZE) > 0:\n",
    "                pool_size += 1\n",
    "\n",
    "            if pool_size == 1:\n",
    "                pad_size = (pool_size * INPUT_SIZE) - len(data)\n",
    "            else:\n",
    "                pad_size = (pool_size * INPUT_SIZE) % len(data)\n",
    "\n",
    "            # Pad Start/End with Start/End value\n",
    "            pad_left = tf.math.floordiv(pad_size, 2) + tf.math.floordiv(INPUT_SIZE, 2)\n",
    "            pad_right = tf.math.floordiv(pad_size, 2) + tf.math.floordiv(INPUT_SIZE, 2)\n",
    "            if tf.math.mod(pad_size, 2) > 0:\n",
    "                pad_right += 1\n",
    "\n",
    "            # Pad By Concatenating Left/Right Edge Values\n",
    "            data = self.pad_edge(data, pad_left, 'LEFT')\n",
    "            data = self.pad_edge(data, pad_right, 'RIGHT')\n",
    "\n",
    "            # Pad Non Empty Frame Indices\n",
    "            non_empty_frames_idxs = self.pad_edge(non_empty_frames_idxs, pad_left, 'LEFT')\n",
    "            non_empty_frames_idxs = self.pad_edge(non_empty_frames_idxs, pad_right, 'RIGHT')\n",
    "\n",
    "            # Reshape to Mean Pool\n",
    "            data = tf.reshape(data, [INPUT_SIZE, -1, N_COLS, N_DIMS])\n",
    "            non_empty_frames_idxs = tf.reshape(non_empty_frames_idxs, [INPUT_SIZE, -1])\n",
    "\n",
    "            # Mean Pool\n",
    "            data = tf.experimental.numpy.nanmean(data, axis=1)\n",
    "            non_empty_frames_idxs = tf.experimental.numpy.nanmean(non_empty_frames_idxs, axis=1)\n",
    "\n",
    "            # Fill NaN Values With 0\n",
    "            data = tf.where(tf.math.is_nan(data), 0.0, data)\n",
    "            \n",
    "            return data, non_empty_frames_idxs\n",
    "    \n",
    "preprocess_layer = PreprocessLayer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "LIPS_MEAN_X = np.zeros([LIPS_IDXS.size], dtype=np.float32)\n",
    "LIPS_MEAN_Y = np.zeros([LIPS_IDXS.size], dtype=np.float32)\n",
    "LIPS_STD_X = np.zeros([LIPS_IDXS.size], dtype=np.float32)\n",
    "LIPS_STD_Y = np.zeros([LIPS_IDXS.size], dtype=np.float32)\n",
    "LIPS_MEAN = np.array([LIPS_MEAN_X, LIPS_MEAN_Y]).T\n",
    "LIPS_STD = np.array([LIPS_STD_X, LIPS_STD_Y]).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "LEFT_HANDS_MEAN_X = np.zeros([LEFT_HAND_IDXS.size], dtype=np.float32)\n",
    "LEFT_HANDS_MEAN_Y = np.zeros([LEFT_HAND_IDXS.size], dtype=np.float32)\n",
    "LEFT_HANDS_STD_X = np.zeros([LEFT_HAND_IDXS.size], dtype=np.float32)\n",
    "LEFT_HANDS_STD_Y = np.zeros([LEFT_HAND_IDXS.size], dtype=np.float32)\n",
    "LEFT_HANDS_MEAN = np.array([LEFT_HANDS_MEAN_X, LEFT_HANDS_MEAN_Y]).T\n",
    "LEFT_HANDS_STD = np.array([LEFT_HANDS_STD_X, LEFT_HANDS_STD_Y]).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "POSE_MEAN_X = np.zeros([POSE_IDXS.size], dtype=np.float32)\n",
    "POSE_MEAN_Y = np.zeros([POSE_IDXS.size], dtype=np.float32)\n",
    "POSE_STD_X = np.zeros([POSE_IDXS.size], dtype=np.float32)\n",
    "POSE_STD_Y = np.zeros([POSE_IDXS.size], dtype=np.float32)\n",
    "POSE_MEAN = np.array([POSE_MEAN_X, POSE_MEAN_Y]).T\n",
    "POSE_STD = np.array([POSE_STD_X, POSE_STD_Y]).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = np.load(f'GISLR/X.npy')\n",
    "y_train = np.load(f'GISLR/y.npy')\n",
    "NON_EMPTY_FRAME_IDXS_TRAIN = np.load(f'GISLR/NON_EMPTY_FRAME_IDXS.npy')\n",
    "validation_data = None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_batch_all_signs(X, y, NON_EMPTY_FRAME_IDXS, n=BATCH_ALL_SIGNS_N):\n",
    "    # Arrays to store batch in\n",
    "    X_batch = np.zeros([NUM_CLASSES*n, INPUT_SIZE, N_COLS, N_DIMS], dtype=np.float32)\n",
    "    y_batch = np.arange(0, NUM_CLASSES, step=1/n, dtype=np.float32).astype(np.int64)\n",
    "    non_empty_frame_idxs_batch = np.zeros([NUM_CLASSES*n, INPUT_SIZE], dtype=np.float32)\n",
    "    \n",
    "    # Dictionary mapping ordinally encoded sign to corresponding sample indices\n",
    "    CLASS2IDXS = {}\n",
    "    for i in range(NUM_CLASSES):\n",
    "        CLASS2IDXS[i] = np.argwhere(y == i).squeeze().astype(np.int32)\n",
    "            \n",
    "    while True:\n",
    "        # Fill batch arrays\n",
    "        for i in range(NUM_CLASSES):\n",
    "            idxs = np.random.choice(CLASS2IDXS[i], n)\n",
    "            X_batch[i*n:(i+1)*n] = X[idxs]\n",
    "            non_empty_frame_idxs_batch[i*n:(i+1)*n] = NON_EMPTY_FRAME_IDXS[idxs]\n",
    "        \n",
    "        yield { 'frames': X_batch, 'non_empty_frame_idxs': non_empty_frame_idxs_batch }, y_batch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Transformer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "UNITS: 512\n"
     ]
    }
   ],
   "source": [
    "LAYER_NORM_EPS = 1e-6\n",
    "\n",
    "# Dense layer units for landmarks\n",
    "LIPS_UNITS = 384\n",
    "HANDS_UNITS = 384\n",
    "POSE_UNITS = 384\n",
    "# final embedding and transformer embedding size\n",
    "UNITS = 512\n",
    "\n",
    "# Transformer\n",
    "NUM_BLOCKS = 2\n",
    "MLP_RATIO = 2\n",
    "\n",
    "# Dropout\n",
    "EMBEDDING_DROPOUT = 0.00\n",
    "MLP_DROPOUT_RATIO = 0.30\n",
    "CLASSIFIER_DROPOUT_RATIO = 0.10\n",
    "\n",
    "# Initiailizers\n",
    "INIT_HE_UNIFORM = tf.keras.initializers.he_uniform\n",
    "INIT_GLOROT_UNIFORM = tf.keras.initializers.glorot_uniform\n",
    "INIT_ZEROS = tf.keras.initializers.constant(0.0)\n",
    "# Activations\n",
    "GELU = tf.keras.activations.gelu\n",
    "\n",
    "print(f'UNITS: {UNITS}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def scaled_dot_product(q,k,v, softmax, attention_mask):\n",
    "    #calculates Q . K(transpose)\n",
    "    qkt = tf.matmul(q,k,transpose_b=True)\n",
    "    #caculates scaling factor\n",
    "    dk = tf.math.sqrt(tf.cast(q.shape[-1],dtype=tf.float32))\n",
    "    scaled_qkt = qkt/dk\n",
    "    softmax = softmax(scaled_qkt, mask=attention_mask)\n",
    "    \n",
    "    z = tf.matmul(softmax,v)\n",
    "    #shape: (m,Tx,depth), same shape as q,k,v\n",
    "    return z\n",
    "\n",
    "class MultiHeadAttention(tf.keras.layers.Layer):\n",
    "    def __init__(self,d_model,num_of_heads):\n",
    "        super(MultiHeadAttention,self).__init__()\n",
    "        self.d_model = d_model\n",
    "        self.num_of_heads = num_of_heads\n",
    "        self.depth = d_model//num_of_heads\n",
    "        self.wq = [tf.keras.layers.Dense(self.depth) for i in range(num_of_heads)]\n",
    "        self.wk = [tf.keras.layers.Dense(self.depth) for i in range(num_of_heads)]\n",
    "        self.wv = [tf.keras.layers.Dense(self.depth) for i in range(num_of_heads)]\n",
    "        self.wo = tf.keras.layers.Dense(d_model)\n",
    "        self.softmax = tf.keras.layers.Softmax()\n",
    "        \n",
    "    def call(self,x, attention_mask):\n",
    "        \n",
    "        multi_attn = []\n",
    "        for i in range(self.num_of_heads):\n",
    "            Q = self.wq[i](x)\n",
    "            K = self.wk[i](x)\n",
    "            V = self.wv[i](x)\n",
    "            multi_attn.append(scaled_dot_product(Q,K,V, self.softmax, attention_mask))\n",
    "            \n",
    "        multi_head = tf.concat(multi_attn,axis=-1)\n",
    "        multi_head_attention = self.wo(multi_head)\n",
    "        return multi_head_attention"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Transformer(tf.keras.Model):\n",
    "    def __init__(self, num_blocks):\n",
    "        super(Transformer, self).__init__(name='transformer')\n",
    "        self.num_blocks = num_blocks\n",
    "    \n",
    "    def build(self, input_shape):\n",
    "        self.ln_1s = []\n",
    "        self.mhas = []\n",
    "        self.ln_2s = []\n",
    "        self.mlps = []\n",
    "        # Make Transformer Blocks\n",
    "        for i in range(self.num_blocks):\n",
    "            # Multi Head Attention\n",
    "            self.mhas.append(MultiHeadAttention(UNITS, 8))\n",
    "            # Multi Layer Perception\n",
    "            self.mlps.append(tf.keras.Sequential([\n",
    "                tf.keras.layers.Dense(UNITS * MLP_RATIO, activation=GELU, kernel_initializer=INIT_GLOROT_UNIFORM),\n",
    "                tf.keras.layers.Dropout(MLP_DROPOUT_RATIO),\n",
    "                tf.keras.layers.Dense(UNITS, kernel_initializer=INIT_HE_UNIFORM),\n",
    "            ]))\n",
    "        \n",
    "    def call(self, x, attention_mask):\n",
    "        # Iterate input over transformer blocks\n",
    "        for mha, mlp in zip(self.mhas, self.mlps):\n",
    "            x = x + mha(x, attention_mask)\n",
    "            x = x + mlp(x)\n",
    "    \n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "class LandmarkEmbedding(tf.keras.Model):\n",
    "    def __init__(self, units, name):\n",
    "        super(LandmarkEmbedding, self).__init__(name=f'{name}_embedding')\n",
    "        self.units = units\n",
    "        \n",
    "    def build(self, input_shape):\n",
    "        # Embedding for missing landmark in frame, initizlied with zeros\n",
    "        self.empty_embedding = self.add_weight(\n",
    "            name=f'{self.name}_empty_embedding',\n",
    "            shape=[self.units],\n",
    "            initializer=INIT_ZEROS,\n",
    "        )\n",
    "        # Embedding\n",
    "        self.dense = tf.keras.Sequential([\n",
    "            tf.keras.layers.Dense(self.units, name=f'{self.name}_dense_1', use_bias=False, kernel_initializer=INIT_GLOROT_UNIFORM),\n",
    "            tf.keras.layers.Activation(GELU),\n",
    "            tf.keras.layers.Dense(self.units, name=f'{self.name}_dense_2', use_bias=False, kernel_initializer=INIT_HE_UNIFORM),\n",
    "        ], name=f'{self.name}_dense')\n",
    "\n",
    "    def call(self, x):\n",
    "        return tf.where(\n",
    "                # Checks whether landmark is missing in frame\n",
    "                tf.reduce_sum(x, axis=2, keepdims=True) == 0,\n",
    "                # If so, the empty embedding is used\n",
    "                self.empty_embedding,\n",
    "                # Otherwise the landmark data is embedded\n",
    "                self.dense(x),\n",
    "            )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Embedding(tf.keras.Model):\n",
    "    def __init__(self):\n",
    "        super(Embedding, self).__init__()\n",
    "        \n",
    "    def get_diffs(self, l):\n",
    "        S = l.shape[2]\n",
    "        other = tf.expand_dims(l, 3)\n",
    "        other = tf.repeat(other, S, axis=3)\n",
    "        other = tf.transpose(other, [0,1,3,2])\n",
    "        diffs = tf.expand_dims(l, 3) - other\n",
    "        diffs = tf.reshape(diffs, [-1, INPUT_SIZE, S*S])\n",
    "        return diffs\n",
    "\n",
    "    def build(self, input_shape):\n",
    "        # Positional Embedding, initialized with zeros\n",
    "        self.positional_embedding = tf.keras.layers.Embedding(INPUT_SIZE+1, UNITS, embeddings_initializer=INIT_ZEROS)\n",
    "        # Embedding layer for Landmarks\n",
    "        self.lips_embedding = LandmarkEmbedding(LIPS_UNITS, 'lips')\n",
    "        self.left_hand_embedding = LandmarkEmbedding(HANDS_UNITS, 'left_hand')\n",
    "        self.pose_embedding = LandmarkEmbedding(POSE_UNITS, 'pose')\n",
    "        # Landmark Weights\n",
    "        self.landmark_weights = tf.Variable(tf.zeros([3], dtype=tf.float32), name='landmark_weights')\n",
    "        # Fully Connected Layers for combined landmarks\n",
    "        self.fc = tf.keras.Sequential([\n",
    "            tf.keras.layers.Dense(UNITS, name='fully_connected_1', use_bias=False, kernel_initializer=INIT_GLOROT_UNIFORM),\n",
    "            tf.keras.layers.Activation(GELU),\n",
    "            tf.keras.layers.Dense(UNITS, name='fully_connected_2', use_bias=False, kernel_initializer=INIT_HE_UNIFORM),\n",
    "        ], name='fc')\n",
    "\n",
    "\n",
    "    def call(self, lips0, left_hand0, pose0, non_empty_frame_idxs, training=False):\n",
    "        # Lips\n",
    "        lips_embedding = self.lips_embedding(lips0)\n",
    "        # Left Hand\n",
    "        left_hand_embedding = self.left_hand_embedding(left_hand0)\n",
    "        # Pose\n",
    "        pose_embedding = self.pose_embedding(pose0)\n",
    "        # Merge Embeddings of all landmarks with mean pooling\n",
    "        x = tf.stack((\n",
    "            lips_embedding, left_hand_embedding, pose_embedding,\n",
    "        ), axis=3)\n",
    "        x = x * tf.nn.softmax(self.landmark_weights)\n",
    "        x = tf.reduce_sum(x, axis=3)\n",
    "        # Fully Connected Layers\n",
    "        x = self.fc(x)\n",
    "        # Add Positional Embedding\n",
    "        max_frame_idxs = tf.clip_by_value(\n",
    "                tf.reduce_max(non_empty_frame_idxs, axis=1, keepdims=True),\n",
    "                1,\n",
    "                np.PINF,\n",
    "            )\n",
    "        normalised_non_empty_frame_idxs = tf.where(\n",
    "            tf.math.equal(non_empty_frame_idxs, -1.0),\n",
    "            INPUT_SIZE,\n",
    "            tf.cast(\n",
    "                non_empty_frame_idxs / max_frame_idxs * INPUT_SIZE,\n",
    "                tf.int32,\n",
    "            ),\n",
    "        )\n",
    "        x = x + self.positional_embedding(normalised_non_empty_frame_idxs)\n",
    "        \n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def scce_with_ls(y_true, y_pred):\n",
    "    # One Hot Encode Sparsely Encoded Target Sign\n",
    "    y_true = tf.cast(y_true, tf.int32)\n",
    "    y_true = tf.one_hot(y_true, NUM_CLASSES, axis=1)\n",
    "    y_true = tf.squeeze(y_true, axis=2)\n",
    "    # Categorical Crossentropy with native label smoothing support\n",
    "    return tf.keras.losses.categorical_crossentropy(y_true, y_pred, label_smoothing=0.25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_model():\n",
    "    # Inputs\n",
    "    frames = tf.keras.layers.Input([INPUT_SIZE, N_COLS, N_DIMS], dtype=tf.float32, name='frames')\n",
    "    non_empty_frame_idxs = tf.keras.layers.Input([INPUT_SIZE], dtype=tf.float32, name='non_empty_frame_idxs')\n",
    "    # Padding Mask\n",
    "    mask0 = tf.cast(tf.math.not_equal(non_empty_frame_idxs, -1), tf.float32)\n",
    "    mask0 = tf.expand_dims(mask0, axis=2)\n",
    "    # Random Frame Masking\n",
    "    mask = tf.where(\n",
    "        (tf.random.uniform(tf.shape(mask0)) > 0.25) & tf.math.not_equal(mask0, 0.0),\n",
    "        1.0,\n",
    "        0.0,\n",
    "    )\n",
    "    # Correct Samples Which are all masked now...\n",
    "    mask = tf.where(\n",
    "        tf.math.equal(tf.reduce_sum(mask, axis=[1,2], keepdims=True), 0.0),\n",
    "        mask0,\n",
    "        mask,\n",
    "    )\n",
    "    \n",
    "    \n",
    "    \"\"\"\n",
    "        left_hand: 468:489\n",
    "        pose: 489:522\n",
    "        right_hand: 522:543\n",
    "    \"\"\"\n",
    "    x = frames\n",
    "    x = tf.slice(x, [0,0,0,0], [-1,INPUT_SIZE, N_COLS, 2])\n",
    "    # LIPS\n",
    "    lips = tf.slice(x, [0,0,LIPS_START,0], [-1,INPUT_SIZE, 40, 2])\n",
    "    lips = tf.where(\n",
    "            tf.math.equal(lips, 0.0),\n",
    "            0.0,\n",
    "            (lips - LIPS_MEAN) / LIPS_STD,\n",
    "        )\n",
    "    # LEFT HAND\n",
    "    left_hand = tf.slice(x, [0,0,40,0], [-1,INPUT_SIZE, 21, 2])\n",
    "    left_hand = tf.where(\n",
    "            tf.math.equal(left_hand, 0.0),\n",
    "            0.0,\n",
    "            (left_hand - LEFT_HANDS_MEAN) / LEFT_HANDS_STD,\n",
    "        )\n",
    "    # POSE\n",
    "    pose = tf.slice(x, [0,0,61,0], [-1,INPUT_SIZE, 5, 2])\n",
    "    pose = tf.where(\n",
    "            tf.math.equal(pose, 0.0),\n",
    "            0.0,\n",
    "            (pose - POSE_MEAN) / POSE_STD,\n",
    "        )\n",
    "    \n",
    "    # Flatten\n",
    "    lips = tf.reshape(lips, [-1, INPUT_SIZE, 40*2])\n",
    "    left_hand = tf.reshape(left_hand, [-1, INPUT_SIZE, 21*2])\n",
    "    pose = tf.reshape(pose, [-1, INPUT_SIZE, 5*2])\n",
    "        \n",
    "    # Embedding\n",
    "    x = Embedding()(lips, left_hand, pose, non_empty_frame_idxs)\n",
    "    \n",
    "    # Encoder Transformer Blocks\n",
    "    x = Transformer(NUM_BLOCKS)(x, mask)\n",
    "    \n",
    "    # Pooling\n",
    "    x = tf.reduce_sum(x * mask, axis=1) / tf.reduce_sum(mask, axis=1)\n",
    "    # Classifier Dropout\n",
    "    x = tf.keras.layers.Dropout(CLASSIFIER_DROPOUT_RATIO)(x)\n",
    "    # Classification Layer\n",
    "    x = tf.keras.layers.Dense(NUM_CLASSES, activation=tf.keras.activations.softmax, kernel_initializer=INIT_GLOROT_UNIFORM)(x)\n",
    "    \n",
    "    outputs = x\n",
    "    \n",
    "    # Create Tensorflow Model\n",
    "    model = tf.keras.models.Model(inputs=[frames, non_empty_frame_idxs], outputs=outputs)\n",
    "    \n",
    "    # Sparse Categorical Cross Entropy With Label Smoothing\n",
    "    loss = scce_with_ls\n",
    "    \n",
    "    # Adam Optimizer with weight decay\n",
    "    optimizer = tfa.optimizers.AdamW(learning_rate=1e-3, weight_decay=1e-5, clipnorm=1.0)\n",
    "    \n",
    "    # TopK Metrics\n",
    "    metrics = [\n",
    "        tf.keras.metrics.SparseCategoricalAccuracy(name='acc'),\n",
    "        tf.keras.metrics.SparseTopKCategoricalAccuracy(k=5, name='top_5_acc'),\n",
    "        tf.keras.metrics.SparseTopKCategoricalAccuracy(k=10, name='top_10_acc'),\n",
    "    ]\n",
    "    \n",
    "    model.compile(loss=loss, optimizer=optimizer, metrics=metrics)\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = get_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                Output Shape                 Param #   Connected to                  \n",
      "==================================================================================================\n",
      " non_empty_frame_idxs (Inpu  [(None, 64)]                 0         []                            \n",
      " tLayer)                                                                                          \n",
      "                                                                                                  \n",
      " tf.math.not_equal (TFOpLam  (None, 64)                   0         ['non_empty_frame_idxs[0][0]']\n",
      " bda)                                                                                             \n",
      "                                                                                                  \n",
      " tf.cast (TFOpLambda)        (None, 64)                   0         ['tf.math.not_equal[0][0]']   \n",
      "                                                                                                  \n",
      " tf.expand_dims (TFOpLambda  (None, 64, 1)                0         ['tf.cast[0][0]']             \n",
      " )                                                                                                \n",
      "                                                                                                  \n",
      " frames (InputLayer)         [(None, 64, 66, 3)]          0         []                            \n",
      "                                                                                                  \n",
      " tf.compat.v1.shape (TFOpLa  (3,)                         0         ['tf.expand_dims[0][0]']      \n",
      " mbda)                                                                                            \n",
      "                                                                                                  \n",
      " tf.slice (TFOpLambda)       (None, 64, 66, 2)            0         ['frames[0][0]']              \n",
      "                                                                                                  \n",
      " tf.random.uniform (TFOpLam  (None, 64, 1)                0         ['tf.compat.v1.shape[0][0]']  \n",
      " bda)                                                                                             \n",
      "                                                                                                  \n",
      " tf.slice_1 (TFOpLambda)     (None, 64, 40, 2)            0         ['tf.slice[0][0]']            \n",
      "                                                                                                  \n",
      " tf.slice_2 (TFOpLambda)     (None, 64, 21, 2)            0         ['tf.slice[0][0]']            \n",
      "                                                                                                  \n",
      " tf.slice_3 (TFOpLambda)     (None, 64, 5, 2)             0         ['tf.slice[0][0]']            \n",
      "                                                                                                  \n",
      " tf.math.greater (TFOpLambd  (None, 64, 1)                0         ['tf.random.uniform[0][0]']   \n",
      " a)                                                                                               \n",
      "                                                                                                  \n",
      " tf.math.not_equal_1 (TFOpL  (None, 64, 1)                0         ['tf.expand_dims[0][0]']      \n",
      " ambda)                                                                                           \n",
      "                                                                                                  \n",
      " tf.math.subtract (TFOpLamb  (None, 64, 40, 2)            0         ['tf.slice_1[0][0]']          \n",
      " da)                                                                                              \n",
      "                                                                                                  \n",
      " tf.math.subtract_1 (TFOpLa  (None, 64, 21, 2)            0         ['tf.slice_2[0][0]']          \n",
      " mbda)                                                                                            \n",
      "                                                                                                  \n",
      " tf.math.subtract_2 (TFOpLa  (None, 64, 5, 2)             0         ['tf.slice_3[0][0]']          \n",
      " mbda)                                                                                            \n",
      "                                                                                                  \n",
      " tf.math.logical_and (TFOpL  (None, 64, 1)                0         ['tf.math.greater[0][0]',     \n",
      " ambda)                                                              'tf.math.not_equal_1[0][0]'] \n",
      "                                                                                                  \n",
      " tf.math.equal_1 (TFOpLambd  (None, 64, 40, 2)            0         ['tf.slice_1[0][0]']          \n",
      " a)                                                                                               \n",
      "                                                                                                  \n",
      " tf.math.truediv (TFOpLambd  (None, 64, 40, 2)            0         ['tf.math.subtract[0][0]']    \n",
      " a)                                                                                               \n",
      "                                                                                                  \n",
      " tf.math.equal_2 (TFOpLambd  (None, 64, 21, 2)            0         ['tf.slice_2[0][0]']          \n",
      " a)                                                                                               \n",
      "                                                                                                  \n",
      " tf.math.truediv_1 (TFOpLam  (None, 64, 21, 2)            0         ['tf.math.subtract_1[0][0]']  \n",
      " bda)                                                                                             \n",
      "                                                                                                  \n",
      " tf.math.equal_3 (TFOpLambd  (None, 64, 5, 2)             0         ['tf.slice_3[0][0]']          \n",
      " a)                                                                                               \n",
      "                                                                                                  \n",
      " tf.math.truediv_2 (TFOpLam  (None, 64, 5, 2)             0         ['tf.math.subtract_2[0][0]']  \n",
      " bda)                                                                                             \n",
      "                                                                                                  \n",
      " tf.where (TFOpLambda)       (None, 64, 1)                0         ['tf.math.logical_and[0][0]'] \n",
      "                                                                                                  \n",
      " tf.where_2 (TFOpLambda)     (None, 64, 40, 2)            0         ['tf.math.equal_1[0][0]',     \n",
      "                                                                     'tf.math.truediv[0][0]']     \n",
      "                                                                                                  \n",
      " tf.where_3 (TFOpLambda)     (None, 64, 21, 2)            0         ['tf.math.equal_2[0][0]',     \n",
      "                                                                     'tf.math.truediv_1[0][0]']   \n",
      "                                                                                                  \n",
      " tf.where_4 (TFOpLambda)     (None, 64, 5, 2)             0         ['tf.math.equal_3[0][0]',     \n",
      "                                                                     'tf.math.truediv_2[0][0]']   \n",
      "                                                                                                  \n",
      " tf.math.reduce_sum (TFOpLa  (None, 1, 1)                 0         ['tf.where[0][0]']            \n",
      " mbda)                                                                                            \n",
      "                                                                                                  \n",
      " tf.reshape (TFOpLambda)     (None, 64, 80)               0         ['tf.where_2[0][0]']          \n",
      "                                                                                                  \n",
      " tf.reshape_1 (TFOpLambda)   (None, 64, 42)               0         ['tf.where_3[0][0]']          \n",
      "                                                                                                  \n",
      " tf.reshape_2 (TFOpLambda)   (None, 64, 10)               0         ['tf.where_4[0][0]']          \n",
      "                                                                                                  \n",
      " tf.math.equal (TFOpLambda)  (None, 1, 1)                 0         ['tf.math.reduce_sum[0][0]']  \n",
      "                                                                                                  \n",
      " embedding (Embedding)       (None, 64, 512)              986243    ['tf.reshape[0][0]',          \n",
      "                                                                     'tf.reshape_1[0][0]',        \n",
      "                                                                     'tf.reshape_2[0][0]',        \n",
      "                                                                     'non_empty_frame_idxs[0][0]']\n",
      "                                                                                                  \n",
      " tf.where_1 (TFOpLambda)     (None, 64, 1)                0         ['tf.math.equal[0][0]',       \n",
      "                                                                     'tf.expand_dims[0][0]',      \n",
      "                                                                     'tf.where[0][0]']            \n",
      "                                                                                                  \n",
      " transformer (Transformer)   (None, 64, 512)              4201472   ['embedding[0][0]',           \n",
      "                                                                     'tf.where_1[0][0]']          \n",
      "                                                                                                  \n",
      " tf.math.multiply (TFOpLamb  (None, 64, 512)              0         ['transformer[0][0]',         \n",
      " da)                                                                 'tf.where_1[0][0]']          \n",
      "                                                                                                  \n",
      " tf.math.reduce_sum_1 (TFOp  (None, 512)                  0         ['tf.math.multiply[0][0]']    \n",
      " Lambda)                                                                                          \n",
      "                                                                                                  \n",
      " tf.math.reduce_sum_2 (TFOp  (None, 1)                    0         ['tf.where_1[0][0]']          \n",
      " Lambda)                                                                                          \n",
      "                                                                                                  \n",
      " tf.math.truediv_3 (TFOpLam  (None, 512)                  0         ['tf.math.reduce_sum_1[0][0]',\n",
      " bda)                                                                'tf.math.reduce_sum_2[0][0]']\n",
      "                                                                                                  \n",
      " dropout (Dropout)           (None, 512)                  0         ['tf.math.truediv_3[0][0]']   \n",
      "                                                                                                  \n",
      " dense (Dense)               (None, 250)                  128250    ['dropout[0][0]']             \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 5315965 (20.28 MB)\n",
      "Trainable params: 5315965 (20.28 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model Training"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Learning Rate Scheduling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "def lrfn(current_step, num_warmup_steps, lr_max, num_cycles=0.50, num_training_steps=N_EPOCHS):\n",
    "    WARMUP_METHOD = 'log'\n",
    "    if current_step < num_warmup_steps:\n",
    "        if WARMUP_METHOD == 'log':\n",
    "            return lr_max * 0.10 ** (num_warmup_steps - current_step)\n",
    "        else:\n",
    "            return lr_max * 2 ** -(num_warmup_steps - current_step)\n",
    "    else:\n",
    "        progress = float(current_step - num_warmup_steps) / float(max(1, num_training_steps - num_warmup_steps))\n",
    "\n",
    "        return max(0.0, 0.5 * (1.0 + math.cos(math.pi * float(num_cycles) * 2.0 * progress))) * lr_max"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2000x1000 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_lr_schedule(lr_schedule, epochs):\n",
    "    fig = plt.figure(figsize=(20, 10))\n",
    "    plt.plot([None] + lr_schedule + [None])\n",
    "    # X Labels\n",
    "    x = np.arange(1, epochs + 1)\n",
    "    x_axis_labels = [i if epochs <= 40 or i % 5 == 0 or i == 1 else None for i in range(1, epochs + 1)]\n",
    "    plt.xlim([1, epochs])\n",
    "    plt.xticks(x, x_axis_labels) # set tick step to 1 and let x axis start at 1\n",
    "    \n",
    "    # Increase y-limit for better readability\n",
    "    plt.ylim([0, max(lr_schedule) * 1.1])\n",
    "    \n",
    "    # Title\n",
    "    schedule_info = f'start: {lr_schedule[0]:.1E}, max: {max(lr_schedule):.1E}, final: {lr_schedule[-1]:.1E}'\n",
    "    plt.title(f'Step Learning Rate Schedule, {schedule_info}', size=18, pad=12)\n",
    "    \n",
    "    # Plot Learning Rates\n",
    "    for x, val in enumerate(lr_schedule):\n",
    "        if epochs <= 40 or x % 5 == 0 or x is epochs - 1:\n",
    "            if x < len(lr_schedule) - 1:\n",
    "                if lr_schedule[x - 1] < val:\n",
    "                    ha = 'right'\n",
    "                else:\n",
    "                    ha = 'left'\n",
    "            elif x == 0:\n",
    "                ha = 'right'\n",
    "            else:\n",
    "                ha = 'left'\n",
    "            plt.plot(x + 1, val, 'o', color='black');\n",
    "            offset_y = (max(lr_schedule) - min(lr_schedule)) * 0.02\n",
    "            plt.annotate(f'{val:.1E}', xy=(x + 1, val + offset_y), size=12, ha=ha)\n",
    "    \n",
    "    plt.xlabel('Epoch', size=16, labelpad=5)\n",
    "    plt.ylabel('Learning Rate', size=16, labelpad=5)\n",
    "    plt.grid()\n",
    "    plt.show()\n",
    "\n",
    "# Learning rate for encoder\n",
    "LR_SCHEDULE = [lrfn(step, num_warmup_steps=N_WARMUP_EPOCHS, lr_max=LR_MAX, num_cycles=0.50) for step in range(N_EPOCHS)]\n",
    "# Plot Learning Rate Schedule\n",
    "plot_lr_schedule(LR_SCHEDULE, epochs=N_EPOCHS)\n",
    "# Learning Rate Callback\n",
    "lr_callback = tf.keras.callbacks.LearningRateScheduler(lambda step: LR_SCHEDULE[step], verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Weight Decay Callback"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "class WeightDecayCallback(tf.keras.callbacks.Callback):\n",
    "    def __init__(self, wd_ratio=WD_RATIO):\n",
    "        self.step_counter = 0\n",
    "        self.wd_ratio = wd_ratio\n",
    "    \n",
    "    def on_epoch_begin(self, epoch, logs=None):\n",
    "        model.optimizer.weight_decay = model.optimizer.learning_rate * self.wd_ratio\n",
    "        print(f'learning rate: {model.optimizer.learning_rate.numpy():.2e}, weight decay: {model.optimizer.weight_decay.numpy():.2e}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 1: LearningRateScheduler setting learning rate to 0.001.\n",
      "learning rate: 1.00e-03, weight decay: 5.00e-05\n",
      "Epoch 1/100\n",
      "94/94 - 63s - loss: 4.2027 - acc: 0.3074 - top_5_acc: 0.5734 - top_10_acc: 0.6754 - lr: 0.0010 - 63s/epoch - 671ms/step\n",
      "\n",
      "Epoch 2: LearningRateScheduler setting learning rate to 0.0009997532801828658.\n",
      "learning rate: 1.00e-03, weight decay: 5.00e-05\n",
      "Epoch 2/100\n",
      "94/94 - 51s - loss: 3.2301 - acc: 0.6054 - top_5_acc: 0.8437 - top_10_acc: 0.8954 - lr: 9.9975e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 3: LearningRateScheduler setting learning rate to 0.0009990133642141358.\n",
      "learning rate: 9.99e-04, weight decay: 5.00e-05\n",
      "Epoch 3/100\n",
      "94/94 - 51s - loss: 2.9735 - acc: 0.6923 - top_5_acc: 0.8923 - top_10_acc: 0.9281 - lr: 9.9901e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 4: LearningRateScheduler setting learning rate to 0.00099778098230154.\n",
      "learning rate: 9.98e-04, weight decay: 4.99e-05\n",
      "Epoch 4/100\n",
      "94/94 - 51s - loss: 2.8335 - acc: 0.7435 - top_5_acc: 0.9141 - top_10_acc: 0.9418 - lr: 9.9778e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 5: LearningRateScheduler setting learning rate to 0.000996057350657239.\n",
      "learning rate: 9.96e-04, weight decay: 4.98e-05\n",
      "Epoch 5/100\n",
      "94/94 - 51s - loss: 2.7497 - acc: 0.7756 - top_5_acc: 0.9272 - top_10_acc: 0.9502 - lr: 9.9606e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 6: LearningRateScheduler setting learning rate to 0.0009938441702975688.\n",
      "learning rate: 9.94e-04, weight decay: 4.97e-05\n",
      "Epoch 6/100\n",
      "94/94 - 51s - loss: 2.6805 - acc: 0.7995 - top_5_acc: 0.9380 - top_10_acc: 0.9577 - lr: 9.9384e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 7: LearningRateScheduler setting learning rate to 0.0009911436253643444.\n",
      "learning rate: 9.91e-04, weight decay: 4.96e-05\n",
      "Epoch 7/100\n",
      "94/94 - 51s - loss: 2.6309 - acc: 0.8196 - top_5_acc: 0.9447 - top_10_acc: 0.9625 - lr: 9.9114e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 8: LearningRateScheduler setting learning rate to 0.0009879583809693738.\n",
      "learning rate: 9.88e-04, weight decay: 4.94e-05\n",
      "Epoch 8/100\n",
      "94/94 - 51s - loss: 2.5812 - acc: 0.8386 - top_5_acc: 0.9520 - top_10_acc: 0.9673 - lr: 9.8796e-04 - 51s/epoch - 538ms/step\n",
      "\n",
      "Epoch 9: LearningRateScheduler setting learning rate to 0.0009842915805643156.\n",
      "learning rate: 9.84e-04, weight decay: 4.92e-05\n",
      "Epoch 9/100\n",
      "94/94 - 51s - loss: 2.5463 - acc: 0.8500 - top_5_acc: 0.9578 - top_10_acc: 0.9717 - lr: 9.8429e-04 - 51s/epoch - 538ms/step\n",
      "\n",
      "Epoch 10: LearningRateScheduler setting learning rate to 0.0009801468428384716.\n",
      "learning rate: 9.80e-04, weight decay: 4.90e-05\n",
      "Epoch 10/100\n",
      "94/94 - 51s - loss: 2.5138 - acc: 0.8623 - top_5_acc: 0.9627 - top_10_acc: 0.9749 - lr: 9.8015e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 11: LearningRateScheduler setting learning rate to 0.0009755282581475768.\n",
      "learning rate: 9.76e-04, weight decay: 4.88e-05\n",
      "Epoch 11/100\n",
      "94/94 - 51s - loss: 2.4872 - acc: 0.8748 - top_5_acc: 0.9655 - top_10_acc: 0.9770 - lr: 9.7553e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 12: LearningRateScheduler setting learning rate to 0.0009704403844771128.\n",
      "learning rate: 9.70e-04, weight decay: 4.85e-05\n",
      "Epoch 12/100\n",
      "94/94 - 51s - loss: 2.4577 - acc: 0.8848 - top_5_acc: 0.9702 - top_10_acc: 0.9799 - lr: 9.7044e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 13: LearningRateScheduler setting learning rate to 0.0009648882429441257.\n",
      "learning rate: 9.65e-04, weight decay: 4.82e-05\n",
      "Epoch 13/100\n",
      "94/94 - 51s - loss: 2.4290 - acc: 0.8950 - top_5_acc: 0.9735 - top_10_acc: 0.9824 - lr: 9.6489e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 14: LearningRateScheduler setting learning rate to 0.0009588773128419905.\n",
      "learning rate: 9.59e-04, weight decay: 4.79e-05\n",
      "Epoch 14/100\n",
      "94/94 - 51s - loss: 2.4052 - acc: 0.9042 - top_5_acc: 0.9770 - top_10_acc: 0.9848 - lr: 9.5888e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 15: LearningRateScheduler setting learning rate to 0.0009524135262330098.\n",
      "learning rate: 9.52e-04, weight decay: 4.76e-05\n",
      "Epoch 15/100\n",
      "94/94 - 51s - loss: 2.3819 - acc: 0.9116 - top_5_acc: 0.9801 - top_10_acc: 0.9872 - lr: 9.5241e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 16: LearningRateScheduler setting learning rate to 0.0009455032620941839.\n",
      "learning rate: 9.46e-04, weight decay: 4.73e-05\n",
      "Epoch 16/100\n",
      "94/94 - 51s - loss: 2.3608 - acc: 0.9198 - top_5_acc: 0.9817 - top_10_acc: 0.9883 - lr: 9.4550e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 17: LearningRateScheduler setting learning rate to 0.0009381533400219318.\n",
      "learning rate: 9.38e-04, weight decay: 4.69e-05\n",
      "Epoch 17/100\n",
      "94/94 - 51s - loss: 2.3434 - acc: 0.9255 - top_5_acc: 0.9843 - top_10_acc: 0.9897 - lr: 9.3815e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 18: LearningRateScheduler setting learning rate to 0.0009303710135019718.\n",
      "learning rate: 9.30e-04, weight decay: 4.65e-05\n",
      "Epoch 18/100\n",
      "94/94 - 51s - loss: 2.3239 - acc: 0.9314 - top_5_acc: 0.9858 - top_10_acc: 0.9913 - lr: 9.3037e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 19: LearningRateScheduler setting learning rate to 0.0009221639627510075.\n",
      "learning rate: 9.22e-04, weight decay: 4.61e-05\n",
      "Epoch 19/100\n",
      "94/94 - 51s - loss: 2.3085 - acc: 0.9378 - top_5_acc: 0.9881 - top_10_acc: 0.9925 - lr: 9.2216e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 20: LearningRateScheduler setting learning rate to 0.0009135402871372809.\n",
      "learning rate: 9.14e-04, weight decay: 4.57e-05\n",
      "Epoch 20/100\n",
      "94/94 - 51s - loss: 2.2897 - acc: 0.9436 - top_5_acc: 0.9892 - top_10_acc: 0.9931 - lr: 9.1354e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 21: LearningRateScheduler setting learning rate to 0.0009045084971874737.\n",
      "learning rate: 9.05e-04, weight decay: 4.52e-05\n",
      "Epoch 21/100\n",
      "94/94 - 51s - loss: 2.2790 - acc: 0.9474 - top_5_acc: 0.9905 - top_10_acc: 0.9944 - lr: 9.0451e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 22: LearningRateScheduler setting learning rate to 0.0008950775061878452.\n",
      "learning rate: 8.95e-04, weight decay: 4.48e-05\n",
      "Epoch 22/100\n",
      "94/94 - 51s - loss: 2.2673 - acc: 0.9526 - top_5_acc: 0.9916 - top_10_acc: 0.9947 - lr: 8.9508e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 23: LearningRateScheduler setting learning rate to 0.0008852566213878947.\n",
      "learning rate: 8.85e-04, weight decay: 4.43e-05\n",
      "Epoch 23/100\n",
      "94/94 - 51s - loss: 2.2584 - acc: 0.9545 - top_5_acc: 0.9922 - top_10_acc: 0.9954 - lr: 8.8526e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 24: LearningRateScheduler setting learning rate to 0.0008750555348152298.\n",
      "learning rate: 8.75e-04, weight decay: 4.38e-05\n",
      "Epoch 24/100\n",
      "94/94 - 51s - loss: 2.2478 - acc: 0.9584 - top_5_acc: 0.9934 - top_10_acc: 0.9960 - lr: 8.7506e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 25: LearningRateScheduler setting learning rate to 0.0008644843137107057.\n",
      "learning rate: 8.64e-04, weight decay: 4.32e-05\n",
      "Epoch 25/100\n",
      "94/94 - 51s - loss: 2.2561 - acc: 0.9563 - top_5_acc: 0.9932 - top_10_acc: 0.9959 - lr: 8.6448e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 26: LearningRateScheduler setting learning rate to 0.0008535533905932737.\n",
      "learning rate: 8.54e-04, weight decay: 4.27e-05\n",
      "Epoch 26/100\n",
      "94/94 - 51s - loss: 2.2451 - acc: 0.9592 - top_5_acc: 0.9940 - top_10_acc: 0.9967 - lr: 8.5355e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 27: LearningRateScheduler setting learning rate to 0.0008422735529643444.\n",
      "learning rate: 8.42e-04, weight decay: 4.21e-05\n",
      "Epoch 27/100\n",
      "94/94 - 51s - loss: 2.2277 - acc: 0.9646 - top_5_acc: 0.9947 - top_10_acc: 0.9970 - lr: 8.4227e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 28: LearningRateScheduler setting learning rate to 0.0008306559326618259.\n",
      "learning rate: 8.31e-04, weight decay: 4.15e-05\n",
      "Epoch 28/100\n",
      "94/94 - 51s - loss: 2.2154 - acc: 0.9678 - top_5_acc: 0.9955 - top_10_acc: 0.9978 - lr: 8.3066e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 29: LearningRateScheduler setting learning rate to 0.0008187119948743449.\n",
      "learning rate: 8.19e-04, weight decay: 4.09e-05\n",
      "Epoch 29/100\n",
      "94/94 - 51s - loss: 2.2106 - acc: 0.9690 - top_5_acc: 0.9962 - top_10_acc: 0.9979 - lr: 8.1871e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 30: LearningRateScheduler setting learning rate to 0.0008064535268264883.\n",
      "learning rate: 8.06e-04, weight decay: 4.03e-05\n",
      "Epoch 30/100\n",
      "94/94 - 51s - loss: 2.2070 - acc: 0.9698 - top_5_acc: 0.9961 - top_10_acc: 0.9980 - lr: 8.0645e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 31: LearningRateScheduler setting learning rate to 0.0007938926261462366.\n",
      "learning rate: 7.94e-04, weight decay: 3.97e-05\n",
      "Epoch 31/100\n",
      "94/94 - 51s - loss: 2.1935 - acc: 0.9743 - top_5_acc: 0.9968 - top_10_acc: 0.9985 - lr: 7.9389e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 32: LearningRateScheduler setting learning rate to 0.0007810416889260653.\n",
      "learning rate: 7.81e-04, weight decay: 3.91e-05\n",
      "Epoch 32/100\n",
      "94/94 - 51s - loss: 2.1832 - acc: 0.9775 - top_5_acc: 0.9973 - top_10_acc: 0.9988 - lr: 7.8104e-04 - 51s/epoch - 538ms/step\n",
      "\n",
      "Epoch 33: LearningRateScheduler setting learning rate to 0.0007679133974894983.\n",
      "learning rate: 7.68e-04, weight decay: 3.84e-05\n",
      "Epoch 33/100\n",
      "94/94 - 51s - loss: 2.1784 - acc: 0.9776 - top_5_acc: 0.9978 - top_10_acc: 0.9989 - lr: 7.6791e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 34: LearningRateScheduler setting learning rate to 0.0007545207078751857.\n",
      "learning rate: 7.55e-04, weight decay: 3.77e-05\n",
      "Epoch 34/100\n",
      "94/94 - 51s - loss: 2.1690 - acc: 0.9805 - top_5_acc: 0.9982 - top_10_acc: 0.9991 - lr: 7.5452e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 35: LearningRateScheduler setting learning rate to 0.0007408768370508576.\n",
      "learning rate: 7.41e-04, weight decay: 3.70e-05\n",
      "Epoch 35/100\n",
      "94/94 - 51s - loss: 2.1633 - acc: 0.9822 - top_5_acc: 0.9984 - top_10_acc: 0.9993 - lr: 7.4088e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 36: LearningRateScheduler setting learning rate to 0.0007269952498697733.\n",
      "learning rate: 7.27e-04, weight decay: 3.63e-05\n",
      "Epoch 36/100\n",
      "94/94 - 51s - loss: 2.1537 - acc: 0.9843 - top_5_acc: 0.9989 - top_10_acc: 0.9996 - lr: 7.2700e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 37: LearningRateScheduler setting learning rate to 0.0007128896457825364.\n",
      "learning rate: 7.13e-04, weight decay: 3.56e-05\n",
      "Epoch 37/100\n",
      "94/94 - 51s - loss: 2.1511 - acc: 0.9848 - top_5_acc: 0.9986 - top_10_acc: 0.9995 - lr: 7.1289e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 38: LearningRateScheduler setting learning rate to 0.0006985739453173903.\n",
      "learning rate: 6.99e-04, weight decay: 3.49e-05\n",
      "Epoch 38/100\n",
      "94/94 - 51s - loss: 2.1403 - acc: 0.9874 - top_5_acc: 0.9990 - top_10_acc: 0.9996 - lr: 6.9857e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 39: LearningRateScheduler setting learning rate to 0.0006840622763423391.\n",
      "learning rate: 6.84e-04, weight decay: 3.42e-05\n",
      "Epoch 39/100\n",
      "94/94 - 51s - loss: 2.1365 - acc: 0.9878 - top_5_acc: 0.9993 - top_10_acc: 0.9998 - lr: 6.8406e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 40: LearningRateScheduler setting learning rate to 0.0006693689601226458.\n",
      "learning rate: 6.69e-04, weight decay: 3.35e-05\n",
      "Epoch 40/100\n",
      "94/94 - 51s - loss: 2.1298 - acc: 0.9888 - top_5_acc: 0.9993 - top_10_acc: 0.9996 - lr: 6.6937e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 41: LearningRateScheduler setting learning rate to 0.0006545084971874737.\n",
      "learning rate: 6.55e-04, weight decay: 3.27e-05\n",
      "Epoch 41/100\n",
      "94/94 - 51s - loss: 2.1280 - acc: 0.9896 - top_5_acc: 0.9994 - top_10_acc: 0.9997 - lr: 6.5451e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 42: LearningRateScheduler setting learning rate to 0.0006394955530196147.\n",
      "learning rate: 6.39e-04, weight decay: 3.20e-05\n",
      "Epoch 42/100\n",
      "94/94 - 51s - loss: 2.1203 - acc: 0.9910 - top_5_acc: 0.9995 - top_10_acc: 0.9998 - lr: 6.3950e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 43: LearningRateScheduler setting learning rate to 0.0006243449435824276.\n",
      "learning rate: 6.24e-04, weight decay: 3.12e-05\n",
      "Epoch 43/100\n",
      "94/94 - 51s - loss: 2.1104 - acc: 0.9929 - top_5_acc: 0.9995 - top_10_acc: 0.9998 - lr: 6.2434e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 44: LearningRateScheduler setting learning rate to 0.0006090716206982714.\n",
      "learning rate: 6.09e-04, weight decay: 3.05e-05\n",
      "Epoch 44/100\n",
      "94/94 - 51s - loss: 2.1030 - acc: 0.9939 - top_5_acc: 0.9998 - top_10_acc: 0.9999 - lr: 6.0907e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 45: LearningRateScheduler setting learning rate to 0.0005936906572928624.\n",
      "learning rate: 5.94e-04, weight decay: 2.97e-05\n",
      "Epoch 45/100\n",
      "94/94 - 51s - loss: 2.1042 - acc: 0.9934 - top_5_acc: 0.9998 - top_10_acc: 0.9999 - lr: 5.9369e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 46: LearningRateScheduler setting learning rate to 0.0005782172325201155.\n",
      "learning rate: 5.78e-04, weight decay: 2.89e-05\n",
      "Epoch 46/100\n",
      "94/94 - 51s - loss: 2.0984 - acc: 0.9944 - top_5_acc: 0.9998 - top_10_acc: 0.9999 - lr: 5.7822e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 47: LearningRateScheduler setting learning rate to 0.0005626666167821521.\n",
      "learning rate: 5.63e-04, weight decay: 2.81e-05\n",
      "Epoch 47/100\n",
      "94/94 - 51s - loss: 2.0933 - acc: 0.9952 - top_5_acc: 0.9999 - top_10_acc: 1.0000 - lr: 5.6267e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 48: LearningRateScheduler setting learning rate to 0.0005470541566592572.\n",
      "learning rate: 5.47e-04, weight decay: 2.74e-05\n",
      "Epoch 48/100\n",
      "94/94 - 51s - loss: 2.0904 - acc: 0.9953 - top_5_acc: 0.9999 - top_10_acc: 1.0000 - lr: 5.4705e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 49: LearningRateScheduler setting learning rate to 0.0005313952597646568.\n",
      "learning rate: 5.31e-04, weight decay: 2.66e-05\n",
      "Epoch 49/100\n",
      "94/94 - 51s - loss: 2.0880 - acc: 0.9955 - top_5_acc: 0.9999 - top_10_acc: 1.0000 - lr: 5.3140e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 50: LearningRateScheduler setting learning rate to 0.0005157053795390641.\n",
      "learning rate: 5.16e-04, weight decay: 2.58e-05\n",
      "Epoch 50/100\n",
      "94/94 - 51s - loss: 2.0840 - acc: 0.9961 - top_5_acc: 0.9999 - top_10_acc: 1.0000 - lr: 5.1571e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 51: LearningRateScheduler setting learning rate to 0.0005.\n",
      "learning rate: 5.00e-04, weight decay: 2.50e-05\n",
      "Epoch 51/100\n",
      "94/94 - 51s - loss: 2.0769 - acc: 0.9971 - top_5_acc: 0.9999 - top_10_acc: 1.0000 - lr: 5.0000e-04 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 52: LearningRateScheduler setting learning rate to 0.00048429462046093585.\n",
      "learning rate: 4.84e-04, weight decay: 2.42e-05\n",
      "Epoch 52/100\n",
      "94/94 - 51s - loss: 2.0745 - acc: 0.9968 - top_5_acc: 0.9999 - top_10_acc: 1.0000 - lr: 4.8429e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 53: LearningRateScheduler setting learning rate to 0.0004686047402353433.\n",
      "learning rate: 4.69e-04, weight decay: 2.34e-05\n",
      "Epoch 53/100\n",
      "94/94 - 51s - loss: 2.0726 - acc: 0.9974 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 4.6860e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 54: LearningRateScheduler setting learning rate to 0.00045294584334074284.\n",
      "learning rate: 4.53e-04, weight decay: 2.26e-05\n",
      "Epoch 54/100\n",
      "94/94 - 51s - loss: 2.0667 - acc: 0.9979 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 4.5295e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 55: LearningRateScheduler setting learning rate to 0.00043733338321784784.\n",
      "learning rate: 4.37e-04, weight decay: 2.19e-05\n",
      "Epoch 55/100\n",
      "94/94 - 51s - loss: 2.0640 - acc: 0.9980 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 4.3733e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 56: LearningRateScheduler setting learning rate to 0.0004217827674798845.\n",
      "learning rate: 4.22e-04, weight decay: 2.11e-05\n",
      "Epoch 56/100\n",
      "94/94 - 51s - loss: 2.0613 - acc: 0.9983 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 4.2178e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 57: LearningRateScheduler setting learning rate to 0.0004063093427071376.\n",
      "learning rate: 4.06e-04, weight decay: 2.03e-05\n",
      "Epoch 57/100\n",
      "94/94 - 51s - loss: 2.0587 - acc: 0.9984 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 4.0631e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 58: LearningRateScheduler setting learning rate to 0.0003909283793017289.\n",
      "learning rate: 3.91e-04, weight decay: 1.95e-05\n",
      "Epoch 58/100\n",
      "94/94 - 51s - loss: 2.0541 - acc: 0.9986 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.9093e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 59: LearningRateScheduler setting learning rate to 0.0003756550564175727.\n",
      "learning rate: 3.76e-04, weight decay: 1.88e-05\n",
      "Epoch 59/100\n",
      "94/94 - 51s - loss: 2.0508 - acc: 0.9990 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.7566e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 60: LearningRateScheduler setting learning rate to 0.0003605044469803854.\n",
      "learning rate: 3.61e-04, weight decay: 1.80e-05\n",
      "Epoch 60/100\n",
      "94/94 - 51s - loss: 2.0487 - acc: 0.9988 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.6050e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 61: LearningRateScheduler setting learning rate to 0.00034549150281252633.\n",
      "learning rate: 3.45e-04, weight decay: 1.73e-05\n",
      "Epoch 61/100\n",
      "94/94 - 51s - loss: 2.0442 - acc: 0.9992 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.4549e-04 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 62: LearningRateScheduler setting learning rate to 0.0003306310398773543.\n",
      "learning rate: 3.31e-04, weight decay: 1.65e-05\n",
      "Epoch 62/100\n",
      "94/94 - 51s - loss: 2.0437 - acc: 0.9991 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.3063e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 63: LearningRateScheduler setting learning rate to 0.00031593772365766105.\n",
      "learning rate: 3.16e-04, weight decay: 1.58e-05\n",
      "Epoch 63/100\n",
      "94/94 - 51s - loss: 2.0392 - acc: 0.9995 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.1594e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 64: LearningRateScheduler setting learning rate to 0.00030142605468260977.\n",
      "learning rate: 3.01e-04, weight decay: 1.51e-05\n",
      "Epoch 64/100\n",
      "94/94 - 51s - loss: 2.0378 - acc: 0.9993 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.0143e-04 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 65: LearningRateScheduler setting learning rate to 0.00028711035421746366.\n",
      "learning rate: 2.87e-04, weight decay: 1.44e-05\n",
      "Epoch 65/100\n",
      "94/94 - 51s - loss: 2.0354 - acc: 0.9995 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.8711e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 66: LearningRateScheduler setting learning rate to 0.00027300475013022663.\n",
      "learning rate: 2.73e-04, weight decay: 1.37e-05\n",
      "Epoch 66/100\n",
      "94/94 - 51s - loss: 2.0332 - acc: 0.9997 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.7300e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 67: LearningRateScheduler setting learning rate to 0.0002591231629491423.\n",
      "learning rate: 2.59e-04, weight decay: 1.30e-05\n",
      "Epoch 67/100\n",
      "94/94 - 51s - loss: 2.0318 - acc: 0.9996 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.5912e-04 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 68: LearningRateScheduler setting learning rate to 0.00024547929212481435.\n",
      "learning rate: 2.45e-04, weight decay: 1.23e-05\n",
      "Epoch 68/100\n",
      "94/94 - 51s - loss: 2.0305 - acc: 0.9996 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.4548e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 69: LearningRateScheduler setting learning rate to 0.00023208660251050156.\n",
      "learning rate: 2.32e-04, weight decay: 1.16e-05\n",
      "Epoch 69/100\n",
      "94/94 - 51s - loss: 2.0284 - acc: 0.9996 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.3209e-04 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 70: LearningRateScheduler setting learning rate to 0.0002189583110739348.\n",
      "learning rate: 2.19e-04, weight decay: 1.09e-05\n",
      "Epoch 70/100\n",
      "94/94 - 51s - loss: 2.0254 - acc: 0.9997 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.1896e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 71: LearningRateScheduler setting learning rate to 0.00020610737385376348.\n",
      "learning rate: 2.06e-04, weight decay: 1.03e-05\n",
      "Epoch 71/100\n",
      "94/94 - 51s - loss: 2.0245 - acc: 0.9997 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.0611e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 72: LearningRateScheduler setting learning rate to 0.00019354647317351188.\n",
      "learning rate: 1.94e-04, weight decay: 9.68e-06\n",
      "Epoch 72/100\n",
      "94/94 - 51s - loss: 2.0231 - acc: 0.9997 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.9355e-04 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 73: LearningRateScheduler setting learning rate to 0.00018128800512565513.\n",
      "learning rate: 1.81e-04, weight decay: 9.06e-06\n",
      "Epoch 73/100\n",
      "94/94 - 51s - loss: 2.0212 - acc: 0.9998 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.8129e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 74: LearningRateScheduler setting learning rate to 0.00016934406733817414.\n",
      "learning rate: 1.69e-04, weight decay: 8.47e-06\n",
      "Epoch 74/100\n",
      "94/94 - 51s - loss: 2.0198 - acc: 0.9998 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.6934e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 75: LearningRateScheduler setting learning rate to 0.00015772644703565563.\n",
      "learning rate: 1.58e-04, weight decay: 7.89e-06\n",
      "Epoch 75/100\n",
      "94/94 - 51s - loss: 2.0173 - acc: 0.9998 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.5773e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 76: LearningRateScheduler setting learning rate to 0.00014644660940672628.\n",
      "learning rate: 1.46e-04, weight decay: 7.32e-06\n",
      "Epoch 76/100\n",
      "94/94 - 51s - loss: 2.0162 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.4645e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 77: LearningRateScheduler setting learning rate to 0.00013551568628929433.\n",
      "learning rate: 1.36e-04, weight decay: 6.78e-06\n",
      "Epoch 77/100\n",
      "94/94 - 51s - loss: 2.0150 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.3552e-04 - 51s/epoch - 542ms/step\n",
      "\n",
      "Epoch 78: LearningRateScheduler setting learning rate to 0.0001249444651847702.\n",
      "learning rate: 1.25e-04, weight decay: 6.25e-06\n",
      "Epoch 78/100\n",
      "94/94 - 51s - loss: 2.0135 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.2494e-04 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 79: LearningRateScheduler setting learning rate to 0.00011474337861210544.\n",
      "learning rate: 1.15e-04, weight decay: 5.74e-06\n",
      "Epoch 79/100\n",
      "94/94 - 51s - loss: 2.0128 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.1474e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 80: LearningRateScheduler setting learning rate to 0.00010492249381215479.\n",
      "learning rate: 1.05e-04, weight decay: 5.25e-06\n",
      "Epoch 80/100\n",
      "94/94 - 51s - loss: 2.0113 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.0492e-04 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 81: LearningRateScheduler setting learning rate to 9.549150281252633e-05.\n",
      "learning rate: 9.55e-05, weight decay: 4.77e-06\n",
      "Epoch 81/100\n",
      "94/94 - 51s - loss: 2.0108 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 9.5492e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 82: LearningRateScheduler setting learning rate to 8.645971286271903e-05.\n",
      "learning rate: 8.65e-05, weight decay: 4.32e-06\n",
      "Epoch 82/100\n",
      "94/94 - 51s - loss: 2.0094 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 8.6460e-05 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 83: LearningRateScheduler setting learning rate to 7.783603724899258e-05.\n",
      "learning rate: 7.78e-05, weight decay: 3.89e-06\n",
      "Epoch 83/100\n",
      "94/94 - 51s - loss: 2.0087 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 7.7836e-05 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 84: LearningRateScheduler setting learning rate to 6.962898649802824e-05.\n",
      "learning rate: 6.96e-05, weight decay: 3.48e-06\n",
      "Epoch 84/100\n",
      "94/94 - 51s - loss: 2.0078 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 6.9629e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 85: LearningRateScheduler setting learning rate to 6.184665997806832e-05.\n",
      "learning rate: 6.18e-05, weight decay: 3.09e-06\n",
      "Epoch 85/100\n",
      "94/94 - 51s - loss: 2.0071 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 6.1847e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 86: LearningRateScheduler setting learning rate to 5.449673790581611e-05.\n",
      "learning rate: 5.45e-05, weight decay: 2.72e-06\n",
      "Epoch 86/100\n",
      "94/94 - 51s - loss: 2.0061 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 5.4497e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 87: LearningRateScheduler setting learning rate to 4.758647376699032e-05.\n",
      "learning rate: 4.76e-05, weight decay: 2.38e-06\n",
      "Epoch 87/100\n",
      "94/94 - 51s - loss: 2.0061 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 4.7586e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 88: LearningRateScheduler setting learning rate to 4.112268715800943e-05.\n",
      "learning rate: 4.11e-05, weight decay: 2.06e-06\n",
      "Epoch 88/100\n",
      "94/94 - 51s - loss: 2.0052 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 4.1123e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 89: LearningRateScheduler setting learning rate to 3.5111757055874326e-05.\n",
      "learning rate: 3.51e-05, weight decay: 1.76e-06\n",
      "Epoch 89/100\n",
      "94/94 - 51s - loss: 2.0043 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.5112e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 90: LearningRateScheduler setting learning rate to 2.9559615522887274e-05.\n",
      "learning rate: 2.96e-05, weight decay: 1.48e-06\n",
      "Epoch 90/100\n",
      "94/94 - 51s - loss: 2.0041 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.9560e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 91: LearningRateScheduler setting learning rate to 2.4471741852423235e-05.\n",
      "learning rate: 2.45e-05, weight decay: 1.22e-06\n",
      "Epoch 91/100\n",
      "94/94 - 51s - loss: 2.0035 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.4472e-05 - 51s/epoch - 539ms/step\n",
      "\n",
      "Epoch 92: LearningRateScheduler setting learning rate to 1.985315716152847e-05.\n",
      "learning rate: 1.99e-05, weight decay: 9.93e-07\n",
      "Epoch 92/100\n",
      "94/94 - 51s - loss: 2.0037 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.9853e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 93: LearningRateScheduler setting learning rate to 1.5708419435684463e-05.\n",
      "learning rate: 1.57e-05, weight decay: 7.85e-07\n",
      "Epoch 93/100\n",
      "94/94 - 51s - loss: 2.0030 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.5708e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 94: LearningRateScheduler setting learning rate to 1.2041619030626282e-05.\n",
      "learning rate: 1.20e-05, weight decay: 6.02e-07\n",
      "Epoch 94/100\n",
      "94/94 - 51s - loss: 2.0027 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 1.2042e-05 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 95: LearningRateScheduler setting learning rate to 8.856374635655695e-06.\n",
      "learning rate: 8.86e-06, weight decay: 4.43e-07\n",
      "Epoch 95/100\n",
      "94/94 - 51s - loss: 2.0024 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 8.8564e-06 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 96: LearningRateScheduler setting learning rate to 6.15582970243117e-06.\n",
      "learning rate: 6.16e-06, weight decay: 3.08e-07\n",
      "Epoch 96/100\n",
      "94/94 - 51s - loss: 2.0020 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 6.1558e-06 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 97: LearningRateScheduler setting learning rate to 3.942649342761117e-06.\n",
      "learning rate: 3.94e-06, weight decay: 1.97e-07\n",
      "Epoch 97/100\n",
      "94/94 - 51s - loss: 2.0021 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 3.9426e-06 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 98: LearningRateScheduler setting learning rate to 2.219017698460002e-06.\n",
      "learning rate: 2.22e-06, weight decay: 1.11e-07\n",
      "Epoch 98/100\n",
      "94/94 - 51s - loss: 2.0022 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.2190e-06 - 51s/epoch - 541ms/step\n",
      "\n",
      "Epoch 99: LearningRateScheduler setting learning rate to 9.866357858642206e-07.\n",
      "learning rate: 9.87e-07, weight decay: 4.93e-08\n",
      "Epoch 99/100\n",
      "94/94 - 51s - loss: 2.0019 - acc: 1.0000 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 9.8664e-07 - 51s/epoch - 540ms/step\n",
      "\n",
      "Epoch 100: LearningRateScheduler setting learning rate to 2.467198171342e-07.\n",
      "learning rate: 2.47e-07, weight decay: 1.23e-08\n",
      "Epoch 100/100\n",
      "94/94 - 51s - loss: 2.0022 - acc: 0.9999 - top_5_acc: 1.0000 - top_10_acc: 1.0000 - lr: 2.4672e-07 - 51s/epoch - 541ms/step\n",
      "\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(\n",
    "            x=get_train_batch_all_signs(X_train, y_train, NON_EMPTY_FRAME_IDXS_TRAIN),\n",
    "            steps_per_epoch=len(X_train) // (NUM_CLASSES * BATCH_ALL_SIGNS_N),\n",
    "            epochs=N_EPOCHS,\n",
    "            # Only used for validation data since training data is a generator\n",
    "            batch_size=BATCH_SIZE,\n",
    "            callbacks=[\n",
    "                lr_callback,\n",
    "                WeightDecayCallback(),\n",
    "            ],\n",
    "            verbose = VERBOSE,\n",
    "        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Accuracy and Loss Curves"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/jpeg": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "plt.figure(figsize=(12, 6))\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.plot(history.history['accuracy'], label='Training accuracy')\n",
    "plt.plot(history.history['val_accuracy'], label='Validation accuracy')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.legend()\n",
    "\n",
    "# Plotting training and validation loss vs epochs\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.plot(history.history['loss'], label='Training loss')\n",
    "plt.plot(history.history['val_loss'], label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Making TFLite Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class TFLiteModel(tf.Module):\n",
    "    def __init__(self, model):\n",
    "        super(TFLiteModel, self).__init__()\n",
    "\n",
    "        # Load the feature generation and main models\n",
    "        self.preprocess_layer = preprocess_layer\n",
    "        self.model = model\n",
    "    \n",
    "    @tf.function(input_signature=[tf.TensorSpec(shape=[None, N_ROWS, N_DIMS], dtype=tf.float32, name='inputs')])\n",
    "    def __call__(self, inputs):\n",
    "        # Preprocess Data\n",
    "        x, non_empty_frame_idxs = self.preprocess_layer(inputs)\n",
    "        # Add Batch Dimension\n",
    "        x = tf.expand_dims(x, axis=0)\n",
    "        non_empty_frame_idxs = tf.expand_dims(non_empty_frame_idxs, axis=0)\n",
    "        # Make Prediction\n",
    "        outputs = self.model({ 'frames': x, 'non_empty_frame_idxs': non_empty_frame_idxs })\n",
    "        # Squeeze Output 1x250 -> 250\n",
    "        outputs = tf.squeeze(outputs, axis=0)\n",
    "\n",
    "        # Return a dictionary with the output tensor\n",
    "        return {'outputs': outputs}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tflite_keras_model = TFLiteModel(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras_model_converter = tf.lite.TFLiteConverter.from_keras_model(tflite_keras_model)\n",
    "# Convert Model\n",
    "tflite_model = keras_model_converter.convert()\n",
    "# Write Model\n",
    "with open('model.tflite', 'wb') as f:\n",
    "    f.write(tflite_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Testing Environment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "objc[20333]: Class CaptureDelegate is implemented in both /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/cv2/cv2.abi3.so (0x311df2620) and /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/mediapipe/.dylibs/libopencv_videoio.3.4.16.dylib (0x138e0c860). One of the two will be used. Which one is undefined.\n",
      "objc[20333]: Class CVWindow is implemented in both /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/cv2/cv2.abi3.so (0x311df2670) and /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/mediapipe/.dylibs/libopencv_highgui.3.4.16.dylib (0x1290b8a68). One of the two will be used. Which one is undefined.\n",
      "objc[20333]: Class CVView is implemented in both /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/cv2/cv2.abi3.so (0x311df2698) and /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/mediapipe/.dylibs/libopencv_highgui.3.4.16.dylib (0x1290b8a90). One of the two will be used. Which one is undefined.\n",
      "objc[20333]: Class CVSlider is implemented in both /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/cv2/cv2.abi3.so (0x311df26c0) and /Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/mediapipe/.dylibs/libopencv_highgui.3.4.16.dylib (0x1290b8ab8). One of the two will be used. Which one is undefined.\n",
      "/Users/devasheeshchopra/Desktop/ASL Words/ASL/lib/python3.10/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "import cv2\n",
    "import mediapipe as mp\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import time\n",
    "import tensorflow as tf\n",
    "import google.generativeai as genai"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "interpreter = tf.lite.Interpreter(model_path=\"model.tflite\")\n",
    "found_signatures = list(interpreter.get_signature_list().keys())\n",
    "pred_fn = interpreter.get_signature_runner(\"serving_default\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv(\"train.csv\")\n",
    "\n",
    "train['sign_ord'] = train['sign'].astype('category').cat.codes\n",
    "\n",
    "SIGN2ORD = train[['sign', 'sign_ord']].set_index('sign').squeeze().to_dict()\n",
    "ORD2SIGN = train[['sign_ord', 'sign']].set_index('sign_ord').squeeze().to_dict()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "mp_drawing = mp.solutions.drawing_utils\n",
    "mp_drawing_styles = mp.solutions.drawing_styles\n",
    "mp_holistic = mp.solutions.holistic\n",
    "\n",
    "unique_signs = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_frame_landmark_df(results, frame, xyz):\n",
    "    xyz_skel = (xyz[[\"type\", \"landmark_index\"]].drop_duplicates().reset_index(drop=True).copy())\n",
    "    face = pd.DataFrame()\n",
    "    pose = pd.DataFrame()\n",
    "    left_hand = pd.DataFrame()\n",
    "    right_hand = pd.DataFrame()\n",
    "    \n",
    "    if results.face_landmarks:\n",
    "        for i, point in enumerate(results.face_landmarks.landmark):\n",
    "            face.loc[i, [\"x\", \"y\", \"z\"]] = [point.x, point.y, point.z]\n",
    "    if results.pose_landmarks:\n",
    "        for i, point in enumerate(results.pose_landmarks.landmark):\n",
    "            pose.loc[i, [\"x\", \"y\", \"z\"]] = [point.x, point.y, point.z]\n",
    "    if results.left_hand_landmarks:\n",
    "        for i, point in enumerate(results.left_hand_landmarks.landmark):\n",
    "            left_hand.loc[i, [\"x\", \"y\", \"z\"]] = [point.x, point.y, point.z]\n",
    "    if results.right_hand_landmarks:\n",
    "        for i, point in enumerate(results.right_hand_landmarks.landmark):\n",
    "            right_hand.loc[i, [\"x\", \"y\", \"z\"]] = [point.x, point.y, point.z]\n",
    "    \n",
    "    face = face.reset_index().rename(columns={\"index\": \"landmark_index\"}).assign(type=\"face\")\n",
    "    pose = pose.reset_index().rename(columns={\"index\": \"landmark_index\"}).assign(type=\"pose\")\n",
    "    left_hand = left_hand.reset_index().rename(columns={\"index\": \"landmark_index\"}).assign(type=\"left_hand\")\n",
    "    right_hand = right_hand.reset_index().rename(columns={\"index\": \"landmark_index\"}).assign(type=\"right_hand\")\n",
    "    \n",
    "    landmarks = pd.concat([face, pose, left_hand, right_hand]).reset_index(drop=True)\n",
    "    landmarks = xyz_skel.merge(landmarks, on=[\"type\", \"landmark_index\"], how=\"left\").assign(frame=frame)\n",
    "    \n",
    "    return landmarks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dotenv import load_dotenv\n",
    "import os\n",
    "load_dotenv()\n",
    "\n",
    "def get_display_message_from_api(recognised_words):\n",
    "\n",
    "    GOOGLE_API_KEY = os.getenv(\"GOOGLE_API_KEY\")\n",
    "    genai.configure(api_key=GOOGLE_API_KEY) \n",
    "    \n",
    "    model = genai.GenerativeModel('gemini-pro')\n",
    "    \n",
    "    prompt = f\"\"\"\n",
    "            Objective:\n",
    "            You have developed an isolated American Sign Language (ASL) word recognition model. At the end of each run, the model stores the recognized words in a list. However, the words may not necessarily be in the correct order. Your objective is to utilize these recognized words to construct a coherent and meaningful English sentence. The resulting sentence should be as simple as possible while still accurately conveying the intended meaning.\n",
    "\n",
    "            Instructions:\n",
    "\n",
    "            - Input: You will be provided with a Python list containing the recognized ASL words from your model. The contents of this list may vary depending on the output of your model.\n",
    "            - Processing: Rearrange the words in the list to form a grammatically correct and logically relevant English sentence. Take into consideration the context and logical flow of the sentence. Always ignore the word \"TV\".\n",
    "            - Output: Generate a concise English sentence that accurately conveys the meaning of the recognized ASL words.\n",
    "\n",
    "            Considerations:\n",
    "\n",
    "            - Simplicity: Aim for simplicity in your sentence structure and vocabulary.\n",
    "            - Clarity: Ensure that the sentence is clear and understandable.\n",
    "            - Relevance: The sentence should reflect the meaning conveyed by the ASL words.\n",
    "            - Grammar: Maintain proper grammar and syntax in the sentence.\n",
    "\n",
    "            Example:\n",
    "\n",
    "            Input: recognized_words = cat mat\n",
    "            output: cat on the mat\n",
    "\n",
    "            Here is the actual input for which you have to produce the relevant output: recognised_words = {' '.join(recognised_words)}\n",
    "            \"\"\"\n",
    "    \n",
    "    response = model.generate_content(prompt)\n",
    "    \n",
    "    return response.text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_relevant_data_subset(pq_path):\n",
    "    data_columns = ['x', 'y', 'z']\n",
    "    data = pd.read_parquet(pq_path, columns=data_columns)\n",
    "    ROWS_PER_FRAME = 543\n",
    "    n_frames = int(len(data) / ROWS_PER_FRAME)\n",
    "    data = data.values.reshape(n_frames, ROWS_PER_FRAME, len(data_columns))\n",
    "    return data.astype(np.float32)\n",
    "\n",
    "def do_capture_loop(xyz, pred_fn):\n",
    "    all_landmarks = []\n",
    "    cap = cv2.VideoCapture(0)\n",
    "    start_time = time.time()\n",
    "    font = cv2.FONT_HERSHEY_SIMPLEX\n",
    "    last_prediction_time = -5\n",
    "    escape_pressed = False\n",
    "    display_message = \"Press Escape to toggle message display\"\n",
    "    unique_signs = []\n",
    "    sign_name = \"\"\n",
    "\n",
    "    with mp_holistic.Holistic(min_detection_confidence=0.5, min_tracking_confidence=0.5) as holistic:\n",
    "        while cap.isOpened():\n",
    "            current_time = time.time()\n",
    "            elapsed_time = int(current_time - start_time)\n",
    "\n",
    "            success, image = cap.read()\n",
    "            if not success:\n",
    "                print(\"Ignoring empty camera frame.\")\n",
    "                continue\n",
    "\n",
    "            image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)\n",
    "            results = holistic.process(image)\n",
    "            landmarks = create_frame_landmark_df(results, elapsed_time, xyz)\n",
    "            all_landmarks.append(landmarks)\n",
    "\n",
    "            image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)\n",
    "            \n",
    "            mp_drawing.draw_landmarks(\n",
    "                image,\n",
    "                results.face_landmarks,\n",
    "                mp_holistic.FACEMESH_CONTOURS,\n",
    "                landmark_drawing_spec=None,\n",
    "                connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_contours_style()\n",
    "            )\n",
    "            mp_drawing.draw_landmarks(\n",
    "                image,\n",
    "                results.pose_landmarks,\n",
    "                mp_holistic.POSE_CONNECTIONS,\n",
    "                landmark_drawing_spec=mp_drawing_styles.get_default_pose_landmarks_style()\n",
    "            )\n",
    "\n",
    "            if current_time - last_prediction_time >= 5:\n",
    "                if all_landmarks:\n",
    "                    concatenated_landmarks = pd.concat(all_landmarks).reset_index(drop=True)\n",
    "                    concatenated_landmarks.to_parquet(\"out.parquet\")\n",
    "                    xyz_np = load_relevant_data_subset(\"out.parquet\")\n",
    "                    p = pred_fn(inputs=xyz_np)\n",
    "                    sign = p['outputs'].argmax()\n",
    "                    sign_name = ORD2SIGN[sign]\n",
    "                    pc = p['outputs'][sign]\n",
    "                    if sign_name not in unique_signs:\n",
    "                        unique_signs.append(sign_name)\n",
    "\n",
    "                    last_prediction_time = current_time\n",
    "                    all_landmarks = []  # Reset landmarks\n",
    "\n",
    "            if sign_name == \"\":\n",
    "                sign_name = \"No Movement Detected\"\n",
    "\n",
    "            cv2.putText(image, f\"Sign: {sign_name}\", (10, 80), font, 2, (0, 255, 0), 3, cv2.LINE_AA)\n",
    "            cv2.putText(image, f\"Time: {elapsed_time}s\", (10, 150), font, 1, (0, 255, 0), 2, cv2.LINE_AA)\n",
    "\n",
    "            if escape_pressed:\n",
    "                cv2.putText(image, display_message, (10, 200), font, 1, (0, 0, 255), 2, cv2.LINE_AA)\n",
    "\n",
    "            cv2.imshow(\"MediaPipe Holistic\", image)\n",
    "\n",
    "            key = cv2.waitKey(5)\n",
    "            if key & 0xFF == 27:  # Escape key\n",
    "                escape_pressed = not escape_pressed\n",
    "                display_message = get_display_message_from_api(unique_signs) if escape_pressed else \"Press Escape to toggle message display\"\n",
    "                if escape_pressed: \n",
    "                    unique_signs = []\n",
    "            elif key & 0xFF == ord('q'):\n",
    "                break\n",
    "\n",
    "    cap.release()\n",
    "    cv2.destroyAllWindows()\n",
    "\n",
    "pq_file = \"10042041.parquet\"\n",
    "xyz = pd.read_parquet(pq_file)\n",
    "interpreter = tf.lite.Interpreter(model_path=\"model.tflite\")\n",
    "interpreter.allocate_tensors()\n",
    "pred_fn = interpreter.get_signature_runner(\"serving_default\")\n",
    "\n",
    "landmarks = do_capture_loop(xyz, pred_fn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualisation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.animation import FuncAnimation\n",
    "from IPython.display import HTML"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "path_to_sign = 'output_every_5s.parquet'\n",
    "sign = pd.read_parquet(path_to_sign)\n",
    "sign.y = sign.y * -1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_hand_points(hand):\n",
    "    x = [[hand.iloc[0].x, hand.iloc[1].x, hand.iloc[2].x, hand.iloc[3].x, hand.iloc[4].x], # Thumb\n",
    "         [hand.iloc[5].x, hand.iloc[6].x, hand.iloc[7].x, hand.iloc[8].x], # Index\n",
    "         [hand.iloc[9].x, hand.iloc[10].x, hand.iloc[11].x, hand.iloc[12].x], \n",
    "         [hand.iloc[13].x, hand.iloc[14].x, hand.iloc[15].x, hand.iloc[16].x], \n",
    "         [hand.iloc[17].x, hand.iloc[18].x, hand.iloc[19].x, hand.iloc[20].x], \n",
    "         [hand.iloc[0].x, hand.iloc[5].x, hand.iloc[9].x, hand.iloc[13].x, hand.iloc[17].x, hand.iloc[0].x]]\n",
    "\n",
    "    y = [[hand.iloc[0].y, hand.iloc[1].y, hand.iloc[2].y, hand.iloc[3].y, hand.iloc[4].y],  #Thumb\n",
    "         [hand.iloc[5].y, hand.iloc[6].y, hand.iloc[7].y, hand.iloc[8].y], # Index\n",
    "         [hand.iloc[9].y, hand.iloc[10].y, hand.iloc[11].y, hand.iloc[12].y], \n",
    "         [hand.iloc[13].y, hand.iloc[14].y, hand.iloc[15].y, hand.iloc[16].y], \n",
    "         [hand.iloc[17].y, hand.iloc[18].y, hand.iloc[19].y, hand.iloc[20].y], \n",
    "         [hand.iloc[0].y, hand.iloc[5].y, hand.iloc[9].y, hand.iloc[13].y, hand.iloc[17].y, hand.iloc[0].y]] \n",
    "    return x, y\n",
    "\n",
    "def get_pose_points(pose):\n",
    "    x = [[pose.iloc[8].x, pose.iloc[6].x, pose.iloc[5].x, pose.iloc[4].x, pose.iloc[0].x, pose.iloc[1].x, pose.iloc[2].x, pose.iloc[3].x, pose.iloc[7].x], \n",
    "         [pose.iloc[10].x, pose.iloc[9].x], \n",
    "         [pose.iloc[22].x, pose.iloc[16].x, pose.iloc[20].x, pose.iloc[18].x, pose.iloc[16].x, pose.iloc[14].x, pose.iloc[12].x, \n",
    "          pose.iloc[11].x, pose.iloc[13].x, pose.iloc[15].x, pose.iloc[17].x, pose.iloc[19].x, pose.iloc[15].x, pose.iloc[21].x], \n",
    "         [pose.iloc[12].x, pose.iloc[24].x, pose.iloc[26].x, pose.iloc[28].x, pose.iloc[30].x, pose.iloc[32].x, pose.iloc[28].x], \n",
    "         [pose.iloc[11].x, pose.iloc[23].x, pose.iloc[25].x, pose.iloc[27].x, pose.iloc[29].x, pose.iloc[31].x, pose.iloc[27].x], \n",
    "         [pose.iloc[24].x, pose.iloc[23].x]\n",
    "        ]\n",
    "\n",
    "    y = [[pose.iloc[8].y, pose.iloc[6].y, pose.iloc[5].y, pose.iloc[4].y, pose.iloc[0].y, pose.iloc[1].y, pose.iloc[2].y, pose.iloc[3].y, pose.iloc[7].y], \n",
    "         [pose.iloc[10].y, pose.iloc[9].y], \n",
    "         [pose.iloc[22].y, pose.iloc[16].y, pose.iloc[20].y, pose.iloc[18].y, pose.iloc[16].y, pose.iloc[14].y, pose.iloc[12].y, \n",
    "          pose.iloc[11].y, pose.iloc[13].y, pose.iloc[15].y, pose.iloc[17].y, pose.iloc[19].y, pose.iloc[15].y, pose.iloc[21].y], \n",
    "         [pose.iloc[12].y, pose.iloc[24].y, pose.iloc[26].y, pose.iloc[28].y, pose.iloc[30].y, pose.iloc[32].y, pose.iloc[28].y], \n",
    "         [pose.iloc[11].y, pose.iloc[23].y, pose.iloc[25].y, pose.iloc[27].y, pose.iloc[29].y, pose.iloc[31].y, pose.iloc[27].y], \n",
    "         [pose.iloc[24].y, pose.iloc[23].y]\n",
    "        ]\n",
    "    return x, y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<video width=\"640\" height=\"480\" controls autoplay loop>\n",
       "  <source type=\"video/mp4\" src=\"data:video/mp4;base64,AAAAIGZ0eXBNNFYgAAACAE00ViBpc29taXNvMmF2YzEAAAAIZnJlZQAARXptZGF0AAACrgYF//+q\n",
       "3EXpvebZSLeWLNgg2SPu73gyNjQgLSBjb3JlIDE2NCByMzEwOCAzMWUxOWY5IC0gSC4yNjQvTVBF\n",
       "Ry00IEFWQyBjb2RlYyAtIENvcHlsZWZ0IDIwMDMtMjAyMyAtIGh0dHA6Ly93d3cudmlkZW9sYW4u\n",
       "b3JnL3gyNjQuaHRtbCAtIG9wdGlvbnM6IGNhYmFjPTEgcmVmPTMgZGVibG9jaz0xOjA6MCBhbmFs\n",
       "eXNlPTB4MzoweDExMyBtZT1oZXggc3VibWU9NyBwc3k9MSBwc3lfcmQ9MS4wMDowLjAwIG1peGVk\n",
       "X3JlZj0xIG1lX3JhbmdlPTE2IGNocm9tYV9tZT0xIHRyZWxsaXM9MSA4eDhkY3Q9MSBjcW09MCBk\n",
       "ZWFkem9uZT0yMSwxMSBmYXN0X3Bza2lwPTEgY2hyb21hX3FwX29mZnNldD0tMiB0aHJlYWRzPTE1\n",
       "IGxvb2thaGVhZF90aHJlYWRzPTIgc2xpY2VkX3RocmVhZHM9MCBucj0wIGRlY2ltYXRlPTEgaW50\n",
       "ZXJsYWNlZD0wIGJsdXJheV9jb21wYXQ9MCBjb25zdHJhaW5lZF9pbnRyYT0wIGJmcmFtZXM9MyBi\n",
       "X3B5cmFtaWQ9MiBiX2FkYXB0PTEgYl9iaWFzPTAgZGlyZWN0PTEgd2VpZ2h0Yj0xIG9wZW5fZ29w\n",
       "PTAgd2VpZ2h0cD0yIGtleWludD0yNTAga2V5aW50X21pbj01IHNjZW5lY3V0PTQwIGludHJhX3Jl\n",
       "ZnJlc2g9MCByY19sb29rYWhlYWQ9NDAgcmM9Y3JmIG1idHJlZT0xIGNyZj0yMy4wIHFjb21wPTAu\n",
       "NjAgcXBtaW49MCBxcG1heD02OSBxcHN0ZXA9NCBpcF9yYXRpbz0xLjQwIGFxPTE6MS4wMACAAAAa\n",
       "e2WIhAAQ//73gb8yy18iuslyJX++vumsaJ7TNFM/UPwoQAAAAwAAAwAAJ8DHjw/erLhKcAAAAwJ2\n",
       "AG+b/8Gaf3AR6LWAx6me4dNglb6gksRomdazZI0G9/EZlXLPnztKDT1Mb/2VI3QhnA4owGT2NYb0\n",
       "pGrsoiutuNvqh+01Yh2/t+yP5ptlcPA6euJ7FIIf7DfUrVQHeQECHKVdpuhWRB96WbVGSa32bm2x\n",
       "mWgzJmPFCObKxzn4txJvlx6eTQH2NaZgOBfA69AqQ4GBEHYOtctUsjCpmMTyZYWJDHuNSOj9iD1q\n",
       "VG/NmRrdJ63h0UZquW5Fddip+228uOlQ3GVCkMm+KPsKq4ngWsj4DzuTxnBJRLWdgZdDYPDbLp/g\n",
       "s6QyEwPjnkn5xWekQQ7w9l2CVNmJCyaHaGR5M3pZ6tA3+zyoZTKRp2AJ/mDuDqXv5q1q/UHafBfa\n",
       "N9Fsoo5daSgW7ne/DOouQZDNj28ahcnbwrgu0hzwBaDOehCFDTWA8bVd38lNkM1Fd/N1J4bBgeQH\n",
       "R193wmGS5xx5r/SkulhEwDxU7j9D2efEpea4Ml9Erx1/AGh+O8P4pNW7KxoB/ys+Bg1xrHSY0nvb\n",
       "w09VyK0siLKs28nwytj2R1cnOM8ryt7hMqfcL63yCtlEqBz9M4jxOp1ig7xCDhFsWXUnlNbHq1++\n",
       "JOXUD01+/RZUSj90DrZCctnNdtDqsWdg51RCkOiFgdtlbcCBLgQWhwtrFOG4Z0c3r7O87f2WpI1N\n",
       "2If4QXLzD+dzJ6Vc+ZBZTOZ19gKqCSPG1/SnzP2f39VEra5unPo3qToX+3E6wCo53HwuVThd91C9\n",
       "egBrUf2Xa8ZhrstAOGmROyvvET7vaAh24zTtORrWcA9tSUe9FiZR1d6ScZxfhFXOM9anNzb75snI\n",
       "nBtL3wO8ciqJSF8EShe3M7/noZn5j3FsyJLqvLr4qLDH9NdjW0lbgrJyjizBcs39bQtRyMbT4FPJ\n",
       "bnLIlJIGZvgBfp86/K7M2w2rteEbkunO29XJvKKu9t69vQ+xCDKgFsdp6utaVRHk3N2ta8KRQcqf\n",
       "ckRlsw5Jxc0wTrpXIO4zje+w2O+TLOJds6v0/dChEZY4E+zN2o70+xp4uLQ92JYXIik8oOO+Zn77\n",
       "wpPPIfcs1ycsL9muG5uhuqgBoxHSoASyP6ulDgvhfpJzyeOP/QYJjJuOH9eT9ZrpJ/hKSlDBu4eo\n",
       "4640dzJKM3dTLNbPh0aJIK+uBY/l7wgNp+1jDM0xaE/8cTr/YSa8YGtDhMd0Ha/yzRYslQjI/7sC\n",
       "IMEMDFkQld0EM4I1xXbL9jaYVq4B1TA5fYzLAvf2EV1u1OzKtHWXEVX4mjMf8Q9+2ikWoAyLQyz9\n",
       "LjtE/efnTFn54e9KcGR7KGjMiBIHIwvGJxmb0qFhqasNdYnq5cNTOUCSEu8mMF5gyxQ7MDsBAWBr\n",
       "UcziYoGG9rMvnufb8y3lXsSCDpa8VCagElJusTmqSI8DvQCbRchvLUQ4cDVNAd+w0Xe2k5Z7eYx0\n",
       "IdGb2rNVvwsMpe3nphjE4zgnxM6jSXT7oBBCCRFipsdYZbdBmWBXWRb1vowdmA42jv8mmX7uaX5d\n",
       "CqVnorzYjDZsK935OsF00ovXxILGgDtL2MHFsDBigCMDr82Bk82VyBzNePFRmTp0ZJeMTWFlamUe\n",
       "oIhsio0C6CaQGJYt4NupMY2ntnuf8iAjqw5NKOV0C9bK86VBCD9UjuEoAoEmMQl3b0XstSVezQcx\n",
       "Uja2QfW0uzwgBYChKzqjSuneHlJQ8DaDCzd5Oun52Z4Vx55XAdMIrzbgqLFuRltLeY3WKxSGkHym\n",
       "qCa2RgzI4IJyxLQXlwXxp/e/Shxset9FiPjBA50peCai3G61yCSja7euiiAQC8igPks5oFIzGJBh\n",
       "FOA1rEcUurjhtsBP2l7JtocE75293SA5ee8DtIT5Y+P5YpEbDGK1yf9AwWVqn0+RdkhISxz7b8ta\n",
       "WV62BDvvgHKXddtF2IKhuTWTE3J6+hkemmkmlE/JDNIzruoENGp/CnEtWM9MrQz52tzZgw/UFr1f\n",
       "+KtbuZN0UVN9k9GH4cuquXFhkmKf1rVUYnxMtcZhTknXFgxy+0X9PVLkJfREkZiloNN8bMviDorM\n",
       "hqzyqxXkR+61cIKvs3uilNV1m6ecdfT2BYTRC4/AsPF0brN4uMpvgd3OULbb3iD+25M5CI25vCdo\n",
       "gbqjDOL+kFa9r+F6TZJr9EgbCccxB2/hwA0QQGjdxitENmf51DoClj9au5fHpegSxiiT9DjjmPcF\n",
       "sIvNqvh2+60dpD7RRFFwBGB2Sw0ATEpnFIdA63lf5YnihiaDM+qJVjctmcc6i2dk8Kdl/nzzxTlC\n",
       "Gns+PuTUnt++zkMHKrriSO+WPpf/3t63sNqz2mZLJ8c15UMpkDdK9K+tj47ItnGOibtPtd3yZWrX\n",
       "lv/WGII99PN1NGxldRByHomCoRCCNs5sS21exlwkhdjpuMWeByMLwLBgf+kQfXISZoqxfK0cXP6V\n",
       "oEG7DAOjeMVzQtZ2bdK+2v0wGgBNXlnvvhNAQ1YolgDNVDxJ9HHaR/pqky/+zUm8zHnTi1Dn3k6W\n",
       "XvtAvjZZbLBidSq+MeXzdL2WSSHKsH4AiGI5WYjc2bkKRbKsJnTDcmJd7PDOm6OCHZL01sQOd7Uf\n",
       "1MmERfDiX/wI+/UGGh6tG26xtzLS62tAnA6uLrh8Mpz6KM0hjMhnxfZuB5EdV6Bvsm9CiaREOmv5\n",
       "W96CZ2x9e5YQiVVmUTBR+sATkbrmPYPt+RteUTN5+5HEAH/4YJPAt2wZ8R8j5VoH5vPfWGcHRGG+\n",
       "2gpIbpKW9JFLCqmtEcRVSD+uyF81S5KF9cwk+sebwqo3iwh5ORawV9Yv0GFNEx3CGSYkVvHaU5SZ\n",
       "rzYUQ/uZn/oBhkEE8T1kcxmFKaaWAk05Njtq3ETWOznNdVr1BhRoxxpIP2fMj2myC/BD24fzyEL2\n",
       "8W6dxH8if7UX7PEbF8A+qNtnyeocFqcR2xxlDumzWqsybiHRf+g3ej5repix8qb6qOaayAXO2CCh\n",
       "Y0xALOeH3eXvPbtWCjXcKSAul27yyW+ngBDdqhYRpOvEAdgwwn82X3ecycdo4+VEXUyOquHolwb8\n",
       "MDwFE4E29jN12LUg7l/P3ubMO8jQ1sual3UYwM3+VaNQrNvuD344IwH9jmwwCd0uipm8kdHoRXlb\n",
       "n/Bh1eT0YkRWy3e5KlbBpfo1d9og3ZuuP4Bu/an+cyImIT8MrULvgBNhry342j+cmxLzkQj9sNU0\n",
       "vpLs7uPP2mU+QauO/5XFyxV3GmfD5doayEned/aulp3eafcfYw27HSmfD6a8aDTPQTfXi1uiFwXG\n",
       "0dnAUfOi8n4HtF+hYF/KMgqpgyqerO8bwf0qC0+SC8l23FE+rnEhV7tQ0PXb8DHng8zAw2G9axwC\n",
       "qVK/+bi5EAkj9LpE6fcBd/U2JyeklwRSIl8gzI2VRbV6e7nhj22WsL+ZCzcv+TroryaV5ftbX2Bq\n",
       "7rf6eLKKVyzQXpqqz2g9A1zaUujUiMUvc6P/Pd11SxSGTn1bwEQ+5dyS23TvCASz8fK53t0HbiCM\n",
       "MP3z+qwmqn7pvG7STPQuAAOgt+8FygoWxb4cPjtZjfBrMZGguhF7zCdOIeTVcmS7xBXjI8eAbw+y\n",
       "bVYbIEMJcQjl149fdLHunKWcLx8ChK8QW3OMU4kae2Uv4rVbqAFASPehsQyL8YS6mpKLxgZx5Vl2\n",
       "aQwuXwn36AqFc4TT4qSowkErGoDPKbBJO8okbCiSyUUL1DcfggRpTaoSBN+X25LZh8cOkxnP1Z1v\n",
       "l88yhW2NeVJ86N4YB/eHRUZGnrEjI308rNSYLkRT2Lut4DIGyn2Hr0uJo7vyJr/UIpZSbduKFmMU\n",
       "KrBAzJS3ApfShD5P1RoW4qfrnHzgsaWltQvZJKR1Ix30V9i7Bqv4NY+A7txFDaoIH2eimbaPxRnl\n",
       "hHigwIlezRldoSPVp4Xue/Jkrx0Roms4XhAKtyBz62ho7lCf6n4vSeHT90xCz0Cp6LcYUh1nrr9r\n",
       "R1breg6OnLk54tjqjMno7Td0s1x3g5uyZXlLHk64IQ+6sADqnYOW5xYvIExQ9TumsAixNr+Hkpav\n",
       "aBbuNcR9VO/iKjji1GanFEL6+AE9w58OHIIQvi0yIU68ZZxvIX5wRoW+O1uMWpJkDmSdpHd7D6di\n",
       "rtgjPDlrJae7eOoaIvm6dBy0tKMy72riEr+/Bgiw61S37LJb+gkvjuj9S2CMnnyhbJJUqHMfQ5rj\n",
       "CkG6jDYevGtQI+Ls2h4SzaMj+z9OAXTB8YO+ifmzBTUuVJCvUnMQIN3tF/mpSsTPHUHtVf5Ux/Pv\n",
       "0TQsTsOkozKCZqIf2n1pUt83KWqVMKgVd9NXN2Iqkq25Il7aPbOvGjW3jYgrBa2Eqkaz3n2bnYxR\n",
       "uKzT0EAKGEtsMqKg+TYnu2jttVo1EJxVfNXf2PqZqe2DmoI/9vYVVXUqf06Kh6CAXmo5UIr8hoW9\n",
       "lce6G5Nz4unBq7NVU1FB8IG2eF1qmNiutoHGF9Q8zfCTQQpPnIuhKWRUoN0fqbebOkSCrRRXTcVe\n",
       "MLTSTXKdbfvrH22LhoJeUJoAoV4IZ/0nFIv2QddrfXJjexOUzVnSR0JlZEHpPOGBZnxdt5ybBmJ9\n",
       "1+d7tXAA8jzTO47y4+Ic2jAtIopWkBzewxplDpCxn6cWlp7iSHT4cxvNYdvMwJPnI/8B1ntNVn6z\n",
       "a3D4YRqUlFmmJ+syaNmcqTsKrvdqFSWpe3ufcI3VFOw/L89Nv0ZI9fBO7oRDI4lD9MZRKJ5kn0Q1\n",
       "LsmmLC4BudNVDs+h/dGDtiXNOr6h17QcVcv4lgYvanE1okvUQg1/90SO0g7FO4CInExPioaO2fq7\n",
       "Ss3mZAkxDdtYqbcwqxj7tnAvH0M+lx8dgkU9QKgdhAy/9NxDkWQTsYbADK0QUQBIBULAcLfP9DiM\n",
       "3NHc25gNm2t1G6/kwTMx+SP7qM1H//ZJI08/m28byOyCYcRbvVIzohOslMymELLI9P/+U1FLFzFQ\n",
       "qGusb1a6H1emxVKcsnqsY0wfbGqDjMu/IKn3ZbTGPWIyP/5ZDVw9tJRhM2KU7PMcAb9Fj0frVDvV\n",
       "miSK3oT176cbilX7idEnncnYtV/GQ/RagVSayOSbJKn10K9bfv+fWIZQo/0nleBByHvRwB4UhLb6\n",
       "FHm3HfPpZQgZsR1oQAOMgyApxkMDwlVWOWyph5YeYzIfaKZck92dP3ZPKEsfIGSy1MkHT31m8MDp\n",
       "HbJF8fulSK2TXjHZvhmodLBDDBod0YsPVV556InwBKEXSGTNhZHD94+jQhp94IgCRCi1C7iiY2ky\n",
       "PQSEih+vrfT0Hosu/PxQOYYJsnQMjgVJEoRHR+RvgdpK1P02aV3G24KHG+E4VVFlLOgqC037no47\n",
       "d/S9kc6Hq3QxGLw/xyz0kvPuYDga8kfVOeY5xYO9m8qCtIfVjSwhCw+23qC/ofv5ddJR2/vnXOvm\n",
       "kySSG/ZT5lW0qO2l0QnsZ51x8WcwyMebl1AsLA97N4NJNHZAVWl+NdtpW4MEbCP7S8kKhGGW8EwC\n",
       "8VeEEJTiv1U9rl/TjXQ5JkTL3kENMLyc6yywcPAEzrC65qR0AhZ50kG9o3Aa0zwDEJ8CowszAGOI\n",
       "0uTuh7Vzb5r6kHYBzL5nE1/y82Q62WeDWjcIAP5+ifQ9ewVJlhg2ie8II3n/gqbQRhAzXiJ06ABM\n",
       "T1V4ylu6RN8K+wTcuv6MhUYX+NeadrasFeq/lHjUCiMO5jtHZXFDaqesCZXwlpARhrFx5TGPv57S\n",
       "H/CCb+XTJeKBlHcAHtP24J+5eQaxQK9Da+GOZq/fRb+TdqnilHovWLIPtqDCHH94JP4VrrejxFNt\n",
       "QHDWNACbetdQAb8Ap6AxBPbpXzrqmfgVa9hoSwf74TKuYyRsdz97g8GY7zzfhqRQp1ZNaOYM0J6q\n",
       "syL7MYTEJaQ8OPVX8c5S7xOKFYhjEntV5dEuJhQRUKZRs/DGeb5rCz7TpglZw5UcsVrNTKB+bIwo\n",
       "Z+6mPsdy4Rd+DOhQssQVufjI6CGw9wpltnccQe3yU+dMvgJHqCp8xfI2cfBgoGpuapPKIgAKHjwN\n",
       "6xHE66ywC0B3TQ35MZW8TNJnsuK5zXJIMrIQPZklqpYJZjpqcVd+r257G/0ni/EBCotCfVUzX2M3\n",
       "qS8ezuyt8SY7KKYLp+DSAb21jbw6d0Fno4Oorno3ro8qvkb2o/ax08iPMoSHT/lGcHZB0tl1vZjl\n",
       "VPXAJKBYhZVvxbyiheJZEM0rjjC5vk6sOtlA4JhRtQ4vhWzv85PbyxleT1dno5hnP2NnCYVRxivR\n",
       "BHlYJYtk9KIe2Ausr/sLmujpcW1uodd4veeotGuK0L0KT8cnYtTbLQSps59uY3vgHTXkQg5NVmep\n",
       "6m+fyF60zMmc11llUHoW9mp/LImUACOShaJj4bWpHca+/na5ZnptPKEUn09+GzI/yp7HA4L6pnbC\n",
       "ZlLiwLsZEU3pKE1iv8VPY3gmXmUQvNZ8v0TwgjOCvbhhoBrRIo0CBk+EBv1DGbP2OqWo8xx2g2Yv\n",
       "KzDcb+oCqz2Qjl4Hp5Ta1Ip2ZjhOah8mTG6I7KykN0sDz6Tm2JEkyW0tRRmgby/VquamcI7d7DwM\n",
       "MXvpsYL2BmsQ1d79ueeMywIsMSoBc/Ok1DwXSF/pV6dqLj8D9CocxR7jW3w1MC1RmcClmgryzpC3\n",
       "8MB85vk3QXpJ8qpcGfZ8uWmoNhBlyLMpZk0EkdVU3LNZ3XP4AHss3X6iLsbKZRMyZWZIdALIy01o\n",
       "l5R0ufBblDdM+d8lO2NwSrzEt6HOOPFkemgzc8VQH7eJhdDhy1pyuNu+3P6gcdwQ8iKt6c+I3xh4\n",
       "LRhu0GE+ZlmNCcACM3ZsJ0/6RWbTbT/Ucl4U1OLt6jfez6QK/mdaZoD/eISFzrFFeZrm6gL/9R4u\n",
       "TPn6NMIg33E56GRL2vsycDd+6ozyJ77+4RD+4+0h62+3QBLhlc1smBfeIkU74BQG1GPSlLUvpORV\n",
       "Ly/vNPN3Q9cK6Z2/YVAYREo60Djx+hzQCDJvK4rbbVxvDjwpAAD7lmUcct/7+ho5/Qfh0+jPaojb\n",
       "tOjtnVb+LqG0cU7HsOzbN6pJspUia+4d9e5pAKLRA8XP+toxik17r80w470Nhdzxt23lE4bsO5Nh\n",
       "mL2LT7jfrmlrus8PEDHmTEeh5hhuk2Eo1BhrZYN8et4RRblx1724t9YY3hzU5l0FkXibj9A3TOn+\n",
       "n+cYEO6Qd/vVBcrqKn4ZaJmMnHkRdf5pa5hFaYhe+Qb0FBeZkeJgLYXTZWblM8OSWuLGchQyigCs\n",
       "niMAtXKL/kZXdIbNHvW7aQAnfb0Ybm8UbzRBUvRhUZPec5fHiMNgJSM1Q7P/pagMuldEiXX0Dp9d\n",
       "BoYRuvDPjkg4Yh1MRXbOf0FIXd2i04c7GA+YT5l4mvRuvsQvs5SvbYC4VvG/XiVGBbQSLIpJeppY\n",
       "F8xWzymFCYKSHCzwcotRnESrnI53mWmhOioLSS76/bunim/krIXxh+2jUunNuy4aPPVg1y1ydQi2\n",
       "1znGDITq9iGrsTPk5liC+w+peg+g8PWqfGowAAADAwOgZybcJW7NZoJqCVr3r2sypLA+cFI6B4eV\n",
       "4ADePZ25PuVGAnnS716hDfJcgYWg8WILNVZtokojFn3v3XOIYrHftxQJxqeaXFSDALqGgkHqe8zk\n",
       "ngkXoSoq4eeUiMge6BfSiUBunsroXcue+cTAy6F62gkfOtgumRvYI+rgKEJnviABOLCAefKYNnx9\n",
       "/IeNR7HJkSlk0hA7hEg5KbWMsWHlUaqnWTB/98f+oxlm11BKKP/xyQ5WkK0r5ayOQ8WKDtvtTJCx\n",
       "x2dPpga94Np/yswtK4pu+Xk/lBGunkvvhRCMkWGqbx8uyKxs+/ky3uJX9/iNpqvtNM/MAPi/VnLd\n",
       "AMPiqaLQI5fTEecligPBV+1hBAa0Aq87+Fgq+3OuUgvOIHZC9izyChOXCURrCrBsETyYyWE8SguG\n",
       "BlBNZ/v8r6xkuKmXE57OBs5BnPEXgyv0/h29pYmqhmhzVSNyjDj7y3tJcDdx7BJNLNdjdsHx9fAf\n",
       "KKEOdz8Qi2GyeI9BdQO1LdPnJTzcfFWV6d7hLMTOtlWIcMTEBU0agNuNg+0kOBJO5INRXfz+T9Nw\n",
       "9SaGlojNai39kDWUQHMo/hxp/NKcqBaAbB1AEDk10+tiK6MpRidXLMnLxLW5a63pb4AFSrpDa8MV\n",
       "wucNpAAoVCvcUUL7iZSLFzaiN/aFnd9rCbv12EpM9EPjLasAnGNrU8mzpIoRe77PNj/7VcZTuWR8\n",
       "+HxxxsokTiwYMrUwwLeMaP5GxpZXfW84Kl1UJ2msxvZPWM+cNWYjKAaZFRFM5lOAW13Lzs+3Ap6J\n",
       "IXPHWy3HzpRK7TDKEdNgG1f23HbVxHOlyE9mGbOy/JtRPTURWVO01XaEEe39P7ERcXReGx5Eh5rl\n",
       "3JAjAayQHi4y3hrAzOmt3P22llCbW0M8raglw2Qna2lXeoWcGM5pw3Lsknw5AORbWaQLDf8yUTG/\n",
       "8Pk712cxyktTkJ6oIPBAr4TbJRE5p964Qh5+j5mPYGQvcScI8vZs/K4L19TImpSdB/QYTkp02P7R\n",
       "flnfw+mOZCIie2B4flKewc+3OsPllc2ScAcC4EUvyipz7r/SiXEbJN8/kstQwC1uETTM625j8o11\n",
       "Na3fmoAQEX5SKSCTn+FXEDMflXiYXqC64ZDwV73b7v//xJbCjrdryPAYJ/DjRitdSk4JGKyM+INT\n",
       "9zm5O4JikXQWkvpJqC31jh4kpfJxyY6eXJolO0lxKoYy64FvbYjI3ZcnOKVJu+9t9MwHahW0n79r\n",
       "zUaDJNM2cG8v0RFka+pSXsxYGkA8EByueuMYnWf+bRDQRdSIbwvBzcitanzp6MR/3CPdoGTf0/9B\n",
       "RUl7IgVrk7WCMVGLfR6Hc/isCzmW+u8hSHnIGw6ocvtWbDD2Jc4LrVqzPCbSn4QvKaQmGWpszPEf\n",
       "JhLne0dtWdyDwUQsbLt9Gna2eAk1GIPIGr8BjmchWu22uCHwl8D82sNkp9j8NEHgh9AAAM6BAAAH\n",
       "MkGaJGxBD/6qVQAG+6RCAIhc87IrV7LUjm0kPwsOHOgaQEMlQnW4T0rTGCi0SdiW1R1s7IfSWHLq\n",
       "hFpdCuCQFThMPp3qXQMb1FgeZEINMwYhtC6PGAV/SHyXlYckmIkkP/q7KbCOlbyOzmUVV7xMPbm/\n",
       "rqSPh2UGK7TELO0kK42vF62721m12SpQ0ZC+puVwlKlrCnDdvX+q3fvpaMYB8VtoXrPmm3RCCfs3\n",
       "SosGEy7ljRSBXsMMmqyyhnYf1CIEPNg8VtgZYnEj9faVShdvrJpl30XaCpiPjOeg+FW2tWHf9NGC\n",
       "oAjnodp6De7guNyjzOxzL83HjJqqtgsCNW8GxKZTN3j2dilqVdEYZyDTmwYHHJj5CTr+BJ00d4zw\n",
       "2gv/1WDj8YJMHrGG0VDKi6IEfs/O2QKPKQlATiSkHpiijtB8845/EQpTYFniz8ED/ePNc2j5rVFs\n",
       "nZx32sw67mp+2OY7Oa3uQMRtolpJe/1B5WP6uNycNSa9Arj9U1LNmSf7CnPrAmX3blXQJjf93ctX\n",
       "5Xx3OYEDZG68UMks/w/lnwbrFLVxEt9qu1QbpCQscM8gVVfJWhZo1zFHOFnbluNlHvevgH0r9Uug\n",
       "WBQ4t8q1dxo5PfcO0ffnzVl6i59wgSo9aA+Z/8cLcSbwBOm9/z5tQLEwknH1cbHQPAJgmu12Qmd4\n",
       "ODUPlJtaiYjN9TN/avFo+1StR4YyNgAfUyZlJb7txh4Usuyyucu/hnj6byS+R3yAvwJASakxjVcu\n",
       "0pFk56YqtUVgQyPOLFQ3qRkLTDSltvgB5S/p/yhdxRTQdVbD4by5YCvAJWOnjUWQrOdzW6LyEfql\n",
       "mOUikU/zRauhqIXDYAtF42FsweXKNTIFTA7whot5X1KpfmcP128YOeePft4yQxWVYfOV7IYmrSAA\n",
       "pmTdBEMWDLnxXECe1meLLzGPBmXT1kJmIPJ7L/Pq1+L/URS1m6QndkN+6J3pO5x5S0ZRZg+MZWM4\n",
       "+Ta26XGcAAXIw7+owa41NRxgiHQSVFhzV/UJvAJQ5dwKOP87BbYpp2jkNEfUQCIejbqH5t1Vvsgo\n",
       "H4sqsaHO1W26zXjb0jaoR7KToxXnhEbfnVjMCQtVJl+e1IST86LFLHI+md0RBKCU7pIS8pcSUvq3\n",
       "saULO20dvlFC7T7twN02ncVBnaCBHcCuIWa+5vCcykR9s81OZ1bK+FY5KKSKnTVBr8xvmwxP7HRT\n",
       "Y2kKCw6QKEEsK+BEG5ksqE0qmZIABIrrz/d05lGVGSyzU+qNxkj6VnKOlAx03e+9t4VkaOoSAZQ9\n",
       "om/QCNzHTFjFDHpEpvGEuxyzrOFIfvcjH4meS8IdYvzZ1NDWXwI7w4dJaVSfAlz3cSL3Ewh1bFJy\n",
       "SwnmyCvvjHbIsV/SGfI0PrJv05u7k/sfYDxtgJCwUe/NiLTDkFtDEwkUoTaI3LtCqc+0KWwnZd++\n",
       "+X7cZBKFa7ovpEbdqNFpQL19RnShZ7W3CqyHJXoBXEBI24GiH066+GyYgeuKXMqUk13M63Pfh2f+\n",
       "S/S8VNPQXXSyXAJGt5zTrv949D4pRDQ4t9G0ffhRXLuzdfEFlyFI/AHVHaCGMfOyakhVaMHIqkps\n",
       "dmQ9BV/6easL7iD8FmgofQbhpQ7Dr6Kin+Bp9AI7ph6tl7tB3jzUz7WNVC78GNuj34ggMDFRWIKp\n",
       "UAmPUCvVJU0ADJQF9ELfNbMS56A70KuNt9Ciz14UW82xgD/lgk3Yhb6gvc9Gav+5be0wFIF+h3vY\n",
       "UmGlYCrmXVpUfgDC772B/9mNiasatd9OcFAccjuQ6a9nPL1EnUEUOAJK5fN37nCgif38UQ2v+fwb\n",
       "QVjncnMz3wKZDNcEj6/Hs+zFS+kq2MV4w52h2EppRYNomMQPdI1fkBXueBpO/lNXV0Q9OWIIjOFO\n",
       "ftYKaw4ykvUNkK3vp17IXzmoubJBnhysITcpxbZhNrH5TG4pPGwj4kJ/3T7FlATjwyLOs8AM9QRz\n",
       "86FTWkYgSVwTQv02AFdZwe0JmS46eIm3cMHARLSxAErBIM8snaBjbQY2JNVG3XimxG8V8LC1sJDN\n",
       "p9J4Vu8Wv8OUy7pxBFLiaD+jX2Zz34jHjXg9ezED2Sj4quY9p+LVG0Dpbb8TKcg+d5V2PmUjizJ4\n",
       "Z9p8Uw7uCZ1wyjb3c6PMJ8cx/DuEfPBG5zHKB65e1l4N3KAU/b2lhM+axTF77nb4wwPWuOhBDL8H\n",
       "JRd3hAn8cAgkOE9DwRNbQVWoVPSG7yvkEH0CbuREBXK38Qg1/RPa3nf56oGv8hR+RAawm+TSK4MC\n",
       "E1/CmraokKod35Qbh+WAJtkU/TKevkBdIO0ntrQ+nq6JKa3QHX+G0jpbZ/fkYNdNKfSpOmsVSVVo\n",
       "FpUy485tJqcbu1KaigiqCD7Jgire70HCkgW1wKDEDjewTIEw943RPAqF7rAaLwJHvGK9YbwCZvy/\n",
       "Z98Q4XoHDn2gb5bUQwL21ADKgAAACuJBnkJ4gh8AAZuZ3Jg4FAAi7WNtWmCCeWPr9zmYGDz937d9\n",
       "s1xbqIHJnnM6D9pOHdKpsST4nyG0p0MRDjdymIcBGEpU8/r9XbA3FcAh4HRGN2MF+txB9mpIgl+X\n",
       "tjRMG3OBTtmqmnx0y9gk8o4A1JT161qlNNnG9HWf79tUmoy3aLTqjN+/Hf5LMh9Y7+ZwI+kpYnWU\n",
       "HyCKdspDXG5Qq2mC+2fsJsCy+x36yuWvDaAMDfn1rwvhJd7GgtwtOUV2+DrTo605FHiZ+8DVmCkJ\n",
       "KxKBmRabqiCBQps+TQizFYbPvWcHu5JEdYDQMEVt+CfZCwEXS0oqx9v+acbpt/d4j1QDk90hgeEs\n",
       "4G/Xd23RwA0Uc+c1eWBWS7s5WJI5HaCG66k42xfXU2c0gvu88vPt/s82kG6oRAwy4FFKv6SiHHDm\n",
       "oXMAIW09wQ7SyOntGwHxFPJjwdALcpKwoQd2c60iVLUOduhuVYlEsbMnhjPq1KU1sdU5SRDRNkLS\n",
       "qEwf1SFRAq9l7j9oFthlvPWB+OaFEngxQbyQgUs68mpRWoENE3snbGSTvt2bLumgYFlL+giddJHY\n",
       "uZwTcHdJBVn2rGPEg/B+qqkCM8Y2YlVVj2J7WFnawhimOJ8Q27WMF0l+ENwH7bphjWNCgEl+gUcg\n",
       "t+sbJsH6lABKcar8lh2ATmZmi8tJYEs20ZvHsf/GHblnTr9WpswwgRsdZEqPi0bTZyerQNJyXJQu\n",
       "Qt0VJ6zQEM8c0Z4Lj8hXKasoLDNjI5+W2v51bkqlbjrALUYTvtWcwT/4rYZ2Iotcsoh4nLsdR+Et\n",
       "Lnjtegfp/4mHC+MriNtk/6OSgBtki7J8sRPpxFBMiYXLnlHNNN5KvtC4MCvCpP5W8C3wVa6PPK0o\n",
       "xlNoxyOODzkdZGxm1ydu+4+boGGcZZqT80uQCTlUJj2YlFspazjgZOTkcmiv2hzsYptJiA5GrdPu\n",
       "Yvj7FMXxFPZJpgQw430uYjwsjTPMOF7EAz1H9a1OUJzTMlEGUALnHYeOS+gVylwz/5HdAXfuvTAH\n",
       "uptnKOPhj3Pvudwb6OIT0N4y6WdFl+VSmWM6t5fMqohK3vetS05NdVq/+Pd/xQql5MuZcM6RhZZr\n",
       "OLzEXlz93oBZvBGNedFUtR2olrj+XBMxYS+6ex1Np9MDNX5PWoMlo1/H+5OsQzd16vXYfXcymue4\n",
       "PdOBpy1MpxOsHkCrlQANbiRqcqzWmmBKl6qsy/bded/DJqWdIgNH/eCwCF9hhF7NITzFVdFteWSt\n",
       "nj12UeTHag26EtHCdxvR4GwClekvsBmGsDMwFXnkK6dH38XZrLIaBenn7D5m+s9cMPx2NhoZFbYt\n",
       "ZeOuqb7eeRxErMeN1LYxqnw8ZlWT+lYXZDPrbMm4MEaM584afTLwuuUPOH130LdeiX5tLvmcNqVc\n",
       "bRbjtG/tT2UbjOF/SlNUXDma3xa1gVhUlBnDfJggugdo76Q8CVvyjfQX47/psMJOzcu1pP79qu9v\n",
       "qHPCi/4Dol+88JOqHG2ggXLdjYT8/E2Wrnlg38fwUyCG0Hl3+feZgP7bBRyotjKAJl5w3GMarNao\n",
       "cOZq2qwIyttVEUDbs5+77Qa8A49ogI1+LGordt+Kn4iyZ16qwn6CFycrE/9YNzH2TCqj9NDj9URl\n",
       "DgvqrklSaeBJ2ZDVJMPPgI23ZlR+VzZgYej12xgZPu9rFWmKF4aAVA4rm6d7rObpYxZav5//rBBY\n",
       "zB7//kz66on8/yWQoM+fDDi3ojktd3L5X2LBkxZ0mnmy818YixVr5cm5L6gvyvSirhQUkfsiDoGR\n",
       "DwHkOmioZkYGF6pVBrN214/TwEunVowf1k3InuZZetsfhsX5aallxrjp+ljNo9cnRYyq5ZxYl0T/\n",
       "0vhpnZ0W/wdN9MHq+XQL5ncswFvrXF11/dW1szzGd3YTInT2M73PW12LhvTBrlCExR+iRXgqlgm4\n",
       "WXpMqPNMC9iE1+oXO9ywQ/Yh6GtOUcE+MKl6/6FSxjYStt93alPNh1vGAu33YGrnBO9i0eA0uJ/0\n",
       "BmakVHB8I97SozWBz5st0n482ELIVXuFHELaJU11n3NMWW8X+9yT5WRGAreO7ba3zQf68YghOXos\n",
       "zwa+eu3KfupfJoRY7xGeWt3hpydLVq+xA3AZj6PwTMBNFHgwSqsMZ3BjUPNkrmjekPpwa3QuBG7Y\n",
       "jkYAXLNHh2phmH4GYursthdflWEIauWhKhD5y6oTKnkbV2WS2K04XIjQARAG2HscY1uvOT9+3IGR\n",
       "CPL1wg6UQIwjn2203XHtQMijQQdKU1L+e+pK9yBYfSW5ZQsvQ0Zj8E365HCa9Yzdz92XSTEQ/jCy\n",
       "K1407a3eB2FbHBHcK70B0KbVTJXWdigA3x5HKnoX2yeQNHkaLrdL0jlJWA/0btbzWyMeq+Dzy2Bu\n",
       "qfOqQPvMDLbwDq1M3Pr9pjDSIh5wqDHPf4mETQ+s1F3cnwmkOw20QQfHQQNauefS47XgRSgbeW5y\n",
       "nJzBpVMTqSpo4hCGNTCcMA0xW173YTQwta4RHyZ+cu04IhdWjits4Z8PK4BlBPNAB0CFg4alBJd+\n",
       "SduRjFKFzqXsFyqOj4K7eAxEoSokOyEUcxXS/Ee1QQNUNfWBKNFC0m+AYj4fjc0tDmrl2p2hvstC\n",
       "RdnUrLAwqTUc+lzzyWxKA+mJCm4fyW56xTHHVu8+JGvadiffCRo56z/FCVln32jC269qg8Ye1lb6\n",
       "iSteDYB0YXMb0VH/KpJr33hc8eSQk2mGtiLM3/hf9GLvkfg49CCbXDFXAiZQnT9VIwO/EoqLVcmF\n",
       "4KmdN2TzOAD14gpwk6qOvOo/uSn/K6OUEjMSnelHwzleKRdGDiXP0b5N5gZ90VhcAsEW8StoKphy\n",
       "IWqW7xjVQwkEi6nXeM3bHMiVVCHyt9X5OwZ/O7x4XTQbt4gN9vkuw6vmfZkVzYnnNV9uJDelpuZW\n",
       "I2l8HKAaMeyRWbG3JAKwB8lFvs8YHhKgHl2UbLaGaJPxSG7nmsCKSv3FTnjvXkfWAFRLEMSXKyvM\n",
       "pMQS/vsX9pzhvM1nfJ06J3b21t86uy7SrMgfpmuY2pBClDtfh6JIF3aIJ18X4jrDyvs8P/Fo1cHq\n",
       "b6rgvbRzcSchziVuUoSVk9uKLXFn1jzTs0Zl8opeBW5kGUY5g0BQxdS2kWiF6e2zHchYtUJYvEmi\n",
       "w/muC/9H+Fs0SPYNJzypVITgLx675C1H/zXCKBX5ZmStDOAjopuu6QQXdFhYluFEleuL2LqunnyB\n",
       "+I4nwip4kuZTCeTmZq2rWtLykbtITaO5iO8a4/Tt6mneZwprH24ZGVvPje0RlZrWpgJ85/Y9Dx+9\n",
       "KT+TogOFBS/6MlLHpiXQW2ybSkfYWCR6fKW09OJFw6sekqgktlpATd5XWtyObBjyXjfQRRZzru3i\n",
       "4jQbbrcv3XctSkmkk07pJDh9cE3mBmRmFBjPP2zp433D0a2F0PSrV5d9cb0rvQXNS41RHFVpnqPh\n",
       "TsMSs7bmErCjq5fTqqBi+H9QvxDTVirNKKmwuAXFlR3ndKIVWrK1UPBXVIuXelWnd+7eTMBA2FES\n",
       "nQRQmLH6hiXV0CafSYm55xO8q/MwqTpnGkS92/LYNVFE2I1SIcRTxTPlOWlQBTHBxZIuIop6xFX1\n",
       "ndYIR72YD4ikLGzBTwXyFksZ30WB7RmGGcDWOrlPsghOg1tuV9N4lK0aR0xVIF1ksAjxKs0CLpJ7\n",
       "U/7wmiM5iVi7WOwHNcVUwQAAB1IBnmF0Q/8AA22M9DoWGADcl0PMeTD8Zh5HVxFnWZiZWyAGC0SM\n",
       "7oATgyWv3dVd9//3tovc68frRnog9MTV6pP3wYOEHBm/jE33kDKYS+4OkbyIOM08j9yiHZBmbzCM\n",
       "/1A2WYpPxlhkoWeSUp4HWWXexyelvzcUWwIzPfTtphchehC7ki4uTm//9ZW2Xv91zeMwJyrDY5ar\n",
       "mCe8A0l6v385aKIuLWnj2yT4jGFMb+J+37CLNbIi8sTFtxvIH75zgelfPm5ES4QKig2e/7k1vby7\n",
       "MrYf6y0TksX9jqOgHz9PulTuvNBq1x4tXr84G8Z//b0fPhAparY8afCCf9RveASHPCAPfl7ebIbe\n",
       "KfHjAf5LgmelQvhN60QNb/n6cmNNVXb2mwdkKunmsw+QPZ959JLwgjovfOgEybFI9xz7wZUBybTm\n",
       "AKCLA1x0ZI0cYDYTYkzTyICut2GTcuFQtvf4AkIoypF+v2ViSgTTuz23FDBWOJF94sBbU7Lp6vp7\n",
       "2GSH+d2PB7qgYkPYt091Inpx05aVeD98SKZHQkATxNFcvAtM+dCFxI0ZdearmDNjVIp254LEYl6v\n",
       "b9nLGoqm8P6kJVX1HwjhKK03U6HjqXSw4D8nvyuHp5HO80NQkvhQh6JXsYCCeAxjNBT28IRcg8Io\n",
       "3ZyG0sCFeVzNDmo/82YOQLu6iABBACaKNC4lVIwZyzUuQos6ut4Ewy6/6lQioWAS375WWea7t/Hk\n",
       "FqK2g6ioa03l9ezvXpO3J1IIufJv2APB8mNrkGEEu5ur0iI41/mzM8KplVMW29XMGoCuXvweAA13\n",
       "ze1cb5eft7V9d3DYYNfhM8rx9Bp2F4Y7eWPckdQxUG24GPhTXMN5HSaHvj5z4cnsAc5D3BLlMQgO\n",
       "FS8HveqeHSxkG5rXCJYzupJApd+EPdmWn6me9LKtaaA71uDUk1m1XbzhALvli4MLNLnQx2onKJWd\n",
       "Xe0nboiACE9+98B+IH38xV3lGssBMUi5Age48VIcVYTNPvY9YtCsDSiB0k7FIfZpGUblN/b+qb9R\n",
       "A6KAKCy69yPlNBmCVoavKriVi2PWAV5fKSkmAYdKslLBtfsXRH4dO0qeP3g0VQv1W08Nq2j30/Y2\n",
       "F59UqltGErjZBG79+IG/85knsY345lZ022HH5FuHTam2CRD0K8yBhqH3lqX3g7xfy7oqcgyyIPia\n",
       "mNL/ZUjoijo1k1vF7JK/cjAcilZMZcn3XtIvWAsqM9kkDPmxiQeuDmyxOAJpCcYUfcSbDHE9suGt\n",
       "gKrVZ+/Cwyl2bxy4dJhJq3ayOCrLuWugsVGamx3rAXBuPxK54Xfo14hpKtk/P4Gr0RhwRTts/Xho\n",
       "zYzrpFpAqEKr4mZJjm4QvMk/u2Gr+ugbbXmcv1Lr7ixio7CsEh2qZ2+fzQ2K08FOtV1ijKApUG8b\n",
       "zV9Qx/ul6NN1b/U1NBe5t2WiDLG2m+I7aqBymPFEGF2kqCO5FQ+XDMcmxqAp7cGQBbKKd8eE/x5J\n",
       "r5OA2QXLrq/Bgcl3oHl4G8NG2Zhb29RyM+JRZQam98nubFaCJ8qyQ/7p2Zj7D0Q5gGTNa1qmBt86\n",
       "zkxGnab/p2J94RmjEtC+7kepBtLPhU+nDytIz9FRyUhuVJmp/+9v/o6VKve18qHBEkFklnRrfoMK\n",
       "MT92joepz9/pBDsQh51ctfMLa0UqbrXPt2Qu5zVIHdentlO+FRgyD/oRUweT4YIE9c9s6X0wt+xW\n",
       "ExeYV3uuoS0oqtWWMoUFqlnGtrNlm1k4javvIN6V4YZ6xs5ZTRfaCVlr4s3gP6RKMdOKH/iz5sc/\n",
       "pL1M8mhd5Gib1jbaBNlLxpTsfJ88EdHl6F5/h/4zlXtn9znTc8kLLoj06zcvsGS31xohnRj0kQqM\n",
       "uJxRTaPFgjU6PpbS1lqvWrOAv+89Vr9S2uOtvWLpAhonbKOlDdTx2C9vo248CXjRo3Dc94LL868L\n",
       "RjYznCkYfrpAl/qPn/8PhmdfnhFqRjVTNKnyJwf2p+HIk9U2ffz4q20lXPeftDMWAoA9gRcXk58H\n",
       "Mjfx0vo7cPcb5Ksqyc0zL+Y8nD9YXl9r7AAgpqNek6IIbU1dHLmL5qssi7eDoGzs9PfzFp4bNwF/\n",
       "7dosFVYAYtI/fAy0EuX+FqiyNruxdJ0yQKXeAUhrPn/sc4PxqupP9JDBeLAzw9A5jVQcbeSAW3+R\n",
       "ho/hgttM+X1RZIjQPhAvqvdIEB16WiXvWr+xSrl0VeYEiwclst/OWk2vrUvl6Ne1Z+yRuPFkxcOF\n",
       "2CgEPGYSasPOxM4QxrIjtrrQ+/SKwU9HeKHIRthCmLwiUtxyUWRG/hKAaCeQD49dnwAfm5I91MUY\n",
       "a/XVquOYXb1lgUyTxYqStU4ghqGV4ExijmH+PsGbEFP5nxlxtwF36tPyyN2mzFNRn9ulPBkMDmDY\n",
       "XA3LuXlStARbw2Uf4qovSaudcJdAiKM0o1vxDYf/1yeNHSNdYSdksT25mwI83kPo9VqI2N3IH6tb\n",
       "Ersab0Ew1R+5sECygAAACoMBnmNqQ/8AA4nPh+BsxPoAOISghmn83lkw7/aLu5kuSnM0BZRBkkxz\n",
       "ncVy8iAD34q+6sl926to4XiTM2IHnXbuYxUqDfque2pImp/ywlSYX1fbh3udAd7i+dOjF5h5AkIw\n",
       "4Dvg9CckNBVC1mu7BZ1O9OcOzr903qzh29VpxcOuPQ8grAwietKrgfnxcv4OfkK3xlq7+BonsARU\n",
       "5ijSA1WLeRJFpZqOvQE3rFTRjcRg1qvLk4Yg39DGhkOQgASvpBh3MaQvjlfwYWW6RQ0RQhCXrHml\n",
       "+5JhRQmA3P4gb70NymS2waq+e/7RBo2bWoWu/Hs/w/aJ6sbjEWXvugGWouL3Ld+IoZQsijIJ1SMz\n",
       "c+/6VxEIt6LQGI3GcTa2/grpNyHjFPgf/DZHrBqEHKgQE7iTN8dKNvvuF2JLx5vYFPq3vzAIBGF7\n",
       "IiT8yqu7tC7aBsO62FW+dICbFiWrnd4WeN/1Bjdqui2HQhV3LuviSBBSZPfqHONyn4eKzkKAHbqW\n",
       "XgeNOGKUup6Xuh5Y7sHVKcR6kqvNdNCO+VIsUtM7QW/O0ACyPyD7eLj1mRoKmHysC4GsPr8yzVcw\n",
       "KEepCC3J+8Kqx4SRQ3b8sVLO69waOGrh6BgUHHhVeSmzEkX2MgSAz4OAhE3lTbkGzI4SBzwpGJHf\n",
       "IZf5cnzlj7Um3wkz7jsZoXQKdpExNi62Cww+zo9Dog3vzICDvtpKur5At5nUmWKY397pgSVszdwT\n",
       "k4K1FBIzFsiToXYlZEJk9QGIr1izcUIk0ejxMdH/WbSmlK9/vzORpEnm+y6RoCDNSW43DAg3+Zln\n",
       "5EMFg02NyzxHRZbVUFw5P5WjMsEa1Xp0TWDeGXVSClkFsM2/nMVHaSiM6urTUFsrc3mj8wG+1igR\n",
       "4biWTJKWc2xjaXTTIWbYk3OZaGtK7GrR6XGd8t/2dwmnr93a3eAmQGaKuWbRfoYHNaoE/zg1r7Zj\n",
       "riJVkHM0Ngls+CZ/uL4f4FBiUr52by35S3zzvzhx3/Rt7pD2RX1dnPp0ZVWshI8uvVuHC4dbS8Z2\n",
       "wjYJ1QYAp4lWpq+syNIT5mBpTaDL8nZncqpb/VsnhYZR3RMsnqPjlFLhGLe9XkN2fIjOS4f8iwDA\n",
       "MEDaHHYmMy5fq4dhLvW+kb1hDQv2P0EpLR0+/7s3ffR2w1plTVjGSiSTRgncu6j1OOmbKkRrVRJK\n",
       "Uc39N4w4HBXPPPFzh22haySd9THzWNe6jH/XXTvY5TgJGWAOVeX1p6GyYDtdenNvMluPChNgVzvh\n",
       "gXzMxUMmzgXvaDd0ToPSLBENvfpy0476sgqea8TNPrhit5qUvdreNpjzoeENsa4+Hs8Zq1qXlWA9\n",
       "Id93XCaG9/gmEaK2/QicrigeKqJQ7rSAX/sDMfNhTPPTBCEJdmF72XqNW00QCZ9wRVbyjbZKBt7S\n",
       "KImDFLEEz81wbkMi+Ob3WvjjNAJZDE55Ar+sN9xdt8i+k2vQlUlfycY+Z7unoN/dV02zrO37qAYR\n",
       "BiAnvNNqGQOijd8FccTFcLwUyxmVVZNwf6gehaNAV4Q6HZckdZHhy+QW0FoiW7fyjj3vv7Du4+Z1\n",
       "LVY+rUnYXbGUvYAhg5PC3diFsOsgJGV01Z2z+renl0DdcxZWav0vZ+U19bVxep5p0PPSDPOVsEl0\n",
       "VWntyg581MfV2lAQ8M/lw6b0A2VNdN73X69/aD0IUAzfAbT8EkJYZCznU+4btpqoZRmM6aXCMEr8\n",
       "B1zFYb8fN/TAdsQNZ3a0Kd69Jp+ruE+6/vBpgKjVSUMEQonckMLRGd8m0YWRZS1Iab5ANTQGF4l1\n",
       "C5tD+vn2MYUFa+E8nKytuSEsMZoSo891Z/hVPd3JMsSU1yqJ+DXijt04XtG4QnBBQkNBAjJKEGUY\n",
       "Q66MV3RrsQDhKzNTDgRUc4A23wuqDyyznv1PoYD0zZ43x6t22BMd8ZF4mCe6XYnCwCyuvDTb6/AN\n",
       "2/FqxpSkY2IZiGyxJxsbv60VxbfX+d33rWC1bo/MzbAD3cXy9o88SprSyEBeuUzgrUfLhK2U3Mrp\n",
       "StNIWoaiHb2qzpXv7mvgpNmSNXodtQQobFqy4ZP+8uWQdeMVA50NCyHJjIgDx0H0NPYpmOW2oahP\n",
       "bzHyEE0OMj6IwWqWK0sM5R04cJ1aNHP2RCbNuUsuOH5ubhIkcLvI0nsuNOjs2V72g5fywEIBULlq\n",
       "NKA1hRU9rkCAYA6rQHmIG8NfuSAU/00JBPZuA7bOW9oXUho57LDZQoqoImPxelrsp9nHMQiv2zbR\n",
       "rnZ7GnfAs5aIQrsQ44WSF3h8n3ekaM1wFe488PssOnsv0gkaPgAfEMmxbKgNibnTrtcLPwZIC3Jn\n",
       "29tLuG8EiBp/9xb2n46S5rc2Jrwpe/lLzITunH/mlqmIhf6s3EWb1lyrTIjy/t3oQ4ayNB50CgJk\n",
       "sqQNljuBgRkiEdrHZufOqgWkhdNliiMOSgNyXgM5OZTJXBHDMgOYRSO0eoRjnRBGZxNoaKYfbKEs\n",
       "wf4nrtg5NAhIeTbqaHXJzlVOa10myIzVxLcReMGaYrI+CtJbOC8u4zjkLO9E/5eWCjm1PLCMgaMz\n",
       "+GAvm20nWRaksIfBVLky612kDMq6CcBr6xbN1k9djB9qGlaR2CxTnyGhN+pJwG6TVtCAdwE31o03\n",
       "FVVLvJiWEwsiroJHDEuMJ5Kt1F3+kRgzOb5RxClOJzCWwnoVAaaBXzX7rxDuBZRk5+/myAwmerAZ\n",
       "aw0FirjpSIDPoGBg0R/3Re5Ng1+yi9Rndld+IeF1MLcjw0iq25Q3bOznF8N9hkKJxl2u1jMn2XRD\n",
       "Ynd4lrkzzvLjNu+lvg4iTS/UvCVpFz8pdjeaS9BmLPAoN+4s1RPQx2eOQTGIiI5WHFI3ZYUc3+3t\n",
       "SZniaEBGmB/taeA6QbezCAMQGxWLLw+HtNlmJ5MwMkjdX1WNtCL+6MQBhrwNv+9FQO0y48me4NU2\n",
       "2VWl6dQetvKeP8CN/nAq6Hr7wDFjsJdQT834MoCRT8CjvY7vIX0WvonIIZd8WFjQb/Os5z+o73Lh\n",
       "IQBqGp+a41Ys/GRBW2WM0BvQdU8J/Pi1HZo4E74Up96N6e+YGY4VSj49u0evNTQAihs6AjFqFUJ0\n",
       "LwDzWQvnE7I1tRWmj7L9NVKoNccj8o4CDPzncWC2itiD8zbDBEVKGGckMvoEo2WOmFBz3O4ulexU\n",
       "Zb3MZMkoogVPLMyy6XxjUfEPgNruqmLqr13ZWdW1o/bVDWyALiKx37mdCTw4eqU5f0o6VA6Kcfor\n",
       "lb2dAp1uhSHOEZ8h559z9Nf0y4aMl1lNXnWXrUT+acThKA6eMTWthKwwW0kAVd+hcJ/JNreenWfe\n",
       "7UwLE9JsAIXi0ExHI7F77QoeMrkaTfvmynlE8LPB0ZTjzKKqO+YsyiUR06OVdB0sBNlIqeGSSaBe\n",
       "vgPS9gn0RZk2LMGZ+JnEFMzvugCW0dr6wCg/El3wFXfaa7IJtnQ4tGvIsC7x4VKEgA1RTBTW6BL5\n",
       "TRRh6rLMYMObcnIpG58ELpRMvBUrhooOLZgu6yL4qFg+2ww9C8YF7gC7B+shCiehUU1XRw60CW9/\n",
       "IwnlyMfUqBuIgPLZ7z+C4vzORCssHM6CfVBRVycAAAREQZplSahBaJlMCH///qmWAAoIrS06AhKH\n",
       "mHJeWPmOfJpRj5y+WQJ+ze1M8j6St+hd353moPOWk060t5rk3HZrGFEIqNVklnO3/Vp3FRgVCTMI\n",
       "b72N4Wq5K21g5WMBRoP0RghEogfnbNht/9lRNiTr6P9hECbUR1v6RKhlfMjkqwqsdFVeDKvyE6nN\n",
       "pdW9ikfASW6KvVp3OD9Kv59O/s3eJUC+vFPWAPmNo8Bt0pIvd746Yge5ljBPOexOoxatt+W/wVLo\n",
       "ZcPeZNiINSq188qTODJ8d//zwe29hLCeGgZETH33Qier1YqjVzfRfwYOoG7lSa2lCMhiiWmgZ2Qr\n",
       "ROi9yWMPuc0prhXsA+Hx5eYAhDDfs0Cw3SSsZkdil1ENdv/Ylu33PtwW4zWqXb0cbj1ip+lDLxT2\n",
       "a8g+I9poi41VyCldPC4rel+3MCBACEASy1toP3zWaxumXjqE+CWj+iJBCHG6RuHjnA+BWvlKtEB9\n",
       "68uPZe6PFUbDgSl6uMJNNwiUjyJoQR8QkoPz68v7W8JTH1bt63U1nArOPcY/lrSRVCkr47KMgTZx\n",
       "PYZqcomTR3FdUzEEDPlLWt7iN8MUuQSGn9fEzPivuaVFYI2TUcJ0c3Ah6kvG18LQdewfO2FiFcrA\n",
       "J+tZl4j5pssiiTKIIiY/M55n5FZWPGyosOtfYtXDXrJ1PdD81ty99Ukrf1R8MeHA6yRJdWWSgjU9\n",
       "/mwfcyxoL9wbnHVp355+pTjv1gShhvtxQdWfiHCmRFivR959M5MFioVGfMtbe7IBSQ/7xZYs+3TY\n",
       "2fgJ8UAYt1r65K+2IocZghveH0zxR8Z7e+Fhp38dCvf4Yk5KUT70kPBEa0s2onpjL4xPlVpQL5Bh\n",
       "DclO+9kuWdnNumYt0Yl99J1cwFLJ8NWDDqc6LpcWjhGIFba2QmiMh1a3aO5v9tOMvpQuImZk1Gz2\n",
       "4Ji5vzmuts6Nr6mxaN0XLKootJQFKOC4RT0fmGpIUjfodxOgo1w+bPVrCJcL1rxRvmrXXSEtquv/\n",
       "c1QbinJGozv7E+++u8LEPlbf42cNZ2RoXrembW4OhB68+GdQesbiJbY/JgCvvDrQ+PQ1qlfYK6gO\n",
       "41oplDQT9kT17a9qQ1QKpo0zTgMQfgyTMwLlo4wqpIv240LiYxuKw693lzAPJbUfXbtJhkN8eS5h\n",
       "7Qp8WSjTnvTybviCfS0gvD+zJIcgo5rMuYL0ETi5zBtkJS5tcQOuBS0u/WWPtlZ2woKQyUHoLiYJ\n",
       "Ise0HNr7nbOpHQEtI/KsysKm++/+fIxC+6neqMtaKSHzvfdCYMrexlhManVl0m74xkgeC1yboVKF\n",
       "BlDQkgbF0yIRK1kh0N/tYX2KYRZw01DNed3B65PYb30YsckOWGzUmjTm539tzJHS93cETVxFQg58\n",
       "Mhh6lJc9NtFjQeH1q9pTtAJGVoh8s6Y9GbF8KGKMNqfFA73prgFeKhNbAAADeW1vb3YAAABsbXZo\n",
       "ZAAAAAAAAAAAAAAAAAAAA+gAAASwAAEAAAEAAAAAAAAAAAAAAAABAAAAAAAAAAAAAAAAAAAAAQAA\n",
       "AAAAAAAAAAAAAAAAQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAIAAAKkdHJhawAAAFx0\n",
       "a2hkAAAAAwAAAAAAAAAAAAAAAQAAAAAAAASwAAAAAAAAAAAAAAAAAAAAAAABAAAAAAAAAAAAAAAA\n",
       "AAAAAQAAAAAAAAAAAAAAAAAAQAAAAAKAAAAB4AAAAAAAJGVkdHMAAAAcZWxzdAAAAAAAAAABAAAE\n",
       "sAAAEAAAAQAAAAACHG1kaWEAAAAgbWRoZAAAAAAAAAAAAAAAAAAAKAAAADAAVcQAAAAAAC1oZGxy\n",
       "AAAAAAAAAAB2aWRlAAAAAAAAAAAAAAAAVmlkZW9IYW5kbGVyAAAAAcdtaW5mAAAAFHZtaGQAAAAB\n",
       "AAAAAAAAAAAAAAAkZGluZgAAABxkcmVmAAAAAAAAAAEAAAAMdXJsIAAAAAEAAAGHc3RibAAAALdz\n",
       "dHNkAAAAAAAAAAEAAACnYXZjMQAAAAAAAAABAAAAAAAAAAAAAAAAAAAAAAKAAeAASAAAAEgAAAAA\n",
       "AAAAAQAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAABj//wAAADVhdmNDAWQAFv/hABhn\n",
       "ZAAWrNlAoD2hAAADAAEAAAMACg8WLZYBAAZo6+PLIsD9+PgAAAAAHHV1aWRraEDyXyRPxbo5pRvP\n",
       "AyPzAAAAAAAAABhzdHRzAAAAAAAAAAEAAAAGAAAIAAAAABRzdHNzAAAAAAAAAAEAAAABAAAAQGN0\n",
       "dHMAAAAAAAAABgAAAAEAABAAAAAAAQAAKAAAAAABAAAQAAAAAAEAAAAAAAAAAQAACAAAAAABAAAQ\n",
       "AAAAABxzdHNjAAAAAAAAAAEAAAABAAAABgAAAAEAAAAsc3RzegAAAAAAAAAAAAAABgAAHTEAAAc2\n",
       "AAAK5gAAB1YAAAqHAAAESAAAABRzdGNvAAAAAAAAAAEAAAAwAAAAYXVkdGEAAABZbWV0YQAAAAAA\n",
       "AAAhaGRscgAAAAAAAAAAbWRpcmFwcGwAAAAAAAAAAAAAAAAsaWxzdAAAACSpdG9vAAAAHGRhdGEA\n",
       "AAABAAAAAExhdmY2MC4zLjEwMA==\n",
       "\">\n",
       "  Your browser does not support the video tag.\n",
       "</video>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "execution_count": 95,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAi8AAAGdCAYAAADaPpOnAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8g+/7EAAAACXBIWXMAAA9hAAAPYQGoP6dpAABObUlEQVR4nO3deXhTZdoG8PskaZKmS7rSBVpKQSjKJnsBEQQFcQFlcMEPZT4HdFxmBhhHcBk/VxyHUUfcl0Ed9wV3XJAdZBMpewsthZaWFrolXbOe7482aQvd0vbk5CT377p6tUnOSZ5GbO8+7/ueVxBFUQQRERGRQqjkLoCIiIjIEwwvREREpCgML0RERKQoDC9ERESkKAwvREREpCgML0RERKQoDC9ERESkKAwvREREpCgauQvobk6nE4WFhQgLC4MgCHKXQ0RERB0giiIqKyuRmJgIlart3orfhZfCwkIkJSXJXQYRERF1Qn5+Pnr16tXmMX4XXsLCwgDUf/Ph4eEyV0NEREQdYTabkZSU5P493ha/Cy+uoaLw8HCGFyIiIoXpyJQPTtglIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkWRNLyUlZXhlltuQXh4OCIiInD77bejqqqqzXMmTZoEQRCafdx5551SlklEREQKopHyyW+55RacPn0aa9euhc1mw+9//3ssXLgQH3zwQZvnLViwAI899pj7tsFgkLJMIvKC06Za5JZUo09MCBKMwZ0+hohIsvBy5MgR/PDDD9i9ezdGjhwJAFi5ciVmzJiBFStWIDExsdVzDQYD4uPjpSqNiLrZG5tzsObAacwYnIAFE/ue9/jHu/OwdPUBiCIgAFg6Iw13nHNce8cw2BCRi2ThZfv27YiIiHAHFwCYOnUqVCoVdu7cieuuu67Vc99//3289957iI+PxzXXXIOHH3641e6LxWKBxWJx3zabzd33TRARAGDdkSJ8ujsfRoMWc8ckY2hSpPux4Y//hLJqGwBgb74Jr2zKwW8PX+F+/LSpFvd/fsB9WwSwfE0mIAJ3XNq3Q8d8vDsPy1YfgFOsf3zOiJ7455xhkn2/ROTbJAsvRUVF6NGjR/MX02gQFRWFoqKiVs+bO3cuevfujcTEROzfvx/3338/srKysHr16haPX758OR599NFurZ2IGl3/8jb8llfhvv3xr6dwyQUxeOZ3Q/DtvkJ3cHEpq7bhXz9mYsm0NADA6j2nWnzep7/PxLXDEpFgDG7zmLGpUVj6+QGITe7/dE8BtuWU4pelUwCwK0MUaDwOL0uXLsU//vGPNo85cuRIpwtauHCh++vBgwcjISEBU6ZMQU5ODvr2Pb8dvWzZMixevNh922w2IykpqdOvTxSoPv01Dz8cLML0QfGYMzIZQH3HpWlwcdlyrATjnl6PpAh9i8+1ckMOekUZcOOoZGSfqWzxGBHAiZIaJBiDsfZIcavHvL45t1lwcSmsqMOnv+ahrNqGp7/PdB8z7cI4/N/MixhiiPyYx+FlyZIlmD9/fpvHpKamIj4+HmfOnGl2v91uR1lZmUfzWcaMGQMAyM7ObjG86HQ66HS6Dj8fEZ1v4jPrkVdWCwBYl3kWT36XiVvH9cbR060Pw4oikFde1+rjS1cfwMT+sRjXLwZfZJxu8ZiUGANOm2qRkW9q9Xm+O9DyuQDw+ubjOHamutl9Px4uxo+Hi/GP2YNx46jkVs8lIuXyOLzExsYiNja23ePS09NRUVGBPXv2YMSIEQCA9evXw+l0ugNJR2RkZAAAEhISPC2ViDrg01/z3MHFpaLWhhfWZXfofJ1GgMV+fm9EFIHfTpbjjNnSwlmNWhsyAoCUaANOlNa0+vi5waWp+z8/gONnq3CypBqDe0Xg+hG92I0h8hOSzXkZOHAgpk+fjgULFuDVV1+FzWbDPffcg5tuusm90qigoABTpkzBu+++i9GjRyMnJwcffPABZsyYgejoaOzfvx+LFi3CxIkTMWTIEKlKJQo4n/6ahy/3FmBQTyN2HC/t0nMlGPU4UVrb4mPbsktRa7W1+BhQP2zU2pARgDaDS0e8tjkXAPDD4TP4509H2Y0h8hOSXufl/fffxz333IMpU6ZApVJh9uzZeOGFF9yP22w2ZGVloaam/geUVqvFzz//jOeffx7V1dVISkrC7Nmz8dBDD0lZJlFAaTpEtC2nrMvP11pwAYAPduXhn78b3OqwUY3V1uqQUZ8YA3JLuhZezrWsYSiLHRgiZRNEUWxpLpximc1mGI1GmEwmhIeHy10OkU/59Nc83PfZgfYP7Ea9IvWotThQWnN+B2ZC3yhs7YYA5YkXb74YVw9t/TpTRCQPT35/c28jogCxL78cr206LtnzB2ta/nFyqryuxeACwOvBBQDu/XAvXtuc4/XXJaLuI+mwERH5hiWfZODz3wokfY1au1PS5+8uLV0kj4iUhZ0XIj+3L79c8uCiRMu/z8RpU+vzdYjIdzG8EPm593ae7PbnVIlO6O0WGC1ViKirhNFShTBrNUKtNQix1sJgq4PebqlfL+3DVnZwOTgR+RYOGxH5uZ3HOz+vJKrWhCtP7MDEwn0ItdZC57BB67AhSHR06Pz80Fg8O/xGZEaldLoGAFA5HXCq1F16jpZ8sCsPvaMNHD4iUhiuNiLyY1e/sAUHCz3crFQUMaQkB1fn/oJxpw9CLXZtLosDAj7uPwUfpF0Oh4cBJLamHEt/fQ8Xlp2EOciA0uBwlOvCUaoPR2lwOEr0RpQGG1ESHIESvREVulBAEDx6DQHAL8su4/JpIpl58vubnRciP7XuSJFHwcVgq8OU/F9xde4vSK5s3NrjQHQfrElJx8nweFjVQbCoguo/qzWwqYPgFFSAKEIFEYIoQmj4bLBb8IeD32Bq/h7MPfozRp7JxD9HzMWpsB5tVNHootJcPLTzbURY66+iG26rQbitBn3Q+saup0JjsXLobOyP7dfh77vpHktEpAwML0R+an3mmfYPApBiOo2rc3/BZfl7EOywAgBq1VqsTxqBb/uk44SxA9dEEQQ4IdS3MRqY1EH414ibsTP+Qvwp4zP0rziFFzc8i7cGXY1v+oxvs0NyxYmduGffagSJDuQYE/HMiLkAgOg6M6IaPmJqTYiuM7k/R9VVolfVWfxj26tYkzIWb110FWqCOhZIDFpO/yNSEg4bEfmpdUeKcPs7e1p9PL3wIK7L2YTBpbnu+/LCeuDbPuOwLmkkaoJa3jG6M6JrTVi092OMOHMUAPBrjwF47uIbUBZsbHacyunAgoPfYNbxrQCAzYlD8OzwG2HRtL/5qsFWi/89tAZXndgOACjRG7Fy2Gzsir+w3XMTjHpsXzbF02+LiLqRJ7+/GV6I/NRrm3Kw/PvMVh+/J+NzXHViOxyCCtsTLsI3fcZjf0xfj+eMdJQgOnHN8V/wv4e+hc5phznIgJXDZmNrz6EAgFBrDZbt/i+Gnz0GAHg3bRo+HDDV43oGl+Tgz3s/Rc/qEgDAhl4X47XBM2HShbZ53lu3jcCUgR3f8Z6IuhfDC8MLBbjTplqkL1/f5jHJ5iJMLNiH71PGovScDoiUkiqLcd+vH+ACU/21Z35OGoHv+ozDfXs+QGJ1KWrVWqwYcTN+SRzc6dfQ2a34n8wfcV32ZqghwqQNwctDrsPmXsNaPWf6RXF4dd7ITr8mEXUNtwcgCnC5JdXtHpMXHo/3Bk7zanABgPywOCy+9F582H8KHBAwNX8Pntu8EonVpSgOjsTiifd0KbgAgEWjxVuDrsHiS+9FbngCjNZq9K/Ib/OcII00HSci6n6csEvkh/rEhMhdQpvsKg3evfBK/BqXhr/u+QgJNaU4EJ2KJ0ff2u7wjieORibjT5P+jKuP/4LvU8a2eawxOKjbXpeIpMXwQuSHEozBmHZhD/x4uP0VRwPiQpFVXOWFqs53OLoP7rpsMfqX5+NQdB+PrwPTEXaVBl/2m9jucafKuFUAkVIwvBD5KUsHN0qUK7i41Gl0Hl2XRSp2hzI2liQiznkh8ls9wrtvqXMgqLZ1bMsDIpIfwwuRn5p2UZzcJShKjcUudwlE1EEML0R+qrCiTu4SFMXpVxeNIPJvnPNC5KfOVHo5vIhAqAg4ANgFwA5AVNDq495RBrlLIKIOYngh8lNTB8Zh5focr72eFsAfzc33EnJChB2uMCPC0RBqzr8tNgs9zW+LsAtAlUpETpB0k2rjjZwjRKQUDC9EfmpoUiQmD4jFhqyzXnk9tVgfNDRNdmdUQYAWgFYEAKF+C+dOKlQ7kRNk6WqZraqxcsIukVIwvBD5sXH9or0WXmpVwHMRdRBEQA1A4/4sQNNwWwMBahGt3Bbc551/W0C5StqlzFcNSZD0+Ymo+zC8EPmx0SlRXn9NscnQUMM9TR/1ej0dkRoTwk0ZiRSEq42I/NjQpEjMHt5T7jJ8ml4DrP/rJLnLICIPMLwQ+bl/3TAMX909DtEG7t3TksdndW0TSCLyPoYXogAwNCkSo1Ii5S7D5yRHBWPOyGS5yyAiDzG8EAWIOaOS5C4BABCm6/7NFztjVEoENv/tMrnLIKJOYHghChBTBsZjeHKE3GVgRLJvdIAeuupCuUsgok5ieCEKIKvvGo+3bhuB6y9OlK+INq66279HiFdKmD28J4Ym+UaIIiLPcak0UYCZMjAeUwbGY0xqNO7//IDXX9/maP1icLFhWhw9Uy3J66bFhWLOyCSMTIlkcCFSOIYXogB146hkTOwfi58PF6OkyoIfDhYhq7hK8tcNUrc+56WtxzyVHBWMR665EBuzzmLSgFhex4XIjzC8EAWwBGMw5qWnAAAWXT4A644U4bv9p9G3RyiuH94LK37Mwue/FXTra45IicTGoyUeP3au+DAdiiobtwsYnhyBm0cn4adDxbjiojj3KiKGFiL/I4ii6JuXvOwks9kMo9EIk8mE8PBwucshUrx9+eX4KqMQADBzWCIe/eYwfsuraPOctLhQZLbQxVk2Iw3XDk1E+vL1Hj02uGc4DhSY3bdnDIrHy/8zAuuOFLGzQuQnPPn9zfBCRB5rGhpKqqx4YPVBOBp+lNw2tjcenTUIH+/Ow9LPD7g3BFh2ZRruuLQvAHTqsX355fj1RDnnrBD5KYYXhhcirzptqsWJkhqkxBiQYAxu9/6uPEZE/onhheGFiIhIUTz5/c3rvBAREZGiMLwQERGRojC8EBERkaIwvBAREZGiMLwQERGRojC8EBERkaIwvFDAE51+dbUAIiK/x72NKGBV5ZfgzEsZEJ1ObKj9GNpgA7TBBugMzT83u99ggO7c+4KDodHqIAiC3N8SEVFAYHihgOSw27F21UqMwlRABZjPnIUIZ6efT6VWNwaa4OD6kGMIabgvuOF+Q5P7m9/nCkIMQURE7WN4oYC08d03cfzQHoxMuQyCoMKNy5bDJlhgra2Bpaam2edm99XWwFpTA2ttrfs2RBFOhwN1VZWoq6rsUl2CSnVeoGkMRa77g93ByN0FMjS/T6NjCCIi/yVZeHnyySfx3XffISMjA1qtFhUVFe2eI4oiHnnkEbzxxhuoqKjA+PHj8corr+CCCy6QqkwKQPt//gEZP35bf0OvAixAbFwKtAkhHj+X6HTCZqlrCDUNgaamuknQqW3ydTUsDaHHWnN+KIIoQnQ6UVddhbrq83dk9oQgqKA1NHR33KEmuCHoNAajxvBzzpBZw31BOj1DEBH5HMnCi9VqxZw5c5Ceno633nqrQ+c888wzeOGFF/DOO++gT58+ePjhhzFt2jQcPnwYer1eqlIpgOQfPoB1/3kFADD+xnkIOmmAvbgGzmorAM/Di6BSuTsjiOp8XaIowmapg7Wmsbtjaej6NA06591XW9sQilz310IUnRBFJyzV1bBUV6MSZztdlyCo6oe42gs6ru5PsAFaQ8h59wXpgxmCiKjbSBZeHn30UQDA22+/3aHjRVHE888/j4ceeggzZ84EALz77ruIi4vDl19+iZtuukmqUilAmM4U4etnl8PpcGDAuIkYc90NKHnjAOwAnFU2WWsTBAFafTC0+mCEIrrTzyOKIuwWCyznBJrm3Z+GYa+aFjpCruBUU9MYgmqqYampRpcGxAShyUTnJmHIENLYEWplsnTT+4J0eggqLpIkCnQ+M+clNzcXRUVFmDp1qvs+o9GIMWPGYPv27Qwv1CXW2hp8+czjqKs0Iy61H6bd+ScIggBVaBAAwFEtb3jpLoIgIEivR5Be3z0hqGmgaaEj5Pq66fyg+vsbO0Ki0wmIojsEdfEbrA957k5QcJOJ0QbozhkqazqButn9eoYgIiXzmfBSVFQEAIiLi2t2f1xcnPuxllgsFlgsFvdts9ksTYGkWKLTiTUvPouS/JMIiYjEzPseQpCufhhSFVIfXpx+El66S9MQhMjOj4eJogi71QJrbW2TidDVDUGn5fvcYahJR8hSU+0OQa6Q1LVZQXCv+Gpprk9r3Z/zOkF6PVQqdRcrISJPeRReli5din/84x9tHnPkyBGkpaV1qShPLF++3D1ERdSSbZ+8h5xfd0AdFISZf30IYVEx7sfUrvAi87CRvxIEAUE6PYJ0eoRERHb6eURRhN1mdQ9puVeBtTD5udUVYw3dIKfDAQANQakWQGmXvscgfXBDdyek2Yqw8yZLGwwoLDcjpf8ApA0Z2qXXJAp0HoWXJUuWYP78+W0ek5qa2qlC4uPjAQDFxcVISEhw319cXIxhw4a1et6yZcuwePFi922z2YykpKRO1UD+58i2Tdj5xScAgCvu+BMSLhjQ7HF/GzbyV4IgIEirQ5BW1+UQ5LDZ2u3+NA9F1Y33NZkT5HTYAQC2ulrY6mqB8rI2X9up1qC632DsOHgYN6tUGDBocKe/D6JA51F4iY2NRWxsrCSF9OnTB/Hx8Vi3bp07rJjNZuzcuRN//OMfWz1Pp9NBp9NJUhMpW1HOMfz0yr8BAKOunY0LL5l83jEcNgosgiBAo9VCo9V2SwhqrfvT0n25pipU2wFVbTWy1n7H8ELUBZLNecnLy0NZWRny8vLgcDiQkZEBAOjXrx9CQ0MBAGlpaVi+fDmuu+46CIKAv/zlL3jiiSdwwQUXuJdKJyYmYtasWVKVSX6qqrwMX614AnabFanDR2HCzbe2eByHjagzmoYggzGi3eOrqqqw99//BmCDruQ0jp04glNHDqLXwEGS10rkjyQLL3//+9/xzjvvuG9ffPHFAIANGzZg0qRJAICsrCyYTCb3MX/7299QXV2NhQsXoqKiAhMmTMAPP/zAa7yQR+xWK75a8QSqykoR1TMJM+69r9VJlapQLQAOG5G0tm3bBpvNhp49e6J3YjQOrPsRG95+A7csf5YTfok6QRBF0a+21DWbzTAajTCZTAgPD5e7HPIyURTx/UvP4siWDdCHhGLuU88iMj6x1eMd1TacfnwHAKDnk+MhqLl8lrpXZWUl/v3vf8Nut+OWW25Bz7ge+M+fF8JSU43LF96LIVOmyV0ikU/w5Pc3f1KTX9n99ec4smUDBJUK1yxe1mZwAQBVsAZouPCrs9ruhQop0Gzbtg12ux09e/ZEv379YAg3Iv13NwMAtn70btevfUMUgBheyG/k7NmFLR/WD1VOnr8QyYPaX44qqAT3pF0OHVF3q6ysxK+//goAmDx5snuLhGHTrkJkYi/Umk3Y/vlHcpZIpEgML+QXSvJPYs3KfwKiiCFTp2PYFVd1+NzGFUdWqcqjALV161bY7Xb06tULffv2dd+v1gRh8q1/AADs/f4blBUWyFUikSIxvJDi1Vaa8eU/H4e1tha9LhyEy35/p0ebAHLFEUnBbDa32HVx6XPxSPS5eCScDjs2/fdNOUokUiyGF1I0h92Ob59/GqbiIhh7xOGaRcug1ni2iI4XqiMpbN26FQ6HA0lJSa1evHPSrX+ASq3G8d92Izdjj5crJFIuhhdStI3vvoG8g/sRpA/GrPsehiHc6PFz8EJ11N1MJhP27KkPIy11XVyiEnvh4unXAAA2vvMGHHZOGifqCIYXUqx9a79Hxo/fAYKAGfcsQUxySqeeh8NG1N1cXZfk5GT06dOnzWPHzr4JwWHhKCs8hX0/feelComUjeGFFCn/0H6sX/UqAGDCjfPQb9TYTj8Xh42oO5lMJvz2228A2u66uOhDQjHhpvorQP/y2QeoMZvaPJ6IGF5IgSqKi/D1c0/D6XAgbfylGD1rTpeej8NG1J22bNkCh8OBlJSUdrsuLoMuuxyxvfvAUl2NXz55X+IKiZSP4YUUxVpbg6/++TjqKs2IS70AV9z5J49WFrWEw0bUXSoqKtxdF9c2KB2hUqkxef5CAMD+n3/A2ZO5UpRH5DcYXkgxRKcTa178F0ryTyIkMgoz73sQQdqu7yjO/Y2ou2zZsgVOpxN9+vRBSkqKR+cmXTgY/cdOgCg6seGdN+BnO7cQdSuGF1KMbZ+8h5xfd0IdFISZf30QYVEx3fK8rmEjsdYO0eHsluekwFNeXo69e/cC8Kzr0tTEW34PTZAW+Yf2I3vX9m6sjsi/MLyQIhzZtgk7v/gEADDtjj8hod+Abntu7m9E3cHVdUlNTUXv3r079RzGHnEYec11AICN/30Ldiuv+kzUEoYX8nlF2Ufx0yv/BgCMmvk7DLxkcrc+P/c3oq4qLy9HRkYGgM53XVxGz5yD0KhomM8WY893X3a5NiJ/xPBCPq2qrBRfrXgCdpsVqcNHYcJN8yR5He5vRF2xefNmOJ1O9O3bF8nJyV16riC9HhPnzgcA7PziE1SVlXZDhUT+heGFfJbNasFX/3oSVeVliO6VjBn33geVSi3Ja3HFEXVWWVlZt3VdXNImTEJC/zTYLHXundKJqBHDC/kkURSx9rWVKMo+Cn1oGGbd9zB0BoNkr8cL1VFnbd68GaIool+/fkhKSuqW5xQEAZfdVr90+vDm9Th9LKtbnpfIXzC8kE/a/fXnOLJ1IwSVCtcsWoqI+ARJX48XqqPOKC0txb59+wB0X9fFJb5ff1x06VQAwIa3X4fo5Eo4IheGF/I5OXt2uVvll82/A8mDhkr+mhw2os5wdV0uuOAC9OrVq9uff8LNtyJIH4zT2Vk4snVjtz8/kVIxvJBPKck/iTUr/wmIIoZefiWGTbvKK6/LYSPyVElJCfbv3w+g+7suLqGRURhz3Q0AgC0fvA1rXa0kr0OkNAwv5DNqK8348p+Pw1pbi6QLB2Py/Du89tocNiJPubou/fv3R8+ePSV7nREzZsIYF4+q8jLs+vIzyV6HSEkYXsgnOOx2fPPc0zAVF8HYIw5XL1oKtUbjtdfnsBF54uzZszhw4AAA6bouLhqtFpfOux0A8Ou3q2E6UyTp6xEpAcML+YQN77yB/EP7EaQPxqz7HoYh3OjV1+f+RuSJTZs2QRRFDBgwAImJiZK/Xr+RY5E8aCgcNhs2vfcfyV+PyNcxvJDs9q1dg30/fQcIAmbc+1fEJKd4vQbub0QddebMGRw8eBCA9F0XF0EQMPm2BRAEFY7t/AV5B/d75XWJfBXDC8kq/9B+rF/1GgBgwo3z0G/kGFnq4P5G1FGbNm0CAKSlpSEhQdol/E3FJKdgyOVXAgA2vvM6nE6H116byNcwvJBsKoqL8PWzy+F0OJA2/lKMnjVHtloElQCVgSuOqG1nzpzBoUOHAHiv69LU+BtugT4kFGfzTuDAup+8/vpEvoLhhWRhra3Bl888hrqqSsT3vQBX3PknCIIga02u5dLc34has3HjRgDAwIEDER8f7/XXDw4LR/qcWwAAWz/+L+qqqrxeA5EvYHghrxOdTqx58V8oPZWHkMgoXPvXBxGk1cldVuOKI3ZeqAXFxcU4fPgwAHm6Li5DL78S0b2SUVdpxvbPP5StDiI5MbyQ1239+L/I+XUn1EFBmPnXBxEWFSN3SQAaJ+06uFyaWuDqulx44YWIi4uTrQ61RoNJty0AAGT8+C1KT+XLVguRXBheyKuObN2IXV9+CgCYdsefkNBvgMwVNWocNmJ4oeaKiopw5MgRAPJ2XVxShlyMviPHwOlwYOO7b0AURblLIvIqhhfymqLso/jp1RcAAKNm/g4DL5ksc0XNcdiIWuPqulx00UXo0aOHvMU0uPR//hcqtQYn9v2G3L2/yl0OkVcxvJBXVJWV4ssVT8BusyJ1+ChMuGme3CWdh8NG1JLTp08jMzMTAHDppZfKXE2jyISeGD7jWgDAxnffhMPOf7cUOBheSHI2qwVfrXgC1eVliO6VjBn33geVSi13WefhsBG1xNV1GTRokM90XVzGXn8TDMYIlJ8uwN4fvpW7HCKvYXghSYmiiLWvrURRzjHoQ8Mw629/h85gkLusFnHYiM5VWFiIrKwsCILgU10XF53BgAk33woA2P7Zh6gxVchbEJGXMLyQpHZ//TmObN0IlVqNaxYtQ0Sc96+N0VEcNqJzNe26xMbGyltMKwZdOhVxqf1gra3B1o//K3c5RF7B8EKSydmzE1s+fAcAMHn+HUgeNETmitrm2pyR+xsRABQUFODo0aM+23VxEVQqTL5tIQDgwPqfUJybI3NFRNJjeCFJlOSfxHcvrABEEUMvn4FhV8yQu6R2NdvfqIb7GwU6V9dl8ODBiInxjWsRtaZn2oUYMG4iIIrY+A6XTpP/Y3ghSWz7+D3Y6mqRdOFgTJ6/UO5yOqTZ/kYcOgpop06dwrFjx3y+69LUxFvmQ6PV4dSRgzi6Y5vc5RBJiuGFJGGz1AEALpo0FWqNRuZqOo77GxHQ2HUZMmQIoqOj5S2mg8JjemDUtbMBAJveews2q0Xmioikw/BCktAZQgAAlpoamSvxDFccUX5+PrKzsxXVdXEZde31CIuORWXJWfy25mu5yyGSDMMLSUIX4govytr1liuOyNV1GTp0KKKiouQtxkNBOj2GTbsKQP0VrYn8FcMLScLdeamulrkSz/BCdYEtLy8POTk5UKlUmDhxotzldIrNUj9cFBIRIW8hRBJieCFJ6ENCAQCWGmWFFw4bBTZX12XYsGGK67q4VJeXAgBCIpVZP1FHMLyQJBTbeeGwUcA6efIkjh8/DpVKhUsuuUTucjqtqqw+vIRGKWOiMVFnMLyQJBQ754XDRgHL1XW5+OKLERkZKW8xXVBVXgYACI1keCH/xfBCkmgML1xtRL7vxIkTyM3NVXzXBWgaXjhsRP6L4YUkoTM0zHnhsBEpgKvrMnz4cEQoeKKrw25DrdkEgHNeyL9JFl6efPJJjBs3DgaDocM/DObPnw9BEJp9TJ8+XaoSSUKunaPrFDZhl/sbBZ7c3FycOHECarVa8V2X6vJyAIBao0FwWLjM1RBJR7JLn1qtVsyZMwfp6el46623Onze9OnTsWrVKvdtnU4nRXkkMfewUXUVRFGEIAgyV9Qx7v2NxPr9jdRhWrlLIgmJotis62I0GuUtqIuq3CuNohXz/xxRZ0gWXh599FEAwNtvv+3ReTqdDvHx8RJURN6kbxg2Ep1O2Cx10OqDZa6oY1z7GzmrbXBU2Rhe/Fxubi5OnjwJtVqNCRMmyF1Ol3G+CwUKn5vzsnHjRvTo0QMDBgzAH//4R5SWlrZ5vMVigdlsbvZB8tPodFCp1QAUOO+F+xsFhKZdlxEjRii+6wIAVWUMLxQYfCq8TJ8+He+++y7WrVuHf/zjH9i0aROuvPJKOByOVs9Zvnw5jEaj+yMpKcmLFVNrBEFocq0XZS2X5oqjwHD8+HHk5eX5TdcFaHKBOoVeYI+oozwKL0uXLj1vQu25H5mZmZ0u5qabbsK1116LwYMHY9asWfj222+xe/du919HLVm2bBlMJpP7Iz8/v9OvT93LNe9FcZN2ueLI7zXtuowcORLh4f4xudV9gTpe44X8nEdzXpYsWYL58+e3eUxqampX6jnvuWJiYpCdnY0pU6a0eIxOp+OkXh+l2OXSvFCd38vJyUF+fj40Go3fdF2AJnNeeHVd8nMehZfY2FjExsZKVct5Tp06hdLSUiQkJHjtNan7uDovVoV1Xjhs5N/O7bqEhYXJW1A34oRdChSSzXnJy8tDRkYG8vLy4HA4kJGRgYyMDFRVNc5/SEtLwxdffAEAqKqqwn333YcdO3bgxIkTWLduHWbOnIl+/fph2rRpUpVJElLstV44bOTXsrOzcerUKWg0GowfP17ucrpVdUN44QXqyN9JtlT673//O9555x337YsvvhgAsGHDBkyaNAkAkJWVBZOp/mqQarUa+/fvxzvvvIOKigokJibiiiuuwOOPP85hIYXisBH5mqZdl1GjRvlV18VWV+fexT2Mw0bk5yQLL2+//Xa713gRRdH9dXBwMH788UepyiEZNO5vpKzwwmEj/3Xs2DEUFBT4ZdfFdYG6IH0wtMEGmashkpZPLZUm/6JX6FJpDhv5p6Zdl9GjRyM0NFTegroZ57tQIGF4Ick0bhGgrM4L9zfyT0ePHkVhYSGCgoL8rusCMLxQYGF4IcnoQur/slXchF3X/kao39+IlO/crktIQ7D2J9Vlrn2NGF7I/zG8kGQar7CrrPDi2t8I4NCRv8jKysLp06eh1Woxbtw4ucuRhGvOC6/xQoGA4YUk0zhhV1lzXgDub+RPAqHrAjTd14jhhfwfwwtJxj1ht6ZG5ko8xxVH/iMzMxNFRUV+3XUBml5dl8NG5P8YXkgy2ibDRk2XxSsBVxz5B6fT6e66jBkzBgaD/y4h5gXqKJAwvJBk9A3teafDDrvVInM1nlGx8+IXMjMzUVxcDJ1Oh/T0dLnLkYwoiu5NGXmBOgoEDC8kmSB9MASh/p+Y0ibtqnmVXcULpK6Lpboadlv9/KyQCHZeyP8xvJBkBEFQ7FV23Z0XDhsp1pEjR3DmzBm/77oAjSuN9KFh0Gi1MldDJD2GF5KUK7zUKazz4p7zws6LIjXtuowdOxbBwcHyFiQxXqCOAg3DC0nKfa0XhS2X5rCRsh0+fBhnz56FTqfD2LFj5S5HcpysS4GG4YUkpVfqFgGcsKtYTbsu6enpft91AeCerMsL1FGgYHghSekM9VsEKDa81HB/I6U5dOgQSkpKoNfrA6LrAjS5ui4vUEcBguGFJKVtWOGhuAm7hiDub6RATqcTmzZtAlDfddHr9TJX5B2NV9flsBEFBoYXkpReoauNmu5vxKEj5Th48CBKSkoQHByMMWPGyF2O17jnvPDquhQgGF5IUkodNgKaXmWX+xspgcPhCMiuCwBUNgwbhXHYiAIEwwtJyr1UWmGdF4ArjpTm4MGDKC0tDbiui+h0svNCAYfhhSTlXipdrayl0gAvVKckTbsu48ePh06nk7ki76kxmyA6nYAgIMQYKXc5RF7B8EKSck3YPbl/r8yVeI4XqlOOAwcOoKysDAaDAaNGjZK7HK/a/fXn9V+IIgRBkLcYIi9heCFJHdmywf31dy/8E7WVZhmr8QyHjZQhULsuNksdfnjleez57kv3ffvX/SBfQURexPBCksnd+yuO7fzFfTtz2ya8veQuHN2xVcaqOo7DRsqwb98+lJeXIyQkJGC6LqWn8vD+A4txaOPPze7f9N4qmM+ekakqIu9heCFJmEvOYs2L/zrv/hpTBb557ml88+xy1JgqvF+YBzhs5PscDgc2b94MoL7rog2ATQkPbvwZ7y1bhNJTedCHhrnvD9IHw1ZXi59eXwlRFGWskEh6DC/U7Rx2G759/mnUVVUiLrUfbn/hTQiq+n9q/UaNhaBS4ejObVi15C4c2bbJZ3/QctjI92VkZKCiogIhISEYOXKk3OVIylZXh+9fehY/vvI87FYLeg+5GFP/cBcAwGCMwP8sfw6aIC1O7t97XkeGyN8wvFC32/LB2zh9LAs6QwiuWbQUEXHx6Dey/jLtIZHRuOWp5xDbuw/qKs1Y88I/8dWKJ9274voS7m/k2+x2u7vrMmHCBL/uupTkncB7y/6Cw5vXQxBUmHDTrZi97FGcyc0BAPQecjGiEnth3I3/AwDY+O6b7v2OiPwRwwt1q2O7fsGe774CAEy/axGMPeIBAEMvnwEAOLJlPSITEnHLU89i3A23QKXWIOfXHXh7yR9xaNM6n+rCqEKCUBW7F2bDbygr3Q6TeR+qqrJQU3MCFksxbDYTnE6LT9UcSDIyMmAymRAaGuq3XRdRFHFg/U94/4HFKCs8hdDIKNzw96cw5robIKhUOLGvfhVfytDhAIARV81EfL/+sNRUY+2bL/HfJvktjdwFkP+oKDqNH15+HgAw4urr0G9U46Z4yYOGIDIhEeWnC3Fky0YMvfxKpM++Gf1GpePHV55H8fFs/PDyc8j6ZTMuX3gvwqJj5PkmmlAZglA45CWIajsK9rV1pAC1OhgqlR4qlc79tVqtr/+s0kOldn0Obn67ncfcz9Hk+bkctr7rsmXLFgD1XZegoCCZK+p+1toa/PzmyziydSMAIGXYCFx592IYwo0AgOqKcpw5Ud95SRlyMQBApVJj2p1/xn/v/zOO79mFzG2bMHDCJDnKJ5IUwwt1C7vVim+eexrW2hok9h+IS26+rdnjgkqFoZfPwMZ338S+n77DkKnTIQgCYpNTMPeJf2H3N6ux/bMPkJuxB28vuQuXzvtfDL5smry/qAVAX5kKp8oCdawaTsECh6MOTmf9hyg6Gg4U4XDUwOGo8UpZqlaCjVqlOz8AqYPr7296nPrc0KSHWnVO4Gr4LAi+2Zzdu3evu+syYsQIucvpdqYzxfj8qb+j/HQBBFX9MNGoa653zx0DgLwDGQCAHil9YTBGuO+PSeqN9Nk3Ydsn72H9qteQPGgoQiJ48TryLwwv1C02vvsGzpzIQXBYOK7+y/1Qa87/p3XhpVOw9cN3cTbvBAqPZqLngIEAAJVajTGz5qDfyLH48dXncfpYFta+/iKytm/FFQvvhbFHnLe/HQCAIAhIOfQInNU2xP1lOILiQ9yPiaIIUbTB6XQFmlp3sHF/dtbC6bA0fK6Dw1kHZzuPOZxNn8P1nBaIYuO8G1d4snths2uVSguV6txA1DTguB7T1QegFgJRW4+p1br651DrIQjqDtXUtOtyySWX+GXXJSQiEkE6PUKjY3D1n/6GnmkXnnfMiX2/AQB6D734vMdGzfwdju76BWdPHMf6Va/hmkVLJa+ZyJsYXqjLjmzdiH1rvwcEATPu/WurQz7BoWEYMG4iDm36GfvWrnGHF5foXkm46bFn8Nuar7Hto/8i70AG1v3nFVy/9P+88F20TAiq/0tXtDmb3y8IEAQtVCotNJqwlk7tVk6nvSH0uEJObUMQssDhqG0IPLX1t92ByAKno7bhHNdjdQ33NX2seWgSRWuT17XC6bTCDpPk36MgaBvCTEO4aSH0qFR6lJVWIia2EDr9UAwfPlzyuuSg0Wpx7ZIHoA0ORnBY+HmPi04nTjRctTplyPnvgVqjwfQ//gXvP7AIR3dsxdGd29B/zHjJ6ybyFoYX6pLSU/lY+/qLAICx19/onjjYmqFXXIlDm37G0e1bMOnWP7jH711UKjVGXn0d+o4YjfWrXsPk+Qslq70jGsOLo50jpaVSaaBShUKDUMlfSxQd53WRmoUjd1iqaxKozglETR9vOK/xscbncDotTV7XCrvdCqCy3Rp79gT0uqv8suvi0lbH8WzeCdSYKhCk0yPxnD8CXHqkpGL0zN9hx+qPse6tV5B04eAWgxCREjG8UKfZ6urwzXPLYbPUIXnQEKT/7uZ2z4nv2x89+vTFmdwcHNr4M0ZdO7vF4yITemL2A491d8kea63z4s8EQQ2NJgRASLvHdpUoOs8LM+d3khrDkcNegw0bf4IoWjB1ymjJ6/NVriGjpIsGQ9MkwImiCIfdDpulDra6OlwwZjz2r/sRNaYKbHjnDcy4Z4lcJRN1K4YX6hRRFPHzmy+h9FQeQiIiMePe+6BStT9nQRAEDLviKvz02gvY9/P3GHn1dc0mIfoaIaj+ewqk8OJNgqCCWh0MtTq4Q8dXV1fjRO4pAEB8/DAJK/NtJ/fXh5fTx7Lw1p8XwGaxwFZXB5ulrn6H6RYc2bIBY2bdgOheSd4slUgSvvtbg3zawQ1rcXjLBgiCClf9+W8erWZIGzcROkMITMVF7nF7X+XuvNgZXnxBaWn9hdeMRqNfDxm1p9Zcv8FpbaUZFUWnUV1eBmttTbPgotZooA8JRWh0DCITeyF1+CiEx8TKVTJRt2LnhTxmLjmL9f951X37m+eehlqthkqjgUqthkrV5Gu1Giq1pvFxlQoqjQbqhl88+9auQZ9hvrvUVdAE3rCRLysrq78Sc1RUlMyVyGvW/Y+gODcbQVodgnR6BOn1CNI1/VoPlbpjq7eIlIjhhTxms9RBFOt/mYuiE7Xmzq9EKc451l1lSSIQ57z4MlfnJTo6WuZK5BUeE8suCgU0hhfyWHTPJPzxjfdRYzbB6XA0+bDDaXfA6XTUf3bY3V87HHaIDgccruMcTjgddvQccP71K3wJw4tvcYWXQO+8EAU6hhfqFJ0hBDqD9KtR5OYrS6WpnmvYKNA7L0SBjhN2idrA1Ua+QxRFDhsREQCGF6I2cdjId1RVVcFms0EQBERERMhdDhHJiOGFqA3u1UZcKi07V9clIiICmhb2ziKiwMHwQtQGdl58ByfrEpELwwtRGzhh13dwsi4RuTC8ELWBnRffwc4LEbkwvBC1geHFd7DzQkQuDC9EbRA0XCrtC5xOJ8MLEbkxvBC1gZ0X31BZWQm73Q6VSgWj0Sh3OUQkM8nCy4kTJ3D77bejT58+CA4ORt++ffHII4/AarW2eV5dXR3uvvtuREdHIzQ0FLNnz0ZxcbFUZRK1SdByqbQvcM13iYyMhJobDhIFPMnCS2ZmJpxOJ1577TUcOnQIzz33HF599VU88MADbZ63aNEifPPNN/j000+xadMmFBYW4vrrr5eqTKI2Ne4qzdVGcuJkXSJqSrIrPU2fPh3Tp093305NTUVWVhZeeeUVrFixosVzTCYT3nrrLXzwwQe47LLLAACrVq3CwIEDsWPHDowdO1aqcolaxGEj38D5LkTUlFfnvJhMpjb/ctqzZw9sNhumTp3qvi8tLQ3JycnYvn17i+dYLBaYzeZmH0TdhXsb+QbuaURETXktvGRnZ2PlypW44447Wj2mqKgIWq32vH1L4uLiUFRU1OI5y5cvh9FodH8kJSV1Z9kU4NydF7sToijKXE3gcnVeOGxEREAnwsvSpUshCEKbH5mZmc3OKSgowPTp0zFnzhwsWLCg24oHgGXLlsFkMrk/8vPzu/X5KbC5wgtEAA6GFzlwmTQRncvjOS9LlizB/Pnz2zwmNTXV/XVhYSEmT56McePG4fXXX2/zvPj4eFitVlRUVDTrvhQXFyM+Pr7Fc3Q6HXQ6XYfrJ/KEa8IuUD901PQ2eUdFRQWcTifUajXCw8PlLoeIfIDH4SU2NhaxsbEdOragoACTJ0/GiBEjsGrVKqhUbf/gHzFiBIKCgrBu3TrMnj0bAJCVlYW8vDykp6d7WipR16kFQAAgNqw4CuZuxt7WdMiovZ8hRBQYJPtJUFBQgEmTJiE5ORkrVqzA2bNnUVRU1GzuSkFBAdLS0rBr1y4AgNFoxO23347Fixdjw4YN2LNnD37/+98jPT2dK41IFoIgcMWRzLhMmojOJdmfkWvXrkV2djays7PRq1evZo+5Jj7abDZkZWWhpqbG/dhzzz0HlUqF2bNnw2KxYNq0aXj55ZelKpOoXUKQGqLVyfAiE853IaJzSRZe5s+f3+7cmJSUlPNWcOj1erz00kt46aWXpCqNyCPsvMiLy6SJ6FwcQCZqB8OLvLhMmojOxfBC1A73FgHc38jrHA4HysvLAbDzQkSNGF6I2tHYeeH+Rt5WXl4OURQRFBSEsLAwucshIh/B8ELUDg4byafpkJEgCDJXQ0S+guGFqB3c30g+XCZNRC1heCFqBzsv8uEyaSJqCcMLUTvcE3YZXryOy6SJqCUML0TtELScsCsXDhsRUUsYXojawaXS8rDZbDCZTADYeSGi5hheiNrBCbvycF3fRavVIiQkROZqiMiXMLwQtYMTduXRdLIul0kTUVMML0TtYHiRByfrElFrGF6I2sEr7MqDk3WJqDUML0Tt4FJpefAaL0TUGoYXonZw2Ege7LwQUWsYXoja4Q4vXCrtNVarFZWVlQDYeSGi8zG8ELWDS6W9zzVkpNfrYTAYZK6GiHwNwwtROzhs5H2c70JEbWF4IWoHw4v3cZk0EbWF4YWoHY1zXrhU2ls4WZeI2sLwQtQOLpX2Pg4bEVFbGF6I2tF0wq4oijJXExjYeSGitjC8ELXDNWwEEYCD4UVqdXV1qK6uBsDOCxG1jOGFqB3u8AIOHXmDa8jIYDBAr9fLXA0R+SKGF6L2qAWgYVNjhhfpcaUREbWH4YWoHYIgNJm0yxVHUuNkXSJqD8MLUQcIWm4R4C2crEtE7WF4IeoAQcMtAryFnRciag/DC1EH8Cq73sPOCxG1h+GFqAMYXryjpqYGtbW1ABheiKh1DC9EHdAYXjhhV0quIaPQ0FDodDqZqyEiX8XwQtQB3CLAO7hMmog6guGFqAM4bOQdnKxLRB3B8ELUAY07SzO8SImTdYmoIxheiDqg6eaMJB12XoioIxheiDqAw0bSE0WRnRci6hCGF6IOYHiRXnV1NSwWCwCGFyJqG8MLUQdwqbT0XENGRqMRQUFBMldDRL6M4YWoA7hUWnocMiKijmJ4IeoATtiVHifrElFHMbwQdYDTWj9c5OrAUPdj54WIOoo/iYk6wF5Sv9+OJkYvcyX+i50XIuoohheiDrCXNoSX6GCZK/FfFRUVAIDIyEh5CyEin8fwQtQOURSbdF4YXqRgs9lQV1cHAAgLC5O5GiLydQwvRO1wVtsg1tXPedFEc9hICpWVlQAAjUYDvZ7vMRG1jeGFqB320vqOgNqoc686ou5VVVUFAAgNDYUgCDJXQ0S+juGFqB2crCs9V+eFQ0ZE1BGShZcTJ07g9ttvR58+fRAcHIy+ffvikUcegdVqbfO8SZMmQRCEZh933nmnVGUStYvzXaTH8EJEntBI9cSZmZlwOp147bXX0K9fPxw8eBALFixAdXU1VqxY0ea5CxYswGOPPea+bTAYpCqTqF1caSS9psNGRETtkSy8TJ8+HdOnT3ffTk1NRVZWFl555ZV2w4vBYEB8fLxUpRF5hJ0X6bHzQkSe8OqcF5PJ1KGrZ77//vuIiYnBoEGDsGzZMtTU1LR6rMVigdlsbvZB1F3ql0nXT9hleJEOOy9E5AnJOi/nys7OxsqVK9vtusydOxe9e/dGYmIi9u/fj/vvvx9ZWVlYvXp1i8cvX74cjz76qBQlE8FZZYNodQACoInihF2psPNCRJ7wuPOydOnS8ybUnvuRmZnZ7JyCggJMnz4dc+bMwYIFC9p8/oULF2LatGkYPHgwbrnlFrz77rv44osvkJOT0+Lxy5Ytg8lkcn/k5+d7+i0Rtco1ZKSO1HNfIwkxvBCRJzzuvCxZsgTz589v85jU1FT314WFhZg8eTLGjRuH119/3eMCx4wZA6C+c9O3b9/zHtfpdNDpdB4/L1FHuOe78OJ0krHb7aitrX+fOWxERB3hcXiJjY1FbGxsh44tKCjA5MmTMWLECKxatQoqled/uWZkZAAAEhISPD6XqKvcK40430UyrvkuKpWKKwuJqEMk64MXFBRg0qRJSE5OxooVK3D27FkUFRWhqKio2TFpaWnYtWsXACAnJwePP/449uzZgxMnTuDrr7/GrbfeiokTJ2LIkCFSlUrUqsbOC8OLVFzhJSwsjFfXJaIOkWzC7tq1a5GdnY3s7Gz06tWr2WOiKAKo34wtKyvLvZpIq9Xi559/xvPPP4/q6mokJSVh9uzZeOihh6Qqk6hNXGkkPdd8Fw4ZEVFHSRZe5s+f3+7cmJSUFHeQAYCkpCRs2rRJqpKIPCKKIoeNvICTdYnIU1w+QdQKp9kK0eYEVIAmkpPCpcJrvBCRpxheiFphc813idRDUPN/Famw80JEnuJPZKJWcMjIOxheiMhTDC9ErXBP1uVKI0lx2IiIPMXwQtQKbsjoHey8EJGnGF6IWsHwIj2Hw4Hq6moA7LwQUccxvBC1QHSKsJdxawCpuYKLIAgICQmRuRoiUgqGF6IWOEwWwC4CagHqCIYXqTS9QF1ntg8hosDEnxZELXAPGUXpIah5yXqpcLIuEXUGwwtRC9zLpLnSSFKcrEtEncHwQtQC7mnkHU03ZSQi6iiGF6IWNK404nwXKXFTRiLqDIYXohZw2Mg7OGxERJ3B8EJ0DtEhwl7WMGwUy/AiJU7YJaLOYHghOoejog5wiIBGBXU4d5OWEjsvRNQZDC9E57CXuvY00kNQcZm0VJxOJyfsElGnMLwQncM9WZfzXSRVU1MDURQBgFfXJSKPMLwQnYN7GnmHa8goJCQEarVa5mqISEkYXojO4V5pxGXSkuKQERF1FsML0Tk4bOQdvMYLEXUWwwtRE6LDCXt5/YTdIA4bSYorjYiosxheiJqwl1sAJyAEqaAK18pdjl/jNV6IqLMYXoiaaDpkJAhcJi0ldl6IqLMYXoia4J5G3sMJu0TUWQwvRE00rjTifBepccIuEXUWwwtRE1xp5B2iKLLzQkSdxvBC1AQvUOcdtbW1cDgcANh5ISLPMbwQNRDtTjgqLAAYXqTmGjIKDg6GRqORuRoiUhqGF6IG9rI6QAQErRqq0CC5y/FrHDIioq5geCFq0HSlEZdJS4uTdYmoKxheiBpwvov38BovRNQVDC9EDbhM2ns4bEREXcHwQtSAy6S9h8NGRNQVDC9EDewl9RsysvMiPQ4bEVFXMLwQARBtDjhMXCbtLRw2IqKuYHghAmAvre+6CHoNVAZed0RKoihy2IiIuoThhQhcJu1NFosFdrsdADsvRNQ5DC9E4Eojb3J1XXQ6HYKCeDFAIvIcwwsRmkzW5UojyXGyLhF1FcMLEQBbw7BREDsvkuNkXSLqKoYXInDYyJs4WZeIuorhhQKe0+qA02wFAGii9TJX4//YeSGirmJ4oYDnWmmkMmigMnACqdQ454WIuorhhQIeh4y8i8NGRNRVDC8U8LjSyLs4bEREXcXwQgGv8QJ1DC/ewM4LEXWVpOHl2muvRXJyMvR6PRISEjBv3jwUFha2eU5dXR3uvvtuREdHIzQ0FLNnz0ZxcbGUZVKAq8ssA1B/dV2SlsVigdVaPzmanRci6ixJw8vkyZPxySefICsrC59//jlycnLwu9/9rs1zFi1ahG+++QaffvopNm3ahMLCQlx//fVSlkkBzHamBs5qGwDAUWWTuRr/5xoy0mq10Ol0MldDREol6Q50ixYtcn/du3dvLF26FLNmzYLNZmvxsuAmkwlvvfUWPvjgA1x22WUAgFWrVmHgwIHYsWMHxo4dK2W5FIDqjpW7v67cdAohw+OgCubGjFLhkBERdQevzXkpKyvD+++/j3HjxrW6n8mePXtgs9kwdepU931paWlITk7G9u3bvVUqBQjR5kTlxlPu206zFeVfHIMoijJW5d84WZeIuoPk4eX+++9HSEgIoqOjkZeXh6+++qrVY4uKiqDVahEREdHs/ri4OBQVFbV4jsVigdlsbvZB1BHVu07DWWmFOkKH2DuHACoBtftLUPPbGblL81vHjh0DAJSWlspcCREpmcfhZenSpRAEoc2PzMxM9/H33Xcf9u7di59++glqtRq33nprt/5lu3z5chiNRvdHUlJStz03+S/R5oR5U33XJWxSEnQpRoRf3hsAUPFVtnuvI+pe+/btA9DYgSEi6gyPB/eXLFmC+fPnt3lMamqq++uYmBjExMSgf//+GDhwIJKSkrBjxw6kp6efd158fDysVisqKiqadV+Ki4sRHx/f4mstW7YMixcvdt82m80MMNSu6t1FcJqtUBt1CBkZBwAIu7QXLMfKYTluQtlHmehx51AIGl5NoLucOdPY0Ro5cqSMlRCR0nkcXmJjYxEbG9upF3M6nQDqh3paMmLECAQFBWHdunWYPXs2ACArKwt5eXkthh0A0Ol0XLVAHqmf65IPAAib3MsdUASVgMgbB6D4+d9gO1UF89qTMF7ZR85S/cqWLVvcXw8YMEDGSohI6ST7s3Lnzp148cUXkZGRgZMnT2L9+vW4+eab0bdvX3cQKSgoQFpaGnbt2gUAMBqNuP3227F48WJs2LABe/bswe9//3ukp6dzpRF1m+pfi+AwW6E2ahEysnlHT2PUIWr2BQCAys2nUJdd3tJTkIfKyspw8OBB9+2oqCgZqyEipZMsvBgMBqxevRpTpkzBgAEDcPvtt2PIkCHYtGmTu1Nis9mQlZWFmpoa93nPPfccrr76asyePRsTJ05EfHw8Vq9eLVWZFGBEuxOVGxq6LpOSWhwWCh4Ug5DR8YAIlH1yFI5qXv+lq7Zu3eqe6yYIAoxGo8wVEZGSCaKfrQs1m80wGo0wmUwIDw+XuxzyMVXbC1HxVQ7U4VrE/21Uq3NanFYHzqzcC/vZWugvjEb0vIEQBMHL1fqHiooKvPDCC+5h44iICPzlL3+Rtygi8jme/P7mbEQKGKK96VyXlrsuLiqtGlE3pwFqAXWHS1G987S3yvQ7v/zyizu4ABwyIqKuY3ihgFH9axEcJivU4efPdWmJNjHUPWG34ttc2IqrpS7R71RWVuK3334DAPcqwMjISDlLIiI/wPBCAaN6R333RBUSBGtBZYeuNxQ6LhG6/pGA3YmyDzMh2pztnkONtm/fDrvdjl69ernnubDzQkRdxfBCAUPXr/4vftvpapx9dT/OvrwPNfvPQnS0HmIElYCoOf2hCg2CragGpu9zvVWu4tXU1GD37t0AgIkTJ6KsrH73boYXIuoqhhcKGBFXpyJu0XCEjIoHNAKs+ZUo+yATRSt2o3JrAZwWe4vnqcO0iJrTHwBQ9UshajPLvFm2Yu3YsQM2mw3x8fG44IILUF5ev+ycw0ZE1FUMLxRQguJCEDn7AiTcPxphU5KhCtHAUW6B6dvjOP3ULlSsOQ57xfkXUdQPiELo+EQAQPmnR+GotHq7dEWpq6vDzp07AdR3XUpKSlBbW7/lAsMLEXUVwwsFJHWYFsbLeyNh6WhEXNcPmthgiBYHqjYXoOiZ3Sj9KBPWU5XNzjFe2QdBCSFwVttQ9kkWRKdfXWWgW+3evRsWiwUxMTHo3bs3PvzwQwBAnz59eEVsIuoyhhcKaEKQGqFjEhC3aASib7sQulQj4BRRm3EWZ17MwJnX9qP2cClEpwhBo0LUTQMgBKlgOVaBqm0Fcpfvk6xWK7Zv3w4AGD9+PD777DOUlZXBaDS6t/0gIuoKj/c2IvJHgkpA8MBoBA+MhrWgClVbTqFmfwmsuSaU5pqgiQlG6ISeMAzvAePVqaj4IhumH05AlxoBbc9Qucv3KXv27EFNTQ0iIyORl5eH3NxcaLVazJ07F6GhfK+IqOvYeSE6h7ZnKKJuSkP8/aMQOrEXBL0a9pJaVHyZjaKnd8FhskDbOxxwiCj7KBNOq0Pukn2GzWbDtm3bAABBQUHYu3cvAGD27NmIi4uTszQi8iMML0St0Bh1iJjRBwnLRsN4dSrUkTo4a+yoXJ8P60kzAMB+thamb4/LXKnvyMjIQFVVFQDgzJkzAIArrriCu0gTUbfisBFRO1Q6DcIm9ERoeiJqD5egaksBrHmNk3mrdxVBd0EkDINjZKxSfg6HA1u3bm123/Dhw927yBMRdReGF6IOEtQCDINjYRgcC8tJM6q2nELtoVJABKq3FwZ8eNm/fz9MJpP7dkpKCmbMmMENLYmo2zG8EHWCrnc4dL0vhL20FjX7ztavUgpgTqcTW7Zscd+OiorCDTfcAI2GP2KIqPvxJwtRF2iigxF+WbLcZcju0KFD7sv/6/V6zJ07FwaDQeaqiMhfccIuEXVJ066LIAiYM2cOYmICewiNiKTF8EJEXZKTk+NeWTRjxgz07dtX5oqIyN9x2IiIukSlUkEQBKSnp2PUqFFyl0NEAUAQRdGvNmgxm80wGo0wmUwIDw+XuxyigGC32zk5l4i6xJPf3xw2IqIuY3AhIm9ieCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRdHIXUB3E0URAGA2m2WuhIiIiDrK9Xvb9Xu8LX4XXiorKwEASUlJMldCREREnqqsrITRaGzzGEHsSMRREKfTicLCQoSFhUEQBLnLkYXZbEZSUhLy8/MRHh4udzl+g++rdPjeSoPvqzT4vkpDFEVUVlYiMTERKlXbs1r8rvOiUqnQq1cvucvwCeHh4fwfSwJ8X6XD91YafF+lwfe1+7XXcXHhhF0iIiJSFIYXIiIiUhSGFz+k0+nwyCOPQKfTyV2KX+H7Kh2+t9Lg+yoNvq/y87sJu0REROTf2HkhIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4UaiXXnoJKSkp0Ov1GDNmDHbt2tXqsW+88QYuueQSREZGIjIyElOnTm3z+EDmyfva1EcffQRBEDBr1ixpC1QwT9/biooK3H333UhISIBOp0P//v2xZs0aL1WrHJ6+r88//zwGDBiA4OBgJCUlYdGiRairq/NStcqwefNmXHPNNUhMTIQgCPjyyy/bPWfjxo0YPnw4dDod+vXrh7ffflvyOgOaSIrz0UcfiVqtVvzPf/4jHjp0SFywYIEYEREhFhcXt3j83LlzxZdeekncu3eveOTIEXH+/Pmi0WgUT5065eXKfZun76tLbm6u2LNnT/GSSy4RZ86c6Z1iFcbT99ZisYgjR44UZ8yYIW7dulXMzc0VN27cKGZkZHi5ct/m6fv6/vvvizqdTnz//ffF3Nxc8ccffxQTEhLERYsWebly37ZmzRrxwQcfFFevXi0CEL/44os2jz9+/LhoMBjExYsXi4cPHxZXrlwpqtVq8YcffvBOwQGI4UWBRo8eLd59993u2w6HQ0xMTBSXL1/eofPtdrsYFhYmvvPOO1KVqEideV/tdrs4btw48c033xRvu+02hpdWePrevvLKK2JqaqpotVq9VaIiefq+3n333eJll13W7L7FixeL48ePl7ROJetIePnb3/4mXnTRRc3uu/HGG8Vp06ZJWFlg47CRwlitVuzZswdTp05136dSqTB16lRs3769Q89RU1MDm82GqKgoqcpUnM6+r4899hh69OiB22+/3RtlKlJn3tuvv/4a6enpuPvuuxEXF4dBgwbhqaeegsPh8FbZPq8z7+u4ceOwZ88e99DS8ePHsWbNGsyYMcMrNfur7du3N/vvAADTpk3r8M9k8pzfbczo70pKSuBwOBAXF9fs/ri4OGRmZnboOe6//34kJiae9z9bIOvM+7p161a89dZbyMjI8EKFytWZ9/b48eNYv349brnlFqxZswbZ2dm46667YLPZ8Mgjj3ijbJ/Xmfd17ty5KCkpwYQJEyCKIux2O+6880488MAD3ijZbxUVFbX438FsNqO2thbBwcEyVea/2HkJME8//TQ++ugjfPHFF9Dr9XKXo1iVlZWYN28e3njjDcTExMhdjt9xOp3o0aMHXn/9dYwYMQI33ngjHnzwQbz66qtyl6ZoGzduxFNPPYWXX34Zv/32G1avXo3vvvsOjz/+uNylEXmEnReFiYmJgVqtRnFxcbP7i4uLER8f3+a5K1aswNNPP42ff/4ZQ4YMkbJMxfH0fc3JycGJEydwzTXXuO9zOp0AAI1Gg6ysLPTt21faohWiM/9mExISEBQUBLVa7b5v4MCBKCoqgtVqhVarlbRmJejM+/rwww9j3rx5+MMf/gAAGDx4MKqrq7Fw4UI8+OCDUKn492xnxMfHt/jfITw8nF0XifBfqsJotVqMGDEC69atc9/ndDqxbt06pKent3reM888g8cffxw//PADRo4c6Y1SFcXT9zUtLQ0HDhxARkaG++Paa6/F5MmTkZGRgaSkJG+W79M68292/PjxyM7OdgdCADh69CgSEhIYXBp05n2tqak5L6C4AqLIbe46LT09vdl/BwBYu3Ztmz+TqYvknjFMnvvoo49EnU4nvv322+Lhw4fFhQsXihEREWJRUZEoiqI4b948cenSpe7jn376aVGr1YqfffaZePr0afdHZWWlXN+CT/L0fT0XVxu1ztP3Ni8vTwwLCxPvueceMSsrS/z222/FHj16iE888YRc34JP8vR9feSRR8SwsDDxww8/FI8fPy7+9NNPYt++fcUbbrhBrm/BJ1VWVop79+4V9+7dKwIQn332WXHv3r3iyZMnRVEUxaVLl4rz5s1zH+9aKn3fffeJR44cEV966SUulZYYw4tCrVy5UkxOTha1Wq04evRocceOHe7HLr30UvG2225z3+7du7cI4LyPRx55xPuF+zhP3tdzMby0zdP39pdffhHHjBkj6nQ6MTU1VXzyySdFu93u5ap9nyfvq81mE//v//5P7Nu3r6jX68WkpCTxrrvuEsvLy71fuA/bsGFDiz8zXe/lbbfdJl566aXnnTNs2DBRq9WKqamp4qpVq7xedyARRJG9QiIiIlIOznkhIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJFYXghIiIiRWF4ISIiIkVheCEiIiJF+X8sqAeqOM8kOAAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def animation_frame(f):\n",
    "    frame = sign[sign.frame==f]\n",
    "    left = frame[frame.type=='left_hand']\n",
    "    right = frame[frame.type=='right_hand']\n",
    "    pose = frame[frame.type=='pose']\n",
    "    face = frame[frame.type=='face'][['x', 'y']].values\n",
    "    lx, ly = get_hand_points(left)\n",
    "    rx, ry = get_hand_points(right)\n",
    "    px, py = get_pose_points(pose)\n",
    "    ax.clear()\n",
    "    ax.plot(face[:,0], face[:,1], '.')\n",
    "    for i in range(len(lx)):\n",
    "        ax.plot(lx[i], ly[i])\n",
    "    for i in range(len(rx)):\n",
    "        ax.plot(rx[i], ry[i])\n",
    "    for i in range(len(px)):\n",
    "        ax.plot(px[i], py[i])\n",
    "    plt.xlim(xmin, xmax)\n",
    "    plt.ylim(ymin, ymax)\n",
    "\n",
    "## These values set the limits on the graph to stabilize the video\n",
    "xmin = sign.x.min() - 0.2\n",
    "xmax = sign.x.max() + 0.2\n",
    "ymin = sign.y.min() - 0.2\n",
    "ymax = sign.y.max() + 0.2\n",
    "\n",
    "fig, ax = plt.subplots()\n",
    "l, = ax.plot([], [])\n",
    "animation = FuncAnimation(fig, func=animation_frame, frames=sign.frame.unique())\n",
    "\n",
    "HTML(animation.to_html5_video())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ASL",
   "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": "3.10.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}