Aerial Cactus Identification (1).ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Aerial Cactus Id Comp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai import *\n",
    "from fastai.vision import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "PosixPath('/home/jupyter/.fastai/data/AerialCactus')"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "path = Config.data_path()/ 'AerialCactus'\n",
    "#path.mkdir(parents=True, exist_ok=True)\n",
    "path"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Getting the Data from Kaggle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#! kaggle competitions download -c aerial-cactus-identification -p {path}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#! unzip -q -n {path}/train.zip -d {path}\n",
    "#! unzip -q -n {path}/test.zip -d {path}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train = pd.read_csv(path/'train.csv')\n",
    "#df_train.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading in Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "tfms = get_transforms(flip_vert=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "?? get_transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "src = (ImageList.from_csv(path, 'train.csv', folder='train')\n",
    "                 .split_by_rand_pct(0.2)\n",
    "                 .label_from_df())\n",
    "data =          (src.transform(tfms, size=16)\n",
    "                 .add_test_folder('test')\n",
    "                 .databunch().normalize(imagenet_stats) )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#data.show_batch(rows = 3, size = (7,6))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Creating and Training a model "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = cnn_learner(data, models.resnet34, metrics = accuracy )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:33 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.357283</td>\n",
       "      <td>0.271261</td>\n",
       "      <td>0.890286</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.241986</td>\n",
       "      <td>0.171896</td>\n",
       "      <td>0.930857</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.212432</td>\n",
       "      <td>0.147447</td>\n",
       "      <td>0.941143</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.188152</td>\n",
       "      <td>0.140474</td>\n",
       "      <td>0.946286</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 01:35 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.049032</td>\n",
       "      <td>0.042120</td>\n",
       "      <td>0.984000</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.061752</td>\n",
       "      <td>0.050960</td>\n",
       "      <td>0.979143</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.072646</td>\n",
       "      <td>0.050031</td>\n",
       "      <td>0.982571</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.064995</td>\n",
       "      <td>0.057849</td>\n",
       "      <td>0.983143</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.062095</td>\n",
       "      <td>0.044915</td>\n",
       "      <td>0.982286</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.040470</td>\n",
       "      <td>0.030220</td>\n",
       "      <td>0.990000</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.041271</td>\n",
       "      <td>0.031187</td>\n",
       "      <td>0.989143</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.033894</td>\n",
       "      <td>0.029032</td>\n",
       "      <td>0.990286</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEKCAYAAAA4t9PUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzt3Xl4VOXZ+PHvnZ0kkBBI2BLWsIVdAoq7IIobtG7F1qqVVmxfuvdttf2pXexiW+vbWqul7kvr1lpRUVERFUQkCGFfQlgSIAuEhCxkv39/zAmOMfvMyUyS+3Ndc2XmzHPOeeYQ5s55lvsRVcUYY4zpqJBAV8AYY0zXZoHEGGOMTyyQGGOM8YkFEmOMMT6xQGKMMcYnFkiMMcb4xAKJMcYYn1ggMcYY4xMLJMYYY3wSFugKdIb+/fvr8OHDA10NY4zpUjZs2HBUVRNbK9cjAsnw4cPJyMgIdDWMMaZLEZEDbSlnTVvGGGN8YoHEGGOMTyyQGGOM8YkFEmOMMT6xQGKMMcYnFkiMMcb4xAKJMcYYn1ggMcaYILXjyAnWZR8LdDVaZYHEGGOCUEV1LTc++jE3Pbae/BOVga5OiyyQGGNMEFr6fjYFpVXU1NXzpxW7A12dFlkgMcaYIFNwopK/v5fNZZMG8bWzhvP8hhx25p0IdLWaZYHEGGOCzL0rdlNbX8+P541lyQWj6RMVzm+X7wx0tZplgcQYY4LIjiMneH5DDjfOGs6wfjHERYfz7dmpvLe7kA/2FAa6ek2yQGKMMUHkN8t30CcqnCWzU09t++qsYaQk9OI3y3dSV68BrF3TLJAYY0yQ8Nx1HOXbs1OJj444tT0yLJQfXzyOHUdO8NLGQwGsYdMskBhjTBCoq1d+89oOhvWL5oZZwz/3/uWTBzElJZ4/vrmLk9V1nV/BFlggMcaYIPBCRg678kv5ybxxRIR9/qtZRPjZpePJO1HJo2v2BaCGzbNAYowxAVZeVcu9b+1m+rC+XDJxYLPlZo5IYG7aAB5ctZejZVUtHrOqto6POmlWvAUSY4wJsL+/n01haRU/u2w8ItJi2dsuGcfJmjr+/PaeJt8vqajhgXezOPued7n+4XWdMiu+R6zZbowxwSqvpJKl7+/l8smDOG1o31bLj0qM5cszh/LPjw9y01nDGZUYC8Ch4pM8unofz358kPLqOs4Z3Z/F544iqXek2x/B3TsSEZknIrtEJEtEbmvi/UgRec55f52IDHe2zxSRTc4jU0S+2NZjGmNMV3Lvil3U18NP5o1r8z7fvXA0vcJDuef1nWw/fILvPbuRc3//Lo9/uJ+LJgxk+XfO4alFp3P26P6t3uH4g2t3JCISCjwAzAVygfUiskxVt3sVWwQcV9VUEVkI3AN8CdgKpKtqrYgMAjJF5BVA23BMY4zpErYdLuHFT3L5xjkjSUmIbvN+/WMjufW8kfxxxW5WbM8nJiKUm84czs1nj2BIfC8Xa9w0N5u2ZgJZqpoNICLPAgsA7y/9BcDPnecvAn8VEVHVCq8yUXgCSFuPaYwxQU9V+cWy7fSNjuB/zk9tfYdGFp09kl35ZYwf1JuvzBxGXHS4C7VsGzcDyRAgx+t1LnB6c2Wcu48SoB9wVEROBx4FhgFfdd5vyzEBEJFbgFsAhg4d6vunMcYYP1qWeZiP9xfxuysndSgI9IoI5f7rprlQs/Zzs4+kqYa5xnP7my2jqutUdQIwA7hdRKLaeEyc/ZeqarqqpicmJraj2sYY467yqlp+s3wHk4bEcU16SqCr4zM3A0ku4H2FkoHDzZURkTAgDijyLqCqO4ByYGIbj2mMMUHt/pVZ5J+o4hcLJhAa4n5nuNvcDCTrgdEiMkJEIoCFwLJGZZYBNzrPrwZWqqo6+4QBiMgwYCywv43HNMaYoLXvaDmPrM7mqtOS2zTctytwrY/E6dNYArwJhAKPquo2EfklkKGqy4BHgKdEJAvPnchCZ/ezgdtEpAaoB76lqkcBmjqmW5/BGBM8isqrKSitZNzAPoGuik9++co2IsNC+cklYwNdFb8R1eBLSexv6enpmpGREehqGGM6aG9hGTc88jEFpZUs/845jB7QO9BV6pB3duSz6IkM/t9l4/n6OSMDXZ1WicgGVU1vrZylSDHGBLXMnGKueWgtlTV1xESGcdt/tlAfhGtytKaypo5fvrqdUYkxTWb37coskBhjgtYHewq57h8fER0RyovfPJM7Lktjw4HjPPPxwUBXrd0eWb2PA8cq+Pn8CU1m9+3KutenMcZ0G69uPszNj69naEI0//7mmYzoH8OVpw3h7NT+/P71neSVuJ+M0F8OF5/kryuzuHjCAM4Z3f2mI1ggMcYEnafW7ufb/9rItJS+PLd4FgP6RAGeNTl+/cWJVNfVc9eyrYGtZDv89vWd1Kvy/y5LC3RVXGGBxBgTNFSV+97azR0vb2POuCSeXDSTuF6fnfU9rF8M37twDG9uy+eNrXkBqmnbfZR9jFcyD7P4vFHtyqfVlVggMcYEhZq6eu54eSt/fmcPV09P5qHrpxMVHtpk2a+fM4Lxg/pw17KtnKis6eSats9vlu9gSHwvvnneqEBXxTUWSIwxAbf1UAnz/7qGpz86yOLzRvKHqycTFtr811N4aAi/u3IShaVV/P6NnZ1Y0/Yprqhmc24JXz59KL0img6K3YEFEmNMwFTV1nHvil184YE1HC2rYulXp3P7Ja2vEggwJSWem84cwdMfHSRjf1Gr5QNhU04xANNS4gNcE3dZIDHGBMSmnGKuuH8196/MYv7Uwbz1/XO5aELz65U35YcXjWFIfC9u/88WqmrrXKppx23KKUYEJiXHBboqrrJAYozpVJU1dfx2+Q6u/NsaTpys5bGbZvCna6cSHx3R7mPFRIZx9xcmsqegjL+/l+1CbX2TmVPM6KRYekcFbq2QzmBrthtjOs2GA8f53xczyS4sZ+GMFH562Xj6+Pgle8G4JK6YMpi/rszi0kmDSE2K9VNtfaOqbMopZm7agEBXxXV2R2KMcZ2q8vAH2Vz797VU1dTz1KKZ/O6qyT4HkQZ3Xp5GdGQoS/75CWVVtX45pq8OFlVwvKKGKd28fwQskBhjXFZaWcO3nvmEu1/bwZxxSbz+vXP8Prs7sXck9183jT0FZXzv2U1BkYuroaN9qgUSY4zpuJ15J5j/1zWs2J7PTy8dx9+/Ot1vdyGNnTM6kTsvT+PtHfn8ccUuV87RHptyiukVHsrYLpqpuD2sj8QY44r/fJLLT1/aQu+ocP759dM5fWQ/1895w6xh7Mov5W+r9jJmQG++MG2I6+dszqacYiYNiWtxPkx3YYHEGONXDenS/7nuIKePSOD+L08jqXdUp5xbRPjF/AnsLSjjx//ezPD+MQFpWqqurWfb4RPcOGtYp587ELp/qDTGdJqjZVVc89Ba/rnuILeeN4pnvn56pwWRBuGhITx4/XQG9InkG09mcKTkZKeeH2DHkRNU19YzNaV7LKXbGgskxhi/efLD/Ww7XMLSr07ntkvGBaxZJyEmgkdunEFFVS23PLmBk9WdO1nxVEf70O7f0Q4WSIwxfrRiez4zhie0e4a6G8YM6M1frpvG1sMl/OjFTDpzWfHMnGISe0cyOK5z78YCxQKJMcYvDhwrZ2deaVAEkQZzxg/gJ/PG8drmI9y/MqvTzrspp5ipKfFtyhnWHVggMcb4xYpt+QBcFGQzuRefO5Irpw3hT2/tZl32MdfPV1JRQ/bR8h4xf6SBBRJjjF+s2J5H2qA+Qbd4k2dVxUkMie/FnS9vo6au3tXzbcrtORMRG1ggMcb4rLC0iowDx7loQnDdjTToFRHKXVeksSu/lCc+3O/quTYd9GT8ndzNM/56s0BijPHZOzvyUYWL0oKnf6SxuWkDuGBsIve9tZu8kkrXzpOZW0xqYvfP+OvNAokxxmcrtueTktCL8YOCNx2IiPDz+ROoqVd+vXyHK+doyPjbExI1enM1kIjIPBHZJSJZInJbE+9HishzzvvrRGS4s32uiGwQkS3Oz9le+1znbN8sIm+ISH83P4MxpmVlVbWszjrKRWkDg36U0rB+MXzzvFG8knmYNVlH/X78nKKTFJVX96j+EXAxkIhIKPAAcAmQBlwnImmNii0CjqtqKnAfcI+z/ShwhapOAm4EnnKOGQb8GbhAVScDm4Elbn0GY0zr3ttVSHVtfdCN1mrON88fxdCEaO58eSvVtf7teN+YcxzoWR3t4O4dyUwgS1WzVbUaeBZY0KjMAuAJ5/mLwBwREVXdqKqHne3bgCgRiQTEecSI50+fPsBhjDEBs2J7HgkxEaQPTwh0VdokKjyUn89PY29hOY+s3ufXY2fmlBAVHsK4gcHbxOcGNwPJECDH63Wus63JMqpaC5QAjVOEXgVsVNUqVa0BvglswRNA0oBH/F91Y0xbVNfWs3JnAReOTyI0JLibtbzNHjeAuWkD+Ms7ezhU7L9cXJtyjveYjL/e3Py0Tf1WNc5R0GIZEZmAp7lrsfM6HE8gmQYMxtO0dXuTJxe5RUQyRCSjsLCw/bU3pg1UtVNTbwSbj7KPUVpZG9SjtZpz5+VpKMrdr273y/Gqa+vZevgEU5J7VrMWuJtGPhdI8XqdzOeboRrK5Dr9H3FAEYCIJAMvATeo6l6n/FSAhtci8jzwuU58p8xSYClAenp6z/2fblx1w6Mf81H2MWIjw4iNCiM2Mpzep56HER0RSk2dUlVbR3VtPdV19VTVOD9r6xiVGMv/uyyNxN6Rgf4oHbJiex69wkM5e3TXG/OSkhDNkgtS+eOK3by3u5Dzxvi2auPOPCfjbw9J1OjNzTuS9cBoERkhIhHAQmBZozLL8HSmA1wNrFRVFZF44DXgdlVd41X+EJAmIg3/4nMBd8bxGdOK+nplXXYRk4bEccWUwcwYlkBy316EhgiFpVVsOVTCyp0FfJR9jO2HT3jW8C6vpq5e6RUeSmJsJG9szePi/3ufN7flBfrjtFt9vfLW9nzOG5NIVHhooKvTId84dyQj+sdw18tbqazxLUNwT1patzHX7khUtVZElgBvAqHAo6q6TUR+CWSo6jI8/RtPiUgWnjuRhc7uS4BU4A4RucPZdpGqHhaRXwDvi0gNcAC4ya3PYExL8ksrqa6r58rTkrn+jI4tYLQnv5TvPbeJxU9t4Nr0ZO68YgKxkV1jvbnNh0rIP1HFxRO7xmitpkSGhfKL+RO44dGPeXDVXr4/d0yHj7Upp5j+sZEMie/lxxp2Da7+xqrqcmB5o213ej2vBK5pYr+7gbubOeZDwEP+rakx7XfwWAUAQ33ILTV6QG9e+tZZ/OWdPfxtVRZrs4/xp2unMqMLjIB6c1seoSHC7LFdN5AAnDsmkQVTB/OXlXsYmRjDgqkdW57Xk/E3Lujn0rihZw0tMMaPDhb5HkgAIsJC+NHFY3nh1lkIwrV/X8s9b+z0+xwHf1uxLY8zRiYQF931U4Hcc9VkZg5P4IfPZ/LuzoJ2719SUUN2Yc/K+OvNAokxHZRTVEGIwGA/NWVMH5bA6989h4UzUnhw1V4WPLCGZ9YdYP3+IkoqavxyDn/JKihjb2F5lxyt1ZSo8FAevjGdcYN6c+vTG1i/v6hd+2eeyvjbM5bWbaxrNMYaE4QOFlUwKK4XEWH++3ssJjKM3145mTnjBvCz/27hZy9tPfXegD6RjBnQ23nEMmN4AiMTY/127vZ4a7tn7ZG5XWQ2e1v0jgrn8a/N5NqH1nLz4+t57pZZpA3u06Z9M3OcjL8pPSfjrzcLJMZ00MGiCp+btZpzYdoA5oxP4lDxSfbkl7E7v5Rd+aXsyS/jmXUHqKypJyI0hBXfP5fh/WNcqUNL3tyWx+TkOL/djQWL/rGRPPX107n6wQ+54dGPefHWWW26vptyihmVGEufHpTx15s1bRnTQQeLTroWSMCTrTa5bzQXjEti8Xmj+NO1U3nl22ez7RfzWP6dc1CUx11eW6Mp+Scq2ZRT3GVya7XXkPhePLXodOpVuf6Rda2mnD+V8bcHTkRsYIHEmA6oqK7laFkVQ/t1/mqAoSFC2uA+zJ8yhOczcig52bn9Jw3NWhcH0drs/paaFMvjX5vB8fJqvvrIOo6XVzdbNvf4SY6VV/fIiYgNrGnLmA7IKfLkZwrksrKLzh7Bvz/J5dmPD7L4vFGddt43t+Uxon8MqUmB6Z/pLJOT4/nHjenc9Nh6bnp8PTed+elcIXGyO4nA1kMlAEzroSO2wAKJMR3ir6G/vkgb3IdZI/vxxIf7WXT2iE5JFFhaWcNH2cf42lkjesR8iTNH9eev103jW898wvefy2y2XHx0OGN7WMZfbxZIjOmAHCeQpPQNbGfzorNH8PUnM3h9ax5XTBns+vk+3HuMmjrlgrFJrp8rWFw0YSDrf3YhJyprUP00q6yqnnqeEB1BeA/L+OvNAokxHXCwqIKYiFASYiICWo/Z45IY0T+Gh1fv4/LJg1y/S1i1q5DYyDDSh/es+RJ9YyLoG+B/62DWc0OoMT7IKaogJSE64M07ISHCzWcNJzOnmE8OHnf1XKrKe7sKOCu1X4/+69t8nv02GNMBbs4haa+rpicT1yvc76v9NbanoIzDJZWc34OatUzbWCAxpp1UNagCSXREGNfNHMobW/NO9d24YdUuTw4qX9ftMN2PBRJj2qmwtIqq2vqAzCFpzo1nDiNExNUJiqt2FTJmQGy3m81ufGeBxJh2ahj6G8g5JI0NiuvFZZMH8dz6HEor/T9BsbyqlvX7i6xZyzTJAokx7RQMc0iasujsEZRV1fJ8Rq7fj90w7Pd8a9YyTbBAYkw7HSyqQISgWwlvcnI8M4b35bE1+6ir19Z3aIdVuwqIiQglvQssuGU6nwUSY9opp+gkA/tEBeU65YvOHkHu8ZOs8OMa8KrKql2FnJna368p8033Yb8VxrRTTlEFKX2Dq1mrwdy0gaQk9PLrUOC9hWUcKj5po7VMsyyQGNNOB53JiMEoNET42pkjyDhwnMycYr8cc9WuQgDOH2uBxDTNAokx7VBZU0feicqg62j3du2MFHpHhrH4qQ38dvkOtuSWoNrxPpP3dheSmhRLcpDehZnAs0BiTDvkHvekjx/aL7g62r3FRobx9xumM25Qbx5ZvY8r/rqa8/+4ij+8uZPth0+0K6hUVNeyLrvIRmuZFlnSRmPaISdIh/42duao/pw5qj/Hy6tZsT2PVzcf4aH3snng3b2MTIxhwZQhLD5vZKsDBtbuPUZ1Xb3NHzEtskBiTDsE42TElvSNieBLM4bypRlDOVZWxetb83h182Hue3s3hWWV3P2FSS3uv2pXIb3CQ5kxomdl+zXtY01bxm8eX7OPP7y506f2+GB3sKiCqPAQEmMjA12VdusXG8n1Zwzj2Vtm8Y1zRvD0Rwd518mf1RRVZdXuAs4c1Y/IsOAb6myChwUS4zdPfnSAB97dy8+Xbeu2waQhWWOg08f76ocXjWXsgN78+MXNza5Hvu9oOTlFJ220lmmVq4FEROaJyC4RyRKR25p4P1JEnnPeXyciw53tc0Vkg4hscX7O9tonQkSWishuEdkpIle5+RlM21TV1nHgWAUD+0TxxNoD3P3ajm4ZTHKCKOuvL6LCQ7nvS1Mprqjmpy9tafLf6tNhv9Y/YlrmWiARkVDgAeASIA24TkTSGhVbBBxX1VTgPuAeZ/tR4ApVnQTcCDzltc/PgAJVHeMc9z23PoNpu31Hy6mrV26/dBw3nTmcR1bv4543dnWrYKKqpxa06g7SBvfhhxeN5fWteby08dDn3l+1u5CRiTHd5vMa97jZ2T4TyFLVbAAReRZYAGz3KrMA+Lnz/EXgryIiqrrRq8w2IEpEIlW1CrgZGAegqvV4go4JsD35ZQCMTurN/CmDqamr56H39hIRKvzgorEBrp1/FJVXU15dF7Sz2jviG+eMZOWOAu56eRszRyScmitysrqOj7KP8ZXThwa4hqYrcLNpawiQ4/U619nWZBlVrQVKgH6NylwFbFTVKhGJd7b9SkQ+EZEXRGSA/6tu2mtPfikhAiMTYxARfrVgIl9KT+EvK7P4yzt7Al09vwjWrL++CA0R7r12Cgr88PlM6p1kjx9lH6O61ob9mrZxM5A01RvZuJ2jxTIiMgFPc9diZ1MYkAysUdXTgLXAH5s8ucgtIpIhIhmFhYXtrbtppz0FZQzrF3NqXkJIiPDbKydx5WlD+NNbu3lw1d4A19B3pwJJEC1o5Q8pCdHcdUUa6/YVncrRtWpXAVHhIZw+wrL9mta5GUhygRSv18nA4ebKiEgYEAcUOa+TgZeAG1S14VvoGFDhbAd4ATitqZOr6lJVTVfV9MREG3Xitt35pYxOiv3MtpAQ4Q9XT2H+lMHc88ZOHv4gO0C184+GyYjdqWmrwdXTk7l4wgD+8OYuduad4L3dhcwa2S8oMxyb4ONmIFkPjBaRESISASwEljUqswxPZzrA1cBKVVWnCes14HZVXdNQWD09t68A5zub5vDZPhcTANW19ew/VsHoAbGfey80RPjTtVO4ZOJA7n5tR4vzFoLdwaIKEntH0iui+325igi/+eIk+vQK5xtPZrD/WIU1a5k2cy2QOH0eS4A3gR3A86q6TUR+KSLznWKPAP1EJAv4AdAwRHgJkArcISKbnEfDb/VPgJ+LyGbgq8AP3foMpm0aRmyNGdC7yffDQkP488Jp9I+N5Pn1OU2W6QoOdpOhv83pFxvJ76+eRE6RJ5+YpY03beVqihRVXQ4sb7TtTq/nlcA1Tex3N3B3M8c8AJzr35oaX+zOLwUgNenzdyQNIsJCuHzyIP758UFOVNbQJyq8s6rnNzlFJ5nZzfsMZo8bwKKzR7Alt4Th/WMCXR3TRbTpjkRERolIpPP8fBH5jtcIKtPD7SkoI0RgVGLzgQTgiimDqa6tZ8W2/E6qmf9U19ZzuORkj5hTccflaTx/66xAV8N0IW1t2vo3UCciqXiao0YA/3StVqZL2ZNfytCE6FY7Zk8bGk9y3168vOnzk9+C3eHik6h2r6G/xvhLWwNJvdPn8UXg/1T1+8Ag96plupI9BWWMbqZ/xJuIMH/KYD7ce4yjZVWdUDP/6Y5zSIzxl7YGkhoRuQ7PCKtXnW1dr5Hb+F11bT37j5Z/buhvcxZMHUJdvbJ8yxGXa+Zfn6aPD94FrYwJlLYGkq8Bs4Bfq+o+ERkBPO1etUxXsf9YObUtjNhqbOzA3owd0JuXNzWeUhTccooqiAgNYUDvqEBXxZig06ZAoqrbVfU7qvovEekL9FbV37lcN9MFNIzYamoOSXPmTx3MhgPHT03w6woOFlWQnNCLkJCunT7eGDe0ddTWKhHpIyIJQCbwmIj8yd2qma5gT37bRmx5mz9lMACvbO46dyXdfQ6JMb5oa9NWnKqeAK4EHlPV6cCF7lXLdBV7Cto2YstbSkI004bGs6yLNG+pKgePWSAxpjltDSRhIjIIuJZPO9uNYXd+GalJbesf8bZgymB25pWeahoLZiUnayitqrVAYkwz2hpIfokn1cleVV0vIiOB7pEb3HRYw4itMe3oH2lw2eTBhAhd4q7k0xFbFkiMaUpbO9tfUNXJqvpN53W2qtoStz1cw4it9nS0N0jsHclZqf1Zlnk46FdRtDkkxrSsrZ3tySLykogUiEi+iPzbSfNuejDvVRE74oopgzlYVMGmnGJ/VsvvGpIY2h2JMU1ra9PWY3hSvg/Gs6rhK842E2TKq2q58+WtbD98wvVz7c4vRdo5YsvbxRMGEhEawrLM4G7eOlhUQb+YCGIjXc1xakyX1dZAkqiqj6lqrfN4HLAc00Fo+ZYjPLn2AAuXrmXjweOuniuroIyhCdEdXp8jrlc4F4xL5NXNR6irD97mrZyiCpLtbsSYZrU1kBwVketFJNR5XI9ntUITZF7fmseAPpHER0dw/cPrWJft3j+TZ1XEjjVrNZg/ZQiFpVV85GI9fWVzSIxpWVsDyc14hv7mAUfwrGb4NbcqZTrmRGUNH+wp5IrJg3l+8SwGxkVx42Mf88Ee/69ZX1NXz76j5R3qaPc2Z3wSMRGhQTt6q7aunkPFJxlqObaMaVZbR20dVNX5qpqoqkmq+gU8kxNNEHl7ez41dcqlkwcxMC6K5xbPYkT/WBY9nsHb2/27Bsj+ow05tnwLJFHhoVw8YSDLtx6hqrbOT7XznyMlldTVq92RGNMCX5ba/YHfamH8YvmWPAbFRTE12bPmWP/YSP71jdMZP6g3tz69gVf9mJJkt48jtrzNnzqY0spa3tvl/zsnX9kcEmNa50sgsex1QaS0sob39xQyb+LAzyQWjI+O4Omvn860ofF8518beXFDrl/Ot6fAtxFb3s5K7U9CTERQjt6yOSTGtM6X8YzBO8ymB1q5s4Dq2nounfT59cZ6R4XzxM0zueXJDfzohUwOHCsnbVAf4qLD6RsdQd/oCOKjw9uVL2tPfhkpfTs+YstbeGgIl00axPMZORwtq6J/bKTPx/SXPfllRIaFMCjO+kiMaU6LgURESmk6YAhg/7OCyPItR0jqHcn0oX2bfD86IoyHb0xnyT83cv/KrCbLRIWH0C8mkp9eOp7LJre8AOaeglKf+0e83XjmcJ5ed4BHV+/jx/PG+e24vsrMLWbSkDhCLX28Mc1qMZCoqu8N4MZ15VW1rNpVyMIZKS2ulxEVHso/bphO3olKiitqOF5RTXFFjdfzat7dVchdy7Yxe1xSs3cbDSO2Zo8b4LfPkJoUy6WTBvHk2gMsPncUcdGBX4Czpq6erYdK+OoZwwJdFWOCmk3V7Qbe3VVAVW09lzTRrNWYiDAorlezTTUXTyji6ofW8uTa/Sw+b1STZQ4cK6emzvcRW40tuSCV1zYf4fEP9/PdC0f79dgdsSuvlKraeqakxAe6KsYENV86202QWL7lCP1jI5kxPMHnY6UPT+DcMYk89N5eyqpqmyzTMGKrrcvrttX4QX24cPwAHl2zr9lzd6aGHGBTLZAY0yILJF1cRXUt7+4sZN7EAX5rx//B3DEcr6jh8TX7mnx/T36Z30ZsNbZkdiolJ2t4+qMDfj92e2XmFJMQE0FyX+sONKYlFki6uPd2FXKypo5LJ7berNVWU1PiuXDqQzceAAAXo0lEQVR8Ekvfz6bkZM3n3t9dUOq3EVtNnfuc0f15+INsTlYHdoLippxipqbEI2Id7ca0xNVAIiLzRGSXiGSJyG1NvB8pIs85768TkeHO9rkiskFEtjg/Zzex7zIR2epm/buC17YcoV9MBDNH+N6s5e37c8dworKWR1Z//q5kT34po5P8fzfS4NuzR3O0rJpn1x907RytKa2sIauwjCnJ1qxlTGtcCyQiEgo8AFwCpAHXiUhao2KLgOOqmgrcB9zjbD8KXKGqk4AbgacaHftKoMytuncVlTV1rNxZwEUTBhIW6t9/ygmD47hk4kAeXb2P4+XVp7Z/mmPLvQF9M0ckMHNEAkvfzw5Y2pQth0pQhSkpcQE5vzFdiZt3JDOBLGc1xWrgWWBBozILgCec5y8Cc0REVHWjqjZMc94GRIlIJICIxOJJz3K3i3XvEt7bXUhFdR2XThroyvG/P3cM5dW1LP0g+9S2hhFbbt6RAHx7dipHSir5zyeHXD1PczJzSgDsjsSYNnAzkAwBcrxe5zrbmiyjqrVACdCvUZmrgI2qWuW8/hVwL1Dh7wp3Na9vOULf6HDOGNn4kvnHmAG9uWLyYB5fs5+jZZ7Lv8elEVuNnZ3anykp8fxtVRa1dfWunqspm3KOM7xfNH1jIjr93MZ0NW4GkqZ6KBvPkm+xjIhMwNPctdh5PRVIVdWXWj25yC0ikiEiGYWFwZcM0FeVNXW8vaOAi9IGEu7nZi1v371wNFW1dTy0ai/w6dDfUUkxrp0TPPNdvn1BKjlFJ1vNwaWqFHk1v/lDZk6JzR8xpo3cDCS5QIrX62Sg8TfCqTIiEgbEAUXO62TgJeAGVd3rlJ8FTBeR/cBqYIyIrGrq5Kq6VFXTVTU9MbH7Lea4es9RyqpqucSlZq0GoxJj+eK0ZJ766AD5JyrZU1BKSkIvoiPcn8s6Z3wS4wf14YF3s5pcQbG+XnlzWx4LHlhD+t1vseGAf1aEzCupJO9EpTVrGdNGbgaS9cBoERkhIhHAQjzrvntbhqczHTyLZa1UVRWReOA14HZVXdNQWFUfVNXBqjocOBvYrarnu/gZgtbyrUeI6xXOWan9XT/Xd+eMpq5e+du7WezJL/NL6vi2EBGWXJDK3sJy3tiad2p7Xb3y8qZDXPLnD1j81AaKK2qIjQzjoff2tnC0tsvM9UxEtDsSY9rGtUDi9HksAd4EdgDPq+o2EfmliMx3ij0C9BORLDwd6A1DhJcAqcAdIrLJeSS5Vdeuprq2nre25zM3bYCrzVoNhvaL5pr0ZP71cQ7ZR8t8XhWxPeZNHMioxBjuX7mHqto6nlt/kDn3ruK7z26iXpX/+9JUVv7wPG46awRvbc8nq8D3wXybcooJCxEmDO7jh09gTPfnavuEqi4HljfadqfX80rgmib2u5tWRmWp6n5gol8q2sWsyTpKaWWta6O1mrJk9mj+veGQJ8dWJ92RAISGCP9zQSo/eD6TWb9dSVF5NROH9OGh60/jorRP1165cdYw/v7eXv7xfjb3XD3Zp3Nm5hQzflCfdqXVN6Yns5ntXdDyLUfoHRXWKc1aDYbE92LhTE+Xl9sjthqbP2UwU5LjGJUYw+Nfm8ErS85m3sRBn8l03C82kmvTU3hp4yHyT1R2+Fz19crm3BKbP2JMO1j23y6muraeN7flMXf8ACLDOvcv5h9dPJa0QX2YOKRzm3zCQkN4ecnZrZb7+jkjeGbdAR5bs5/bLunYmibZR8soq6q1jnZj2sHuSLqY1VmFnKis5fIp/sut1VZ9osJZOHNo0OaeGtYvhksmDeKZjw5QWvn5HGFtsfGgp6N92lALJMa0lQWSLubVzUfoExXG2andb0izPyw+dySlVbX86+OO5enKzC0mNjKMkf07b0CBMV2dBZIupLKmjre25XPxhIFEhNk/XVMmJ8dz5qh+PLp6P9W17Z8Rn5lTwuTkuBZXmjTGfJZ9G3Uh7+8upLSqlsunDA50VYLa4vNGkXeistUZ8Y1V1tSx48gJmz9iTDtZIOlCXnNya505yp3cWt3FuaP7M25gb5a+v5f6JmbEN2fb4RPU1qutiGhMO1kg6SIqa+p4e3s+8ya6m1urOxARbj1vFLvzy1i1u6DN+2Xa0rrGdIh9I3UR7+4soLy6jssmWbNWW1w2eRBD4nvx0HvZrRd2ZOYWM7BPFAP6RLlYM2O6HwskXcSrzkqIZ4z070qI3VV4aAiLzh7Bx/uK+ORg25I5ZuYU20REYzrAAkkXUFFdy8odBVwyyf8rIXZnX5qRQlyvcJa24a7keHk1+49VMDWlbyfUzJjuxb6VuoB3dhRwssaatdorJjKMr54xjDe355Fd2HIyx08z/todiTHtZYGkC3ht8xESe0cyc4Q1a7XXjWcOJzw0hPtXZqHa/AiuzJwSRGDSEAskxrSXBZIgV1ZVy7u7Crh04kBCbZJcuyX2juTms0bw0sZD3LVsW7PDgTNzixmdFEvvqPBOrqExXZ8lbQxyb2/Pp6q23iYh+uAn88ZSr8rS97M5XlHDvddM+UxmAFVlU04xc8bZkjfGdIQFkiD36uYjDOwTxfSh1gncUSLCTy8dT0JMBL97fSclJ2t46PrTTi0XnHv8JEXl1Taj3ZgOsqatIFZysob3dxdy6aRBlvvJD249bxT3XDWJ1XsK+fI/1nG8vBrwrIgINhHRmI6yQBLE3tqeT3VdfUBSxndXX5oxlL99ZTrbj5zg2r+v5UjJSTJziokMC2HswM5dsMuY7sICSRB7bfNhhsT3Ypr9pexX8yYO5PGvzeBISSVXP7iWd3cVMHFInKWeMaaD7H9OkCquqOaDPUe5bPKgoF1Iqis7c1R/nr3lDCpr6thbWG4rIhrjAwskQWrFtnxq65XLJ1uzllsmDonjhVtnce6YRL44bUigq2NMl2WjtoLUK5sPMzQh2ibIuWxkYixP3jwz0NUwpkuzO5IgdKysig/3HrNmLWNMl2CBJAg9snofddasZYzpIiyQBJl3duTzt1V7uXp6MhMGW7OWMSb4WSAJIvuPlvO95zYxcUgf7v7CxEBXxxhj2sTVQCIi80Rkl4hkichtTbwfKSLPOe+vE5Hhzva5IrJBRLY4P2c726NF5DUR2Ski20Tkd27W3xeVNXUUOTOn2+JkdR23Pr2B0BDhwa9MJyo81MXaGWOM/7gWSEQkFHgAuARIA64TkbRGxRYBx1U1FbgPuMfZfhS4QlUnATcCT3nt80dVHQdMA84SkUvc+gy++Ol/tnDm797hhYycVsuqKrf/ZzO78kv588JppCREd0INjTHGP9y8I5kJZKlqtqpWA88CCxqVWQA84Tx/EZgjIqKqG1X1sLN9GxAlIpGqWqGq7wI4x/wESHbxM3RIcUU1r24+QnhICP/74mZ+/GImlTV1zZZ/4sP9/HfTYX5w4RjOG5PYiTU1xhjfuRlIhgDef47nOtuaLKOqtUAJ0K9RmauAjapa5b1RROKBK4B3mjq5iNwiIhkiklFYWNjhD9ERyzIPU11XzzPfOJ1vz07l+YxcvvDAGvYdLf9c2Yz9Rdz92g4uHJ/E/1yQ2qn1NMYYf3AzkDQ1AaLxqkItlhGRCXiauxZ/ZieRMOBfwF9UtckFuVV1qaqmq2p6YmLn/pX/QkYu4wf1YXJyPD+8aCyPfW0GeScqueL+1SzfcuRUuYLSSr71zCck9+3FvddOtQy/xpguyc1AkgukeL1OBg43V8YJDnFAkfM6GXgJuEFV9zbabymwR1X/z4V6+2Rn3gm2HCrhmumftrhdMDaJ175zDqlJsXzrmU/4+bJtVFTXsuSZjZRW1vLQV6cT18tW5jPGdE1upkhZD4wWkRHAIWAh8OVGZZbh6UxfC1wNrFRVdZqtXgNuV9U13juIyN14As7XXax7h72QkUt4qPCFRrmbhsT34vnFs/jt6zt4bM1+lmUepqi8mj8vnMq4gX0CVFtjjPGda3ckTp/HEuBNYAfwvKpuE5Ffish8p9gjQD8RyQJ+ADQMEV4CpAJ3iMgm55Hk3KX8DM8osE+c7UETUGrq6vnvxkPMGTeAhJiIz70fERbCXVdM4G9fOY2aunq+cc4IFky1ZIHGmK5NVBt3W3Q/6enpmpGR4fp53tyWx+KnNvDoTenMHjegxbK1dfWE2foXxpggJiIbVDW9tXL2TeZHL2TkktQ7knNHt965b0HEGNNd2LeZnxSWVvHurgK+eNoQCxLGmB7FvvH85L8bD1FXr1wzPaX1wsYY041YIPEDVeWFDTlMGxpPalJsoKtjjDGdygKJH2zOLWF3fpndjRhjeiQLJH7wwoYcosJDuHyKLURljOl5LJD4qLKmjmWbDjNvwkD6RNnsdGNMz2OBxEcrtudzorKWa9KtWcsY0zNZIPHRCxk5DInvxayRjZMWG2NMz2CBxAeHi0+yOusoV01Ptsy9xpgeywKJD/69IRdVPpPp1xhjehoLJB2kqrz4SS5njEywpXGNMT2aBZIO+uRgMQeOVdjcEWNMj2eBpIPe2p5PWIgwd0LLWX6NMaa7s0DSQe/syGfmiASbO2KM6fEskHTAwWMV7CkoY854uxsxxhgLJB3w9o58AC4cnxTgmhhjTOBZIOmAd3bmk5oUy7B+MYGuijHGBJwFknYqraxhXXYRc8bZ3YgxxoAFknZ7f/dRauvV+keMMcZhgaSd3tmRT3x0OKcNjQ90VYwxJihYIGmHunrl3V0FXDA2ydZlN8YYh30btsMnB49zvKKGOTZayxhjTrFA0g5v7/DMZj93TGKgq2KMMUHDAkk7vLOjgNNH2mx2Y4zxZoGkjQ4cKyeroIw542y0ljHGeHM1kIjIPBHZJSJZInJbE+9HishzzvvrRGS4s32uiGwQkS3Oz9le+0x3tmeJyF9EpFNWlHp7RwEAF9qwX2OM+QzXAomIhAIPAJcAacB1IpLWqNgi4LiqpgL3Afc4248CV6jqJOBG4CmvfR4EbgFGO495bn0Gb+/syGd0UixD+9naI8YY483NO5KZQJaqZqtqNfAssKBRmQXAE87zF4E5IiKqulFVDzvbtwFRzt3LIKCPqq5VVQWeBL7g4mcA4ERlDR/vK7JJiMYY0wQ3A8kQIMfrda6zrckyqloLlAD9GpW5CtioqlVO+dxWjgmAiNwiIhkiklFYWNjhDwHw3q5CauvVkjQaY0wT3AwkTfVdaHvKiMgEPM1di9txTM9G1aWqmq6q6YmJvg3XfWdHPgkxEUwb2ten4xhjTHfkZiDJBbzXoU0GDjdXRkTCgDigyHmdDLwE3KCqe73KJ7dyTL+qravn3V2FnD82kdCQTunXN8aYLsXNQLIeGC0iI0QkAlgILGtUZhmeznSAq4GVqqoiEg+8BtyuqmsaCqvqEaBURM5wRmvdALzs4mdgw4HjlJyssWG/xhjTDNcCidPnsQR4E9gBPK+q20TklyIy3yn2CNBPRLKAHwANQ4SXAKnAHSKyyXk0dFB8E3gYyAL2Aq+79RkA3tlZQHiocO6Y/m6exhhjuizxDH7q3tLT0zUjI6ND+86+dxWD43rx9NdP93OtjDEmuInIBlVNb62czWxvwb6j5WQXlluSRmOMaYEFkha8c2ptdusfMcaY5lggacE7OwoYMyCWlASbzW6MMc0JC3QFgpWqkja4D4PiogJdFWOMCWoWSJohItxxeePUYMYYYxqzpi1jjDE+sUBijDHGJxZIjDHG+MQCiTHGGJ9YIDHGGOMTCyTGGGN8YoHEGGOMTyyQGGOM8UmPyP4rIoXAgUab4/As7duc5t5vz/bG2/oDR1usrP+19jn9vX9bytu1d2f/tpZvqZxd+44dw81r39x7nXHth6lq60vMqmqPfABLO/J+e7Y33gZkBNvn9Pf+bSlv1z5w1761cnbtO3YMN699W69zIK99T27aeqWD77dne2vn6Ay+1qG9+7elvF17d/Zva/mWytm179gx3Lz2zb0XNNe+RzRtBQsRydA2LBJj/M+ufeDYtQ+czrr2PfmOJBCWBroCPZhd+8Cxax84nXLt7Y7EGGOMT+yOxBhjjE8skHSQiDwqIgUisrUD+04XkS0ikiUifxER8Xrv2yKyS0S2icjv/Vvr7sGNay8iPxeRQyKyyXlc6v+ad31u/d477/9IRFRE+vuvxt2HS7/3vxKRzc7v/AoRGdyRulkg6bjHgXkd3PdB4BZgtPOYByAiFwALgMmqOgH4o+/V7JYex8/X3nGfqk51Hst9q2K39TguXHsRSQHmAgd9rF939jj+v/Z/UNXJqjoVeBW4syMHt0DSQar6PlDkvU1ERonIGyKyQUQ+EJFxjfcTkUFAH1Vdq54OqieBLzhvfxP4napWOecocPdTdE0uXXvTBi5e+/uAHwPWadsMN669qp7wKhpDB6+/BRL/Wgp8W1WnAz8C/tZEmSFArtfrXGcbwBjgHBFZJyLvicgMV2vbvfh67QGWOLf5j4pIX/eq2u34dO1FZD5wSFUz3a5oN+Tz772I/FpEcoCv0ME7Eluz3U9EJBY4E3jBq+k3sqmiTWxr+CsgDOgLnAHMAJ4XkZFqQ+ta5Kdr/yDwK+f1r4B7gZv9W9Pux9drLyLRwM+Ai9ypYfflp997VPVnwM9E5HZgCXBXe+tigcR/QoBip63xFBEJBTY4L5fh+cJK9iqSDBx2nucC/3ECx8ciUo8nV06hmxXvBny+9qqa77XfP/C0F5vW+XrtRwEjgEznyzAZ+EREZqpqnst17+r88Z3j7Z/Aa3QgkFjTlp84bY37ROQaAPGYoqp1Xh24d6rqEaBURM5wRk7cALzsHOa/wGxn/zFABJ2f7K7L8ce1d9qRG3wRaPfImJ7I12uvqltUNUlVh6vqcDx/TJ1mQaR1fvq9H+11yPnAzo5Wxh4deAD/Ao4ANXh++Rfh+cvqDSAT2A7c2cy+6Xi+qPYCf+XTiaERwNPOe58AswP9OYPx4dK1fwrYAmzG81fcoEB/zmB8uHHtG5XZD/QP9OcMxodLv/f/drZvxpOna0hH6mYz240xxvjEmraMMcb4xAKJMcYYn1ggMcYY4xMLJMYYY3xigcQYY4xPLJCYHklEyjr5fA+LSJqfjlXnZGvdKiKviEh8K+XjReRb/ji3MU2x4b+mRxKRMlWN9ePxwlS11l/Ha+Vcp+ouIk8Au1X11y2UHw68qqoTO6N+puexOxJjHCKSKCL/FpH1zuMsZ/tMEflQRDY6P8c6228SkRdE5BVghYicLyKrRORFEdkpIs84M4lxtqc7z8ucRHmZIvKRiAxwto9yXq8XkV+28a5pLZ8mP4wVkXdE5BPxrD2xwCnzO2CUcxfzB6fs/zrn2Swiv/DjZTQ9kAUSYz71ZzxrkswArgIedrbvBM5V1Wl4sqP+xmufWcCNqjrbeT0N+B6QBowEzmriPDHAR6o6BXgf+IbX+f/snL+pXEif4eRUmoNnJj5AJfBFVT0NuAC41wlktwF71ZMy439F5CI8a1LMBKYC00Xk3NbOZ0xzLGmjMZ+6EEjzyqTaR0R6A3HAE05eIgXCvfZ5S1W914j4WFVzAURkEzAcWN3oPNV8mhRyA54FncATlBrW6PgnzS9s1svr2BuAt5ztAvzGCQr1eO5UBjSx/0XOY6PzOhZPYHm/mfMZ0yILJMZ8KgSYpaonvTeKyP3Au6r6Rae/YZXX2+WNjlHl9byOpv+P1einnZPNlWnJSVWdKiJxeALS/wB/wbOeRCIwXVVrRGQ/ENXE/gL8VlX/3s7zGtMka9oy5lMr8KzHAICINKTnjgMOOc9vcvH8H+FpUgNY2FphVS0BvgP8SETC8dSzwAkiFwDDnKKlQG+vXd8EbnbWs0BEhohIkp8+g+mBLJCYnipaRHK9Hj/A86Wc7nRAbwdudcr+HvitiKwBQl2s0/eAH4jIx8AgoKS1HVR1I57MrwuBZ/DUPwPP3clOp8wxYI0zXPgPqroCT9PZWhHZArzIZwONMe1iw3+NCRLiWS3wpKqqiCwErlPVBa3tZ0ygWR+JMcFjOvBXZ6RVMbbUr+ki7I7EGGOMT6yPxBhjjE8skBhjjPGJBRJjjDE+sUBijDHGJxZIjDHG+MQCiTHGGJ/8fwgImRngZ/6cAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:47 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.035484</td>\n",
       "      <td>0.027525</td>\n",
       "      <td>0.990286</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.036864</td>\n",
       "      <td>0.026836</td>\n",
       "      <td>0.991143</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.029393</td>\n",
       "      <td>0.027737</td>\n",
       "      <td>0.990571</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.034428</td>\n",
       "      <td>0.025910</td>\n",
       "      <td>0.990571</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4, max_lr=slice(1e-5,1e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:47 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.035168</td>\n",
       "      <td>0.026299</td>\n",
       "      <td>0.990571</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.033477</td>\n",
       "      <td>0.026339</td>\n",
       "      <td>0.990571</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.030585</td>\n",
       "      <td>0.025375</td>\n",
       "      <td>0.990571</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.028298</td>\n",
       "      <td>0.025906</td>\n",
       "      <td>0.990000</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4, max_lr = slice(1e-6,1e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:47 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.032324</td>\n",
       "      <td>0.025630</td>\n",
       "      <td>0.990857</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.029159</td>\n",
       "      <td>0.027806</td>\n",
       "      <td>0.990286</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.033450</td>\n",
       "      <td>0.026886</td>\n",
       "      <td>0.991143</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.029236</td>\n",
       "      <td>0.026772</td>\n",
       "      <td>0.990000</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4, max_lr = slice(1e-4/5,1e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:47 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.034933</td>\n",
       "      <td>0.030732</td>\n",
       "      <td>0.988571</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.046961</td>\n",
       "      <td>0.029223</td>\n",
       "      <td>0.989714</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.039652</td>\n",
       "      <td>0.027428</td>\n",
       "      <td>0.991429</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.031172</td>\n",
       "      <td>0.025339</td>\n",
       "      <td>0.991143</td>\n",
       "      <td>00:11</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4, max_lr = slice(1e-4,1e-3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.freeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:33 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.029466</td>\n",
       "      <td>0.023790</td>\n",
       "      <td>0.992286</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.028243</td>\n",
       "      <td>0.023435</td>\n",
       "      <td>0.992571</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.029157</td>\n",
       "      <td>0.024193</td>\n",
       "      <td>0.991714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.026156</td>\n",
       "      <td>0.023572</td>\n",
       "      <td>0.991714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4, max_lr= slice(1e-5/5,1e-5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('stage-1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = (src.transform(tfms, size=32)\n",
    "        .databunch().normalize(imagenet_stats))\n",
    "\n",
    "learn.data = data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:35 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.094779</td>\n",
       "      <td>0.059584</td>\n",
       "      <td>0.976571</td>\n",
       "      <td>00:09</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.057245</td>\n",
       "      <td>0.034519</td>\n",
       "      <td>0.989429</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.037514</td>\n",
       "      <td>0.028941</td>\n",
       "      <td>0.989714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.040391</td>\n",
       "      <td>0.029408</td>\n",
       "      <td>0.989429</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 01:41 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.022341</td>\n",
       "      <td>0.016919</td>\n",
       "      <td>0.992857</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.019891</td>\n",
       "      <td>0.016414</td>\n",
       "      <td>0.994571</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.022543</td>\n",
       "      <td>0.016107</td>\n",
       "      <td>0.993714</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.022827</td>\n",
       "      <td>0.014848</td>\n",
       "      <td>0.994571</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.020127</td>\n",
       "      <td>0.014771</td>\n",
       "      <td>0.994286</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.020210</td>\n",
       "      <td>0.013542</td>\n",
       "      <td>0.994286</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.016318</td>\n",
       "      <td>0.014273</td>\n",
       "      <td>0.994286</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.016766</td>\n",
       "      <td>0.015624</td>\n",
       "      <td>0.993429</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(8, max_lr=slice(1e-6,1e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:51 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.025022</td>\n",
       "      <td>0.013003</td>\n",
       "      <td>0.994000</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.016866</td>\n",
       "      <td>0.023115</td>\n",
       "      <td>0.989429</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.014168</td>\n",
       "      <td>0.013252</td>\n",
       "      <td>0.995429</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.013503</td>\n",
       "      <td>0.012800</td>\n",
       "      <td>0.995714</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4, max_lr = slice(1e-4/5,1e-3/5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('stage-2')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:50 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.010580</td>\n",
       "      <td>0.011261</td>\n",
       "      <td>0.995714</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.012290</td>\n",
       "      <td>0.011235</td>\n",
       "      <td>0.996000</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.011360</td>\n",
       "      <td>0.012245</td>\n",
       "      <td>0.995143</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.012793</td>\n",
       "      <td>0.012949</td>\n",
       "      <td>0.994857</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4, max_lr =slice(7e-7,9e-7))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:34 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.014384</td>\n",
       "      <td>0.011381</td>\n",
       "      <td>0.995143</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.014181</td>\n",
       "      <td>0.014596</td>\n",
       "      <td>0.993714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.016994</td>\n",
       "      <td>0.012827</td>\n",
       "      <td>0.994286</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.017030</td>\n",
       "      <td>0.064029</td>\n",
       "      <td>0.994286</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:35 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.015684</td>\n",
       "      <td>0.011011</td>\n",
       "      <td>0.996000</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.019093</td>\n",
       "      <td>0.014502</td>\n",
       "      <td>0.994000</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.010311</td>\n",
       "      <td>0.010329</td>\n",
       "      <td>0.995429</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.009357</td>\n",
       "      <td>0.010825</td>\n",
       "      <td>0.995714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.freeze()\n",
    "learn.fit_one_cycle(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:35 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.012350</td>\n",
       "      <td>0.011588</td>\n",
       "      <td>0.994571</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.013257</td>\n",
       "      <td>0.010873</td>\n",
       "      <td>0.995429</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.008684</td>\n",
       "      <td>0.453531</td>\n",
       "      <td>0.992286</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.007513</td>\n",
       "      <td>0.441817</td>\n",
       "      <td>0.992571</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.freeze()\n",
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:35 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.011534</td>\n",
       "      <td>0.247484</td>\n",
       "      <td>0.992571</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.007787</td>\n",
       "      <td>0.049706</td>\n",
       "      <td>0.993714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.010151</td>\n",
       "      <td>0.010926</td>\n",
       "      <td>0.995714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.008105</td>\n",
       "      <td>1.195883</td>\n",
       "      <td>0.990571</td>\n",
       "      <td>00:09</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4,max_lr = slice(1e-4/5,1e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.unfreeze()\n",
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:50 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.009547</td>\n",
       "      <td>0.507598</td>\n",
       "      <td>0.992571</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.016935</td>\n",
       "      <td>1.526829</td>\n",
       "      <td>0.988286</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.014439</td>\n",
       "      <td>0.096752</td>\n",
       "      <td>0.994000</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.008756</td>\n",
       "      <td>0.435084</td>\n",
       "      <td>0.994286</td>\n",
       "      <td>00:12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(4,max_lr = slice(1e-5,1e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 00:35 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.011524</td>\n",
       "      <td>0.011103</td>\n",
       "      <td>0.995714</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.009552</td>\n",
       "      <td>1.963145</td>\n",
       "      <td>0.991429</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.010317</td>\n",
       "      <td>0.729466</td>\n",
       "      <td>0.993143</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.011787</td>\n",
       "      <td>1.669949</td>\n",
       "      <td>0.991143</td>\n",
       "      <td>00:08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.freeze()\n",
    "learn.fit_one_cycle(4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n"
     ]
    }
   ],
   "source": [
    "learn.freeze()\n",
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds, _ = learn.load('stage-2').get_preds(ds_type=DatasetType.Test)\n",
    "predictions = np.argmax(preds, axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "fnames = [f.name[:-4] for f in learn.data.test_ds.items]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "submission = pd.DataFrame({\n",
    "    \"Id\":fnames,\n",
    "    \"has_cactus\": predictions\n",
    "})\n",
    "submission.to_csv(path/'submission1.csv', index=False)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}