keras-efficientnetb2-for-classifying-cloud.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:28.144237Z",
     "start_time": "2019-11-16T15:18:23.607292Z"
    },
    "_kg_hide-input": true,
    "_kg_hide-output": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import os, glob\n",
    "import random\n",
    "from sklearn.model_selection import train_test_split\n",
    "import cv2\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import multiprocessing\n",
    "from copy import deepcopy\n",
    "from sklearn.metrics import precision_recall_curve, auc\n",
    "import keras\n",
    "import keras.backend as K\n",
    "from keras.optimizers import Adam\n",
    "from keras.callbacks import Callback\n",
    "from keras.applications.densenet import DenseNet201\n",
    "from keras.layers import Dense, Flatten\n",
    "from keras.models import Model, load_model\n",
    "from keras.utils import Sequence\n",
    "from albumentations import Compose, VerticalFlip, HorizontalFlip, Rotate, GridDistortion\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.display import Image\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "from numpy.random import seed\n",
    "seed(10)\n",
    "from tensorflow import set_random_seed\n",
    "set_random_seed(10)\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:28.154239Z",
     "start_time": "2019-11-16T15:18:28.151574Z"
    },
    "_kg_hide-input": true,
    "_kg_hide-output": true
   },
   "outputs": [],
   "source": [
    "test_imgs_folder = '/home/dex/Desktop/ml/cloud data//test_images/'\n",
    "train_imgs_folder = '/home/dex/Desktop/ml/cloud data//train_images/'\n",
    "num_cores = multiprocessing.cpu_count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Generators"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## One-hot encoding classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:29.700482Z",
     "start_time": "2019-11-16T15:18:28.160867Z"
    },
    "_kg_hide-input": true
   },
   "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>Image_Label</th>\n",
       "      <th>EncodedPixels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0011165.jpg_Fish</td>\n",
       "      <td>264918 937 266318 937 267718 937 269118 937 27...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0011165.jpg_Flower</td>\n",
       "      <td>1355565 1002 1356965 1002 1358365 1002 1359765...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0011165.jpg_Gravel</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0011165.jpg_Sugar</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>002be4f.jpg_Fish</td>\n",
       "      <td>233813 878 235213 878 236613 878 238010 881 23...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Image_Label                                      EncodedPixels\n",
       "0    0011165.jpg_Fish  264918 937 266318 937 267718 937 269118 937 27...\n",
       "1  0011165.jpg_Flower  1355565 1002 1356965 1002 1358365 1002 1359765...\n",
       "2  0011165.jpg_Gravel                                                NaN\n",
       "3   0011165.jpg_Sugar                                                NaN\n",
       "4    002be4f.jpg_Fish  233813 878 235213 878 236613 878 238010 881 23..."
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = pd.read_csv('/home/dex/Desktop/ml/cloud data/train.csv')\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:29.891133Z",
     "start_time": "2019-11-16T15:18:29.714729Z"
    },
    "_kg_hide-input": false
   },
   "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>Image</th>\n",
       "      <th>Class</th>\n",
       "      <th>Fish</th>\n",
       "      <th>Flower</th>\n",
       "      <th>Sugar</th>\n",
       "      <th>Gravel</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0011165.jpg</td>\n",
       "      <td>{Flower, Fish}</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>002be4f.jpg</td>\n",
       "      <td>{Flower, Fish, Sugar}</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0031ae9.jpg</td>\n",
       "      <td>{Flower, Fish, Sugar}</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0035239.jpg</td>\n",
       "      <td>{Flower, Gravel}</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>003994e.jpg</td>\n",
       "      <td>{Fish, Gravel, Sugar}</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         Image                  Class  Fish  Flower  Sugar  Gravel\n",
       "0  0011165.jpg         {Flower, Fish}     1       1      0       0\n",
       "1  002be4f.jpg  {Flower, Fish, Sugar}     1       1      1       0\n",
       "2  0031ae9.jpg  {Flower, Fish, Sugar}     1       1      1       0\n",
       "3  0035239.jpg       {Flower, Gravel}     0       1      0       1\n",
       "4  003994e.jpg  {Fish, Gravel, Sugar}     1       0      1       1"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = train_df[~train_df['EncodedPixels'].isnull()]\n",
    "train_df['Image'] = train_df['Image_Label'].map(lambda x: x.split('_')[0])\n",
    "train_df['Class'] = train_df['Image_Label'].map(lambda x: x.split('_')[1])\n",
    "classes = train_df['Class'].unique()\n",
    "train_df = train_df.groupby('Image')['Class'].agg(set).reset_index()\n",
    "for class_name in classes:\n",
    "    train_df[class_name] = train_df['Class'].map(lambda x: 1 if class_name in x else 0)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-12T18:40:12.951297Z",
     "start_time": "2019-11-12T18:40:12.939657Z"
    }
   },
   "outputs": [],
   "source": [
    "# dictionary for fast access to ohe vectors\n",
    "img_2_ohe_vector = {img:vec for img, vec in zip(train_df['Image'], train_df.iloc[:, 2:].values)}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Stratified split into train/val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:31.261937Z",
     "start_time": "2019-11-16T15:18:31.246149Z"
    }
   },
   "outputs": [],
   "source": [
    "train_imgs, val_imgs = train_test_split(train_df['Image'].values, \n",
    "                                        test_size=0.2, \n",
    "                                        stratify=train_df['Class'].map(lambda x: str(sorted(list(x)))), # sorting present classes in lexicographical order, just to be sure\n",
    "                                        random_state=2019)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Generator class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:33.380430Z",
     "start_time": "2019-11-16T15:18:33.361057Z"
    }
   },
   "outputs": [],
   "source": [
    "class DataGenenerator(Sequence):\n",
    "    def __init__(self, images_list=None, folder_imgs=train_imgs_folder, \n",
    "                 batch_size=32, shuffle=True, augmentation=None,\n",
    "                 resized_height=260, resized_width=260, num_channels=3):\n",
    "        self.batch_size = batch_size\n",
    "        self.shuffle = shuffle\n",
    "        self.augmentation = augmentation\n",
    "        if images_list is None:\n",
    "            self.images_list = os.listdir(folder_imgs)\n",
    "        else:\n",
    "            self.images_list = deepcopy(images_list)\n",
    "        self.folder_imgs = folder_imgs\n",
    "        self.len = len(self.images_list) // self.batch_size\n",
    "        self.resized_height = resized_height\n",
    "        self.resized_width = resized_width\n",
    "        self.num_channels = num_channels\n",
    "        self.num_classes = 4\n",
    "        self.is_test = not 'train' in folder_imgs\n",
    "        if not shuffle and not self.is_test:\n",
    "            self.labels = [img_2_ohe_vector[img] for img in self.images_list[:self.len*self.batch_size]]\n",
    "\n",
    "    def __len__(self):\n",
    "        return self.len\n",
    "    \n",
    "    def on_epoch_start(self):\n",
    "        if self.shuffle:\n",
    "            random.shuffle(self.images_list)\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        current_batch = self.images_list[idx * self.batch_size: (idx + 1) * self.batch_size]\n",
    "        X = np.empty((self.batch_size, self.resized_height, self.resized_width, self.num_channels))\n",
    "        y = np.empty((self.batch_size, self.num_classes))\n",
    "\n",
    "        for i, image_name in enumerate(current_batch):\n",
    "            path = os.path.join(self.folder_imgs, image_name)\n",
    "            img = cv2.resize(cv2.imread(path), (self.resized_height, self.resized_width)).astype(np.float32)\n",
    "            if not self.augmentation is None:\n",
    "                augmented = self.augmentation(image=img)\n",
    "                img = augmented['image']\n",
    "            X[i, :, :, :] = img/255.0\n",
    "            if not self.is_test:\n",
    "                y[i, :] = img_2_ohe_vector[image_name]\n",
    "        return X, y\n",
    "\n",
    "    def get_labels(self):\n",
    "        if self.shuffle:\n",
    "            images_current = self.images_list[:self.len*self.batch_size]\n",
    "            labels = [img_2_ohe_vector[img] for img in images_current]\n",
    "        else:\n",
    "            labels = self.labels\n",
    "        return np.array(labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:35.810955Z",
     "start_time": "2019-11-16T15:18:35.806775Z"
    }
   },
   "outputs": [],
   "source": [
    "albumentations_train = Compose([\n",
    "    VerticalFlip(), HorizontalFlip(), Rotate(limit=20), GridDistortion()\n",
    "], p=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generator instances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:37.175360Z",
     "start_time": "2019-11-16T15:18:36.934649Z"
    }
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'img_2_ohe_vector' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-8-aa6369792a56>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0mdata_generator_train\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDataGenenerator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_imgs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0maugmentation\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0malbumentations_train\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mdata_generator_train_eval\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDataGenenerator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_imgs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0mdata_generator_val\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDataGenenerator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mval_imgs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-6-d31628ee795c>\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, images_list, folder_imgs, batch_size, shuffle, augmentation, resized_height, resized_width, num_channels)\u001b[0m\n\u001b[1;32m     18\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_test\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;34m'train'\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mfolder_imgs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mshuffle\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_test\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 20\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlabels\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mimg_2_ohe_vector\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mimg\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mimg\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimages_list\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     21\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     22\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__len__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-6-d31628ee795c>\u001b[0m in \u001b[0;36m<listcomp>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m     18\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_test\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;34m'train'\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mfolder_imgs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mshuffle\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_test\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 20\u001b[0;31m             \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlabels\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mimg_2_ohe_vector\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mimg\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mimg\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mimages_list\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     21\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     22\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__len__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'img_2_ohe_vector' is not defined"
     ]
    }
   ],
   "source": [
    "data_generator_train = DataGenenerator(train_imgs, augmentation=albumentations_train)\n",
    "data_generator_train_eval = DataGenenerator(train_imgs, shuffle=False)\n",
    "data_generator_val = DataGenenerator(val_imgs, shuffle=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PR-AUC-based Callback"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The callback would be used:\n",
    "1. to estimate AUC under precision recall curve for each class,\n",
    "2. to early stop after 5 epochs of no improvement in mean PR AUC,\n",
    "3. save a model with the best PR AUC in validation,\n",
    "4. to reduce learning rate on PR AUC plateau."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-12T18:40:19.769985Z",
     "start_time": "2019-11-12T18:40:19.750104Z"
    }
   },
   "outputs": [],
   "source": [
    "class PrAucCallback(Callback):\n",
    "    def __init__(self, data_generator, num_workers=num_cores, \n",
    "                 early_stopping_patience=5, \n",
    "                 plateau_patience=3, reduction_rate=0.5,\n",
    "                 stage='train', checkpoints_path='checkpoints/'):\n",
    "        super(Callback, self).__init__()\n",
    "        self.data_generator = data_generator\n",
    "        self.num_workers = num_workers\n",
    "        self.class_names = ['Fish', 'Flower', 'Sugar', 'Gravel']\n",
    "        self.history = [[] for _ in range(len(self.class_names) + 1)] # to store per each class and also mean PR AUC\n",
    "        self.early_stopping_patience = early_stopping_patience\n",
    "        self.plateau_patience = plateau_patience\n",
    "        self.reduction_rate = reduction_rate\n",
    "        self.stage = stage\n",
    "        self.best_pr_auc = -float('inf')\n",
    "        if not os.path.exists(checkpoints_path):\n",
    "            os.makedirs(checkpoints_path)\n",
    "        self.checkpoints_path = checkpoints_path\n",
    "        \n",
    "    def compute_pr_auc(self, y_true, y_pred):\n",
    "        pr_auc_mean = 0\n",
    "        print(f\"\\n{'#'*30}\\n\")\n",
    "        for class_i in range(len(self.class_names)):\n",
    "            precision, recall, _ = precision_recall_curve(y_true[:, class_i], y_pred[:, class_i])\n",
    "            pr_auc = auc(recall, precision)\n",
    "            pr_auc_mean += pr_auc/len(self.class_names)\n",
    "            print(f\"PR AUC {self.class_names[class_i]}, {self.stage}: {pr_auc:.3f}\\n\")\n",
    "            self.history[class_i].append(pr_auc)        \n",
    "        print(f\"\\n{'#'*20}\\n PR AUC mean, {self.stage}: {pr_auc_mean:.3f}\\n{'#'*20}\\n\")\n",
    "        self.history[-1].append(pr_auc_mean)\n",
    "        return pr_auc_mean\n",
    "              \n",
    "    def is_patience_lost(self, patience):\n",
    "        if len(self.history[-1]) > patience:\n",
    "            best_performance = max(self.history[-1][-(patience + 1):-1])\n",
    "            return best_performance == self.history[-1][-(patience + 1)] and best_performance >= self.history[-1][-1]    \n",
    "              \n",
    "    def early_stopping_check(self, pr_auc_mean):\n",
    "        if self.is_patience_lost(self.early_stopping_patience):\n",
    "            self.model.stop_training = True    \n",
    "              \n",
    "    def model_checkpoint(self, pr_auc_mean, epoch):\n",
    "        if pr_auc_mean > self.best_pr_auc:\n",
    "            # remove previous checkpoints to save space\n",
    "            for checkpoint in glob.glob(os.path.join(self.checkpoints_path, 'classifier_densenet169_epoch_*')):\n",
    "                os.remove(checkpoint)\n",
    "            self.best_pr_auc = pr_auc_mean\n",
    "            self.model.save(os.path.join(self.checkpoints_path, f'classifier_densenet169_epoch_{epoch}_val_pr_auc_{pr_auc_mean}.h5'))              \n",
    "            print(f\"\\n{'#'*20}\\nSaved new checkpoint\\n{'#'*20}\\n\")\n",
    "              \n",
    "    def reduce_lr_on_plateau(self):\n",
    "        if self.is_patience_lost(self.plateau_patience):\n",
    "            new_lr = float(keras.backend.get_value(self.model.optimizer.lr)) * self.reduction_rate\n",
    "            keras.backend.set_value(self.model.optimizer.lr, new_lr)\n",
    "            print(f\"\\n{'#'*20}\\nReduced learning rate to {new_lr}.\\n{'#'*20}\\n\")\n",
    "        \n",
    "    def on_epoch_end(self, epoch, logs={}):\n",
    "        y_pred = self.model.predict_generator(self.data_generator, workers=self.num_workers)\n",
    "        y_true = self.data_generator.get_labels()\n",
    "        # estimate AUC under precision recall curve for each class\n",
    "        pr_auc_mean = self.compute_pr_auc(y_true, y_pred)\n",
    "              \n",
    "        if self.stage == 'val':\n",
    "            # early stop after early_stopping_patience=4 epochs of no improvement in mean PR AUC\n",
    "            self.early_stopping_check(pr_auc_mean)\n",
    "\n",
    "            # save a model with the best PR AUC in validation\n",
    "            self.model_checkpoint(pr_auc_mean, epoch)\n",
    "\n",
    "            # reduce learning rate on PR AUC plateau\n",
    "            self.reduce_lr_on_plateau()            \n",
    "        \n",
    "    def get_pr_auc_history(self):\n",
    "        return self.history"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Callback instances"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:46.057156Z",
     "start_time": "2019-11-16T15:18:46.033120Z"
    }
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'PrAucCallback' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-9-5597f53d23cb>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mtrain_metric_callback\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mPrAucCallback\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_generator_train_eval\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      2\u001b[0m \u001b[0mval_callback\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mPrAucCallback\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_generator_val\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstage\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'val'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'PrAucCallback' is not defined"
     ]
    }
   ],
   "source": [
    "train_metric_callback = PrAucCallback(data_generator_train_eval)\n",
    "val_callback = PrAucCallback(data_generator_val, stage='val')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Defining a model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:18:46.972521Z",
     "start_time": "2019-11-16T15:18:46.959900Z"
    }
   },
   "outputs": [],
   "source": [
    "from keras.losses import binary_crossentropy\n",
    "def dice_coef(y_true, y_pred, smooth=1):\n",
    "    y_true_f = K.flatten(y_true)\n",
    "    y_pred_f = K.flatten(y_pred)\n",
    "    intersection = K.sum(y_true_f * y_pred_f)\n",
    "    return (2. * intersection + smooth) / (K.sum(y_true_f) + K.sum(y_pred_f) + smooth)\n",
    "\n",
    "def dice_loss(y_true, y_pred):\n",
    "    smooth = 1.\n",
    "    y_true_f = K.flatten(y_true)\n",
    "    y_pred_f = K.flatten(y_pred)\n",
    "    intersection = y_true_f * y_pred_f\n",
    "    score = (2. * K.sum(intersection) + smooth) / (K.sum(y_true_f) + K.sum(y_pred_f) + smooth)\n",
    "    return 1. - score\n",
    "\n",
    "def bce_dice_loss(y_true, y_pred):\n",
    "    return binary_crossentropy(y_true, y_pred) + dice_loss(y_true, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-09T06:29:19.569056Z",
     "start_time": "2019-11-09T06:29:16.302065Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting git+https://github.com/qubvel/efficientnet\n",
      "  Cloning https://github.com/qubvel/efficientnet to /tmp/pip-req-build-9dw_xs7j\n",
      "  Running command git clone -q https://github.com/qubvel/efficientnet /tmp/pip-req-build-9dw_xs7j\n",
      "Requirement already satisfied, skipping upgrade: keras_applications<=1.0.8,>=1.0.7 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from efficientnet==1.0.0) (1.0.8)\n",
      "Requirement already satisfied, skipping upgrade: scikit-image in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from efficientnet==1.0.0) (0.15.0)\n",
      "Requirement already satisfied, skipping upgrade: h5py in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from keras_applications<=1.0.8,>=1.0.7->efficientnet==1.0.0) (2.9.0)\n",
      "Requirement already satisfied, skipping upgrade: numpy>=1.9.1 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from keras_applications<=1.0.8,>=1.0.7->efficientnet==1.0.0) (1.16.4)\n",
      "Requirement already satisfied, skipping upgrade: PyWavelets>=0.4.0 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from scikit-image->efficientnet==1.0.0) (1.0.3)\n",
      "Requirement already satisfied, skipping upgrade: scipy>=0.17.0 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from scikit-image->efficientnet==1.0.0) (1.3.0)\n",
      "Requirement already satisfied, skipping upgrade: pillow>=4.3.0 in /home/dex/.local/lib/python3.7/site-packages (from scikit-image->efficientnet==1.0.0) (6.1.0)\n",
      "Requirement already satisfied, skipping upgrade: imageio>=2.0.1 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from scikit-image->efficientnet==1.0.0) (2.5.0)\n",
      "Requirement already satisfied, skipping upgrade: networkx>=2.0 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from scikit-image->efficientnet==1.0.0) (2.3)\n",
      "Requirement already satisfied, skipping upgrade: matplotlib!=3.0.0,>=2.0.0 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from scikit-image->efficientnet==1.0.0) (3.1.0)\n",
      "Requirement already satisfied, skipping upgrade: six in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from h5py->keras_applications<=1.0.8,>=1.0.7->efficientnet==1.0.0) (1.12.0)\n",
      "Requirement already satisfied, skipping upgrade: decorator>=4.3.0 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from networkx>=2.0->scikit-image->efficientnet==1.0.0) (4.4.0)\n",
      "Requirement already satisfied, skipping upgrade: cycler>=0.10 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0) (0.10.0)\n",
      "Requirement already satisfied, skipping upgrade: kiwisolver>=1.0.1 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0) (1.1.0)\n",
      "Requirement already satisfied, skipping upgrade: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0) (2.4.0)\n",
      "Requirement already satisfied, skipping upgrade: python-dateutil>=2.1 in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0) (2.8.0)\n",
      "Requirement already satisfied, skipping upgrade: setuptools in /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages (from kiwisolver>=1.0.1->matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet==1.0.0) (41.0.1)\n",
      "Building wheels for collected packages: efficientnet\n",
      "  Building wheel for efficientnet (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Stored in directory: /tmp/pip-ephem-wheel-cache-ltry0q8d/wheels/64/60/2e/30ebaa76ed1626e86bfb0cc0579b737fdb7d9ff8cb9522663a\n",
      "Successfully built efficientnet\n",
      "Installing collected packages: efficientnet\n",
      "  Found existing installation: efficientnet 1.0.0\n",
      "    Uninstalling efficientnet-1.0.0:\n",
      "      Successfully uninstalled efficientnet-1.0.0\n",
      "Successfully installed efficientnet-1.0.0\n"
     ]
    }
   ],
   "source": [
    "!pip install -U git+https://github.com/qubvel/efficientnet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:19:03.209192Z",
     "start_time": "2019-11-16T15:18:51.878324Z"
    },
    "_kg_hide-output": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: Logging before flag parsing goes to stderr.\n",
      "W1116 18:18:51.900685 140150228264768 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:95: The name tf.reset_default_graph is deprecated. Please use tf.compat.v1.reset_default_graph instead.\n",
      "\n",
      "W1116 18:18:51.901755 140150228264768 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:98: The name tf.placeholder_with_default is deprecated. Please use tf.compat.v1.placeholder_with_default instead.\n",
      "\n",
      "W1116 18:18:51.909141 140150228264768 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:102: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
      "\n",
      "W1116 18:18:51.910609 140150228264768 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
      "\n",
      "W1116 18:18:51.913436 140150228264768 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:4185: The name tf.truncated_normal is deprecated. Please use tf.random.truncated_normal instead.\n",
      "\n",
      "W1116 18:18:54.584877 140150228264768 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:1834: The name tf.nn.fused_batch_norm is deprecated. Please use tf.compat.v1.nn.fused_batch_norm instead.\n",
      "\n",
      "W1116 18:18:54.928814 140150228264768 deprecation.py:506] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n"
     ]
    }
   ],
   "source": [
    "import efficientnet.keras as efn \n",
    "def get_model():\n",
    "    K.clear_session()\n",
    "    base_model =  efn.EfficientNetB2(weights='imagenet', include_top=False, pooling='avg', input_shape=(260, 260, 3))\n",
    "    x = base_model.output\n",
    "    y_pred = Dense(4, activation='sigmoid')(x)\n",
    "    return Model(inputs=base_model.input, outputs=y_pred)\n",
    "\n",
    "model = get_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:19:03.252299Z",
     "start_time": "2019-11-16T15:19:03.234324Z"
    }
   },
   "outputs": [],
   "source": [
    "from keras_radam import RAdam"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Initial tuning of the added fully-connected layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-12T18:40:48.163090Z",
     "start_time": "2019-11-12T18:40:42.207968Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "W1112 21:40:42.237008 139670237226816 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
      "\n",
      "W1112 21:40:42.315240 139670237226816 deprecation.py:323] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-21-62fe69b460be>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      8\u001b[0m                               \u001b[0mcallbacks\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtrain_metric_callback\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_callback\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m                               \u001b[0mworkers\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnum_cores\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m                               \u001b[0mverbose\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     11\u001b[0m                              )\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/legacy/interfaces.py\u001b[0m in \u001b[0;36mwrapper\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m     89\u001b[0m                 warnings.warn('Update your `' + object_name + '` call to the ' +\n\u001b[1;32m     90\u001b[0m                               'Keras 2 API: ' + signature, stacklevel=2)\n\u001b[0;32m---> 91\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     92\u001b[0m         \u001b[0mwrapper\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_original_function\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     93\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mwrapper\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/engine/training.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(self, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m   1416\u001b[0m             \u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0muse_multiprocessing\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1417\u001b[0m             \u001b[0mshuffle\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mshuffle\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1418\u001b[0;31m             initial_epoch=initial_epoch)\n\u001b[0m\u001b[1;32m   1419\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1420\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0minterfaces\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlegacy_generator_methods_support\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/engine/training_generator.py\u001b[0m in \u001b[0;36mfit_generator\u001b[0;34m(model, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)\u001b[0m\n\u001b[1;32m    215\u001b[0m                 outs = model.train_on_batch(x, y,\n\u001b[1;32m    216\u001b[0m                                             \u001b[0msample_weight\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0msample_weight\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 217\u001b[0;31m                                             class_weight=class_weight)\n\u001b[0m\u001b[1;32m    218\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    219\u001b[0m                 \u001b[0mouts\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mto_list\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mouts\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/engine/training.py\u001b[0m in \u001b[0;36mtrain_on_batch\u001b[0;34m(self, x, y, sample_weight, class_weight)\u001b[0m\n\u001b[1;32m   1215\u001b[0m             \u001b[0mins\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mx\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0my\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0msample_weights\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1216\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_train_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1217\u001b[0;31m         \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_function\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mins\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1218\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0munpack_singleton\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1219\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m   2713\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_legacy_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2714\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2715\u001b[0;31m             \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2716\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2717\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mpy_any\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mis_tensor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mx\u001b[0m \u001b[0;32min\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py\u001b[0m in \u001b[0;36m_call\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m   2669\u001b[0m                                 \u001b[0mfeed_symbols\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2670\u001b[0m                                 \u001b[0msymbol_vals\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2671\u001b[0;31m                                 session)\n\u001b[0m\u001b[1;32m   2672\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2673\u001b[0m             \u001b[0mfetched\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_callable_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0marray_vals\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun_metadata\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py\u001b[0m in \u001b[0;36m_make_callable\u001b[0;34m(self, feed_arrays, feed_symbols, symbol_vals, session)\u001b[0m\n\u001b[1;32m   2621\u001b[0m             \u001b[0mcallable_opts\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun_options\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mCopyFrom\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun_options\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2622\u001b[0m         \u001b[0;31m# Create callable.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2623\u001b[0;31m         \u001b[0mcallable_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msession\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_make_callable_from_options\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcallable_opts\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2624\u001b[0m         \u001b[0;31m# Cache parameters corresponding to the generated callable, so that\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2625\u001b[0m         \u001b[0;31m# we can detect future mismatches and refresh the callable.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_make_callable_from_options\u001b[0;34m(self, callable_options)\u001b[0m\n\u001b[1;32m   1487\u001b[0m     \"\"\"\n\u001b[1;32m   1488\u001b[0m     \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_extend_graph\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1489\u001b[0;31m     \u001b[0;32mreturn\u001b[0m \u001b[0mBaseSession\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_Callable\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcallable_options\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1490\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1491\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/envs/keras_env/lib/python3.7/site-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, session, callable_options)\u001b[0m\n\u001b[1;32m   1444\u001b[0m       \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1445\u001b[0m         self._handle = tf_session.TF_SessionMakeCallable(\n\u001b[0;32m-> 1446\u001b[0;31m             session._session, options_ptr)\n\u001b[0m\u001b[1;32m   1447\u001b[0m       \u001b[0;32mfinally\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1448\u001b[0m         \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_DeleteBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moptions_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "for base_layer in model.layers[:-3]:\n",
    "    base_layer.trainable = False\n",
    "    \n",
    "model.compile(optimizer=RAdam(warmup_proportion=0.1, min_lr=1e-5),  loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "history_0 = model.fit_generator(generator=data_generator_train,\n",
    "                              validation_data=data_generator_val,\n",
    "                              epochs=20,\n",
    "                              callbacks=[train_metric_callback, val_callback],\n",
    "                              workers=num_cores,\n",
    "                              verbose=1\n",
    "                             )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fine-tuning the whole model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After unfreezing all the layers(except last 3) I set a less aggressive initial learning rate and train until early stopping (or 100 epochs max)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2/20\n",
      "138/138 [==============================] - 231s 2s/step - loss: 2.7147 - accuracy: 0.4382 - val_loss: 3.0616 - val_accuracy: 0.4090\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.742\n",
      "\n",
      "PR AUC Flower, train: 0.865\n",
      "\n",
      "PR AUC Sugar, train: 0.840\n",
      "\n",
      "PR AUC Gravel, train: 0.782\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.807\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.701\n",
      "\n",
      "PR AUC Flower, val: 0.851\n",
      "\n",
      "PR AUC Sugar, val: 0.843\n",
      "\n",
      "PR AUC Gravel, val: 0.777\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.793\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Saved new checkpoint\n",
      "####################\n",
      "\n",
      "Epoch 3/20\n",
      "138/138 [==============================] - 193s 1s/step - loss: 2.6637 - accuracy: 0.4708 - val_loss: 2.9207 - val_accuracy: 0.4577\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.776\n",
      "\n",
      "PR AUC Flower, train: 0.876\n",
      "\n",
      "PR AUC Sugar, train: 0.876\n",
      "\n",
      "PR AUC Gravel, train: 0.800\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.832\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.736\n",
      "\n",
      "PR AUC Flower, val: 0.849\n",
      "\n",
      "PR AUC Sugar, val: 0.863\n",
      "\n",
      "PR AUC Gravel, val: 0.771\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.805\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Saved new checkpoint\n",
      "####################\n",
      "\n",
      "Epoch 4/20\n",
      "138/138 [==============================] - 195s 1s/step - loss: 2.6310 - accuracy: 0.4735 - val_loss: 2.8695 - val_accuracy: 0.5221\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.801\n",
      "\n",
      "PR AUC Flower, train: 0.887\n",
      "\n",
      "PR AUC Sugar, train: 0.886\n",
      "\n",
      "PR AUC Gravel, train: 0.825\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.849\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.743\n",
      "\n",
      "PR AUC Flower, val: 0.868\n",
      "\n",
      "PR AUC Sugar, val: 0.872\n",
      "\n",
      "PR AUC Gravel, val: 0.802\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.821\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Saved new checkpoint\n",
      "####################\n",
      "\n",
      "Epoch 5/20\n",
      "138/138 [==============================] - 194s 1s/step - loss: 2.6088 - accuracy: 0.4742 - val_loss: 3.0059 - val_accuracy: 0.4338\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.784\n",
      "\n",
      "PR AUC Flower, train: 0.874\n",
      "\n",
      "PR AUC Sugar, train: 0.892\n",
      "\n",
      "PR AUC Gravel, train: 0.818\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.842\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.710\n",
      "\n",
      "PR AUC Flower, val: 0.839\n",
      "\n",
      "PR AUC Sugar, val: 0.869\n",
      "\n",
      "PR AUC Gravel, val: 0.799\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.804\n",
      "####################\n",
      "\n",
      "Epoch 6/20\n",
      "138/138 [==============================] - 196s 1s/step - loss: 2.5976 - accuracy: 0.4712 - val_loss: 2.8583 - val_accuracy: 0.5046\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.816\n",
      "\n",
      "PR AUC Flower, train: 0.912\n",
      "\n",
      "PR AUC Sugar, train: 0.900\n",
      "\n",
      "PR AUC Gravel, train: 0.837\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.866\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.739\n",
      "\n",
      "PR AUC Flower, val: 0.874\n",
      "\n",
      "PR AUC Sugar, val: 0.903\n",
      "\n",
      "PR AUC Gravel, val: 0.798\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.828\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Saved new checkpoint\n",
      "####################\n",
      "\n",
      "Epoch 7/20\n",
      "138/138 [==============================] - 195s 1s/step - loss: 2.5805 - accuracy: 0.4871 - val_loss: 2.8421 - val_accuracy: 0.4403\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.804\n",
      "\n",
      "PR AUC Flower, train: 0.904\n",
      "\n",
      "PR AUC Sugar, train: 0.900\n",
      "\n",
      "PR AUC Gravel, train: 0.844\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.863\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.733\n",
      "\n",
      "PR AUC Flower, val: 0.868\n",
      "\n",
      "PR AUC Sugar, val: 0.885\n",
      "\n",
      "PR AUC Gravel, val: 0.799\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.821\n",
      "####################\n",
      "\n",
      "Epoch 8/20\n",
      "138/138 [==============================] - 198s 1s/step - loss: 2.5521 - accuracy: 0.4812 - val_loss: 2.9676 - val_accuracy: 0.3805\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.843\n",
      "\n",
      "PR AUC Flower, train: 0.912\n",
      "\n",
      "PR AUC Sugar, train: 0.900\n",
      "\n",
      "PR AUC Gravel, train: 0.840\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.874\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.758\n",
      "\n",
      "PR AUC Flower, val: 0.864\n",
      "\n",
      "PR AUC Sugar, val: 0.872\n",
      "\n",
      "PR AUC Gravel, val: 0.784\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.819\n",
      "####################\n",
      "\n",
      "Epoch 9/20\n",
      "138/138 [==============================] - 200s 1s/step - loss: 2.5402 - accuracy: 0.4835 - val_loss: 2.8360 - val_accuracy: 0.4926\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.844\n",
      "\n",
      "PR AUC Flower, train: 0.912\n",
      "\n",
      "PR AUC Sugar, train: 0.894\n",
      "\n",
      "PR AUC Gravel, train: 0.847\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.875\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.765\n",
      "\n",
      "PR AUC Flower, val: 0.853\n",
      "\n",
      "PR AUC Sugar, val: 0.889\n",
      "\n",
      "PR AUC Gravel, val: 0.806\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.828\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Reduced learning rate to 0.0005000000237487257.\n",
      "####################\n",
      "\n",
      "Epoch 10/20\n",
      "138/138 [==============================] - 195s 1s/step - loss: 2.4682 - accuracy: 0.5016 - val_loss: 2.8715 - val_accuracy: 0.4972\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.879\n",
      "\n",
      "PR AUC Flower, train: 0.930\n",
      "\n",
      "PR AUC Sugar, train: 0.919\n",
      "\n",
      "PR AUC Gravel, train: 0.887\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.904\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.783\n",
      "\n",
      "PR AUC Flower, val: 0.882\n",
      "\n",
      "PR AUC Sugar, val: 0.901\n",
      "\n",
      "PR AUC Gravel, val: 0.818\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.846\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Saved new checkpoint\n",
      "####################\n",
      "\n",
      "Epoch 11/20\n",
      "138/138 [==============================] - 196s 1s/step - loss: 2.4290 - accuracy: 0.4891 - val_loss: 2.9092 - val_accuracy: 0.4724\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.892\n",
      "\n",
      "PR AUC Flower, train: 0.941\n",
      "\n",
      "PR AUC Sugar, train: 0.919\n",
      "\n",
      "PR AUC Gravel, train: 0.895\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.912\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.775\n",
      "\n",
      "PR AUC Flower, val: 0.880\n",
      "\n",
      "PR AUC Sugar, val: 0.886\n",
      "\n",
      "PR AUC Gravel, val: 0.817\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.839\n",
      "####################\n",
      "\n",
      "Epoch 12/20\n",
      "138/138 [==============================] - 199s 1s/step - loss: 2.4015 - accuracy: 0.5034 - val_loss: 2.9748 - val_accuracy: 0.4697\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.895\n",
      "\n",
      "PR AUC Flower, train: 0.941\n",
      "\n",
      "PR AUC Sugar, train: 0.924\n",
      "\n",
      "PR AUC Gravel, train: 0.897\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.914\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.754\n",
      "\n",
      "PR AUC Flower, val: 0.873\n",
      "\n",
      "PR AUC Sugar, val: 0.888\n",
      "\n",
      "PR AUC Gravel, val: 0.799\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.829\n",
      "####################\n",
      "\n",
      "Epoch 13/20\n",
      "138/138 [==============================] - 197s 1s/step - loss: 2.3766 - accuracy: 0.5036 - val_loss: 2.8061 - val_accuracy: 0.5368\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.888\n",
      "\n",
      "PR AUC Flower, train: 0.948\n",
      "\n",
      "PR AUC Sugar, train: 0.922\n",
      "\n",
      "PR AUC Gravel, train: 0.908\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.916\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.724\n",
      "\n",
      "PR AUC Flower, val: 0.878\n",
      "\n",
      "PR AUC Sugar, val: 0.882\n",
      "\n",
      "PR AUC Gravel, val: 0.800\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.821\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Reduced learning rate to 0.0002500000118743628.\n",
      "####################\n",
      "\n",
      "Epoch 14/20\n",
      "138/138 [==============================] - 195s 1s/step - loss: 2.3018 - accuracy: 0.4946 - val_loss: 2.9496 - val_accuracy: 0.5074\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.922\n",
      "\n",
      "PR AUC Flower, train: 0.953\n",
      "\n",
      "PR AUC Sugar, train: 0.934\n",
      "\n",
      "PR AUC Gravel, train: 0.929\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.934\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.763\n",
      "\n",
      "PR AUC Flower, val: 0.895\n",
      "\n",
      "PR AUC Sugar, val: 0.874\n",
      "\n",
      "PR AUC Gravel, val: 0.801\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.833\n",
      "####################\n",
      "\n",
      "\n",
      "####################\n",
      "Reduced learning rate to 0.0001250000059371814.\n",
      "####################\n",
      "\n",
      "Epoch 15/20\n",
      "138/138 [==============================] - 197s 1s/step - loss: 2.2442 - accuracy: 0.5054 - val_loss: 3.0402 - val_accuracy: 0.4559\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, train: 0.938\n",
      "\n",
      "PR AUC Flower, train: 0.962\n",
      "\n",
      "PR AUC Sugar, train: 0.941\n",
      "\n",
      "PR AUC Gravel, train: 0.939\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, train: 0.945\n",
      "####################\n",
      "\n",
      "\n",
      "##############################\n",
      "\n",
      "PR AUC Fish, val: 0.763\n",
      "\n",
      "PR AUC Flower, val: 0.888\n",
      "\n",
      "PR AUC Sugar, val: 0.893\n",
      "\n",
      "PR AUC Gravel, val: 0.802\n",
      "\n",
      "\n",
      "####################\n",
      " PR AUC mean, val: 0.837\n",
      "####################\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for base_layer in model.layers[:-3]:\n",
    "    base_layer.trainable = True\n",
    "    \n",
    "model.compile(optimizer=RAdam(warmup_proportion=0.1, min_lr=1e-5),  loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "history_1 = model.fit_generator(generator=data_generator_train,\n",
    "                              validation_data=data_generator_val,\n",
    "                              epochs=20,\n",
    "                              callbacks=[train_metric_callback, val_callback],\n",
    "                              workers=num_cores,\n",
    "                              verbose=1,\n",
    "                              initial_epoch=1\n",
    "                             )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualizing train and val PR AUC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_with_dots(ax, np_array):\n",
    "    ax.scatter(list(range(1, len(np_array) + 1)), np_array, s=50)\n",
    "    ax.plot(list(range(1, len(np_array) + 1)), np_array)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmkAAAHDCAYAAAB/Ho3HAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzs3Xl8VOXZ//HPlYQkZAEkQVYhgFRBBMSAa+ta64I7bq1769Lt+VVrW0sXrXaxdtM+rbbqA1prResuWm3r2rqxqKCCKMoAIWwJCGSG7Pfvj3sCcZgJkzBr8n2/XnlN5pwz51wzmWSu3Mt1m3MOEREREcksOekOQERERER2piRNREREJAMpSRMRERHJQErSRERERDKQkjQRERGRDKQkTURERCQDKUkT6SQzKzEzZ2ZzEnCu+WZWl4i4uotEvr6JZGaTwnHdErH9sfD2fp041ydm9nbio/zUNaLGKyLZQ0maZI3wB05nvi5Od8ySXGb2RPhnfUEcx74WPvbYVMSWbGbWL/x8Hkt3LPFqlzi2/2oys+pwsntUnMevNbMnzewLuxnP19udd2oHx90SPuZbHRzzrY6SYjPra2bfM7OXzWxD+HlsMrNXzex6Mxu2O89Fuqe8dAcg0gk/ibLtW0Bf4Fbgk4h9yWqpCAJjgUS0gJ0JFCTgPD3VHcDJwGXAvbEOMrNxwMHAx8BzCY7h68C1wJYEn3d3LcG/TzemO5Ao1gF/Cn9fBBwInAqcamYXO+fu6eD43sAEYBowzcwuc87d1cU4LgMcYMDlwNwunqdDZnY0MBsYAASAJ/DPqRT/3H8EfM/M9nfOLUtGDJKdlKRJ1nDOXR+5Ldxa1he4xTkXSFEcDng/QedakYjz9GD/AFYBnzWzfZxzS2Mcd1n49i6X4GVWnHOrE3m+RHHONZCg92kSrI38fTazq4DfAr82s78651p2cfyVwO3ATWY2K+L4XTKzg4CJwEPAJOBcM7vKObe108+m4+scCMwBegHfAG53zrVGHLMP8BugJJHXluyn7k7p9trGfZlZbzP7qZktM7NGM/tDeH+ZmV1rZi+Fu10azWydmT1sZpOjnC/qmCkz+3V4e6WZfcnMFpjZNjOrMbN7zWzPWLFFbJsWPs81ZjbVzJ41s83h5/Dv8B/9aM9zuJn9NXy9UPj657Q/X5yvV5dfDzMbZGZ3m9l6M6s3s0Vmdl6M6xSa2Y1mFjCzBjP7yMx+jP8wi0v4g3lm+O5XYlynALgAaAZmtdteEb7+6+Hn12hmq8zsHjMbFW8MFmNMmpnlmtl3zGxp+LVYGX6PFMU4T7mZzYh43dea2UNmdkDEsd8CNoXvnmqf7g78VviYmGPSzGyEmd0Zfr5tP98HzGx8lGPbuvFOM7MTzewVMwuaH1f3SGdeq134E9AClAPxnPPu8PFlwMguXK8tcZ8F/AUoBr7YhfPsyu341r8fOOf+GJmgATjnljrnpgGLk3B9yWJqSZOeIgf/3+w+wLNALdDWinUAviv1ReBxYDP+j/4p+O6UzzvnXu7Etb6L74p5HHgBOAw4HxhvZpWd+I//cOCn4bjuxH9wnQa8aGbj27fCmR/P8howBN+dNw8YCtyDb23qjK6+HgOA1/HJw/34D71zgL+ZWaNz7uF28ebgu3w+DywFfo/v9voGvvunM/4P3110oZnNcM41Rew/Hf9B/qhzbm277cfju8tfwL9e24B9gS8BJ5vZ1N3sepoJXAisZEdX3VnAZKL/g1wJ/JhPv+6j2PG6H+OceyV87OvAL4Hv4V+/2e3O83pHQZnv+n0Jnwz9A/hr+Dpnhq9zknPuxSgP/RK+S3IOcBv+fXI6cGD4/ZiIFigL38bb2tl2fOTPvOMHmZUC5wJr8H8P3sW/5y8H/tyZc+3iOgcAU/C/E7/f1fHOucZEXVu6CeecvvSVtV/48R0OqOjgmPnhY+YC/aLs7w/sEWX7aKAGmBexvSR8vjkR238d3l4LfKbddsN/6DrgxCix1UVsmxY+1gHTI/Z9O7z95ojtD4S3/yhi+8H4FiQHXBPna9rV18PhxwbmtNtXCbQCcyMec3n4+OeAXu22DwSqor2+u4j5qWivV3jfc+F9x0dsHwwURTn+UKABuD9i+6TweW6J2P5YeHu/dttODG9bBJS0214a3uaAt6O87tHen2Pw48pei9jeL3yex2K8JrHifT28/esR278Q3r4KyG+3/Vvh7fXAQRGPuT287/I4f05tMb0dZd/V4X3rgdw4jv9GeF+g/XsuzjiuDD/2l1HeJwdGOf6W8L5vdXDOttfplnbbvtnZ97K+9NX+S92d0pN83zkXObkA59xG59ymKNs/wrf2VJpZWSeu8yvn3AftzuOAtoHNMWeQRfGsc+6hiG13RJ4n3CpwBv7D7VftD3bOvQ78vRPX3J3XYxNwrWvXneOcmw+8BRxgZu1b7i8J317r2rV8OefWATd1Jt6wttflU12e4a64o/Ctpv+MeD5rnHOhyBM5514F3sAnLV3V9vyuc85t7852vrXpx9EeEH7do70/P8S3Xh1kZn13IybMbD/gIHy32m0R13k2fJ1h+FbGSHc5596I2LbT+zFOg8zPaLzezH5pZv/Gj8kC+I7bubU58vhngP/Ft35e5qJ0Ie5CW1fn3e223R2xLxEGh2+rEnhO6UHU3Sk9ScyZW+an/n8T/2GzJzuPixqCbyGLx/wo21aFb/eI8xxRz+Oc22pmmyPOMx7/u7zAOVcf5Tz/xXftxK2Lr8di59y2KKdbhe/iK2XHOKoDgG3OuXlRjn+xM7GGPQVUA583sxFuR1fwV/AtmXdF+yA3s+nhYw7At2TlRezvHeM57Urb2L2Xoux7IdaDzOwYfAvRVHz3cbTXfXMX4omM64XwPw+Rnse35B6AT8jbS9T7GnyL6XXh71vwLbRPALc6557fxfFt6vCto69EOT6m8LjKycAbzrkl7XY9DPwR+KKZfds5F+zMeWNdLnyb0Mkq0nMoSZOeIuRijJkxs/PxA4frgH8By/FlNhxwHHAInSuTsVNrCL7LESB3N8/Tdq7252lrXVkX4/hY26Pajdejo3ghHLOZFYYfH4hx/NoY22NyzjWb2SzgB8ClwHXhlruL8UnArMjHmNl1wPXABnxX1yp8y4zDJ7X7hOPsSpLWF2h2zu1U/sI5t9nMGqLEcxG+NWcL/nUPsON1Px7fAra75Vra3itrYuxv2x6tMG+i3tcAC51zk7pyfHiCxjT8OL/HzGyK69zM7ivCt3e33+icC5nZg8CXgfPY0foNvsseOp5s17av/T8D1eFb1UCTLlGSJj1FR//J/hTYChzgnPu4/Q4zG4NPSjJZW32ugTH2x9oeS1JfD+dcfThJiRXXoC6e+i7g+8ClZvYT4CR8d9MTLqJMhpkVh4/9GJgSmUyZ2QldjKHNZqDMzPpHOXdfoidbP8MnQgdEJh1mNhafpO2utla4WK/x4IjjMk64S/ivZubwkx5mAkfH81gzK8EnYAC3m9ntMQ69nE8naW2vR0fDHsrDt+2T2f+Gbw8zs8IYLd0iMWlMmvRo4daWEfiByZEJSS8yP0EDeAffonFguJUq0uHxniiFr8dbQG8zmxJl35FdOWE4sfkXO8ZUtY1PuyPK4XvhE6UXoyRRZcB+XYmhnTfDt0dE2XdU5Ibwz20o8GaUBC0fPwEkUtu4rc60Yr3VFpeZWZT9bbG9GWVfRnHO3Yef0XyUmZ0U58POxXe7v4OfFRztazUwxcwmtnvcwvBtR+//tn1tx+Kcews/c3gPfDd2h8I/a5HtlKRJj+aca8b/Ud7PzNr+E24rEfELulZ/KaXC3biP4ceOfaf9PvMFO8/qxLlS9Xq0dT/eFE7+2q4zEF+9v6vuDN/+ADgBP2D7mSjHrcR3Sx1svo5a2/UL8TMWi3cjBtjx/H4Sbr1pO38pcEPkweEWlnXA/u0nZZhZLnAzMDzKNYL4WajR9kXlnHsXnzTsj+/W2878clkn43/+0V6zTPSj8O1PYySdkS4P317jnPtKtC/gdxHHgn891uITwtMjTxredmT4mGcjdn8VPzP252Z2Zfh3KfLxe5vZE8C4OJ6D9CDq7hTxf5R/DSwys0fwH95HABX4OlK72/WVCt/Gt5jdYGafw38QDwPOBp7E11eLdwZcKl6Pu/DJ47Hh68zBF/w8G986MrSL530C/0F5aPj+/0WZKdg2/ugOfCmGRWb2FD4xOwbID8fQ5VZD59zTZnYvvojue+HXEfws3I/x3cmRfoef2brQzB4NbzsCn4Q9S8RsU+dcq5m9ABxvZn/H1/pqAf7pnOtoeaNL8RMa7jSz0/AlQdrqpNUDF7osqdflnHvOzF7Cv05n40vRRBVuGZuCn+n77w5Oew/wc+BLZvYd51wo3EV/Pv799XB4Nmpba+OB+PdNCDg/skvTObcg3NL3AP4fgO+a2fP4pLwEP4nhEKARX4ZEZDu1pIn4pWiuxM9WvBQ/ZuUD/Ay7rKgA7pxbie8Sux//R/8qfJfdRfgabRD/2pJJfz3CMy1PwY/DKgL+B5/8/QE/2L+r521ix4DwVnz3VSzfYkdLzNfwg9Ffwj/P9V2NoZ1L8MVmG8Lnn45fgmga0RPmX4WP24Rv5ToHv6zTFGIv73Q5flbikfjSHjeyI0GNKtyadiB+LNckfOvr0cCj+Dpo0WZXZrK2n+EN4ZbHWNpaxmZ2VLLDOVeD/53pi/8ZtG1/Dj/rta2w9P8Lf40Mb5sUPibaOZ8H9sa3Eq/G/9P0XfzvZwH+92CM07qdEsGiz8IWke7CzG7FJ0GHd7ZcgYiIpI+SNJFuwsyGOOeqI7ZNAV7GV6wfER5zJiIiWUBj0kS6jyVm9ibwHn5s0T7sGD/2dSVoIiLZRS1pIt2Emf0Cv2bkcPyA5E3Aq/h1Pl9NZ2wiItJ5StJEREREMlC36O4sLy93FRUV6Q5DREREZJcWLFhQ45wbsKvjukWSVlFRwfz50db+FREREcksZrYinuNUJ01EREQkAylJExEREclAStJEREREMlC3GJMWTVNTE1VVVdTX1+/64G6isLCQYcOG0atXr10fLCIiIhmt2yZpVVVVlJaWUlFRgZmlO5ykc85RW1tLVVUVI0eOTHc4IiIispu6bXdnfX09ZWVlPSJBAzAzysrKelTLoYiISHfWbZM0oMckaG162vMVERHpzrp1kiYiIiKSrZSkJUltbS2TJk1i0qRJDBo0iKFDh26/39jYGNc5LrnkEpYuXZrkSEVERCQTdduJA+lWVlbG22+/DcD1119PSUkJ11xzzaeOcc7hnCMnJ3quPGvWrKTHKSIiIplJLWkptmzZMsaPH8+VV17J5MmTWbNmDZdffjmVlZXst99+3HDDDduPPfzww3n77bdpbm6mX79+XHvttUycOJFDDjmE9evXp/FZiIiISLL1iJa0nzz5HourtyT0nOOG9OG6k/fr0mMXL17MrFmz+NOf/gTATTfdRP/+/Wlubuaoo45i+vTpjBs37lOP2bx5M0cccQQ33XQTV199NTNnzuTaa6/d7echIiIimUktaWkwevRopkyZsv3+/fffz+TJk5k8eTJLlixh8eLFOz2md+/enHDCCQAceOCBBAKBVIUrIiIiadAjWtK62uKVLMXFxdu///DDD7n11luZO3cu/fr14/zzz49a6yw/P3/797m5uTQ3N6ckVhEREUkPtaSl2ZYtWygtLaVPnz6sWbOGZ599Nt0hiYiISAboES1pmWzy5MmMGzeO8ePHM2rUKA477LB0hyQiIiIZwJxz6Y5ht1VWVrr58+d/atuSJUsYO3ZsmiJKn576vEVERLKFmS1wzlXu6jh1d4qIiIhkICVpIiIiIhlIY9JERESkx6lraGbOwmoCtUEqyoqZNnEIJQWZlRZlVjQiIiIiSTYvsJGLZ83FOQg1tlCUn8uNTy3m7kumMqWif7rD207dnSIiItJj1DU0c/GsuQQbWgg1tgA+UQs2tIS3Z04dUiVpIiIi0mPMWVhNrMIWzsGcRdWpDagDStKS5Mgjj9ypMO0tt9zC1772tZiPKSkpSXZYIiIiPdrymrrtLWiRQo0tBGpCKY4oNiVpSXLeeecxe/bsT22bPXs25513XpoiEhER6blaWh1PLqzmsbdjt5QV5edSUV6Uwqg6piQtSaZPn86cOXNoaGgAIBAIUF1dzaRJkzjmmGOYPHky+++/P48//niaIxUREem+mltaeeyt1Rz3u5f45v1vUVyQR0GeRT3WDKZNGJLiCGPrGbM7/3EtrH0nsecctD+ccFPM3WVlZUydOpVnnnmGU089ldmzZ3POOefQu3dvHn30Ufr06UNNTQ0HH3wwp5xyCmbR3zAiIiLSec0trTz+djV/eGEZy2uC7DOwlD9+cTInjB/EgpWbdprdaQZ3XzKV4gwqw5E5kXRDbV2ebUnazJkzcc4xY8YMXn75ZXJycli9ejXr1q1j0KBB6Q5XREQk6zW1tPLom6v5wwvLWLkxxNjBffjT+ZM5btwgcnJ8g8iUiv7MnXEscxZVE6gJUVFexLQJQzIqQYOekqR10OKVTKeddhpXX301b775Jtu2bWPy5MncfffdbNiwgQULFtCrVy8qKiqor69PS3wiIiLZJlYR2sbmVh5aUMVtLy6jatM29h/alzsvrOTYsXtG7a0qLsjjnCnD0/AM4tczkrQ0KSkp4cgjj+TSSy/dPmFg8+bN7LnnnvTq1YsXXniBFStWpDlKERGR7BCtCO0Nc97jiweN4OlFa6jeXM/Evfpx46njOXKfAVk/lEhJWpKdd955nHHGGdtnen7pS1/i5JNPprKykkmTJrHvvvumOUIREZHM174IbZu2Uhp3/Wc5k/bqxy/OnMDnxpRnfXLWRklakp1++um4dlXzysvLee2116IeW1dXl6qwREREskpHRWgL8nI4d8owjvjMgNQGlWQqwSEiIiIZL1AbjFmEtqG5lRW121IcUfIpSRMREZGMN6Rvb3JzondjZloR2kTp1kmai9Uu2k31tOcrIiI9Q/Un25g9bxUtrdE/5zKtCG2idNskrbCwkNra2h6TuDjnqK2tpbCwMN2hiIiIJMyCFRs55Q+vsHJjiGtP2JfiglyK8nMB34JWXJCbcUVoE6X7PaOwYcOGUVVVxYYNG9IdSsoUFhYybNiwdIchIiKSEA/OW8UPHnuHof16c/9lBzFmYCkXHDwi44vQJkr3fFZAr169GDlyZLrDEBERkU5qbmnlZ08vYdYrAQ7fu5w/fPEA+hXlA9lRhDZRum2SJiIiItnnk1Aj3/jbW/x3WQ2XHjaSGSfuS15utx2d1SElaSIiIpIRPly3la/8ZT7Vn2zj5jMncPaUvdIdUlopSRMREZG0e27JOv7f7Lcp7JXL7MsP5sAR/dMdUtopSRMREZG0cc5x+0sf8atnl7LfkD7ccUElQ/r1TndYGUFJmoiIiKRFfVML331oEU8srObkiUO4+cwJ9A6X1xAlaSIiIpICdQ3NzFlYTaA2SEVZMZUV/bnqgbd5t3oz3/nCPnztyNHdZmH0RFGSJiIiIh2KTLCmTRxCSSdqk80LbOTiWXNxDkKNLRTk5dDY3EpBrxzuvKCSY8cNTGL02UtJmoiIiMQUmWAV5edy41OLufuSqUyp2PXg/rqGZi6eNZdgw47F0RuaWwEw4JDRZckKPespSRMREemmdrcFLFqCFWr03188ay5zZxxLcUEeocZm1m1pYN2WetZvbWD9lvrt379TtZlQu8e3Z2bMWVTdY4rTdpaSNBERkW5od1vAAB57qyrmoubbGls48lcvsq2phbqG5p32F+TlMLBPIY3NLcRaRTvU2EKgJhTvU+pxlKSJiIh0M/G0gOXmGOu3NFC9eRtrN9ezZnM9azdv87db/P0NWxtiXqPVQWlhHidNGMzAPoXsWVrAwD6FDOxTwJ6lhfTpnYeZMXvuSm6Ys3j79dsrys+lorwo8S9AN6EkTUREpJuZs7AaF6P5KtTYwtSf//tTCVyb0sI8BvctZHDf3owb3IeaugZe/mADjS07n6woP5crjhi1y67KaROHcONTi6PuM4NpE4bs+gn1UErSREREupHWVscrH9VEbbkCcA5G9C/mhPGDGBROyAb1LWRQ38KdxqvVNTRz0M//TWPLzueKN8EqKcjj7kum7tT1agZ3XzKV4k6Mketp9MqIiIh0A8vW1/HoW1U89lY1qz/ZFvO4ovxcLjp0RFyD9ROVYE2p6M/cGccyZ1E1gZoQFeVFTJswRAnaLpiL1R6aRSorK938+fPTHYaIiEhKbQw28uTCah55s4qFVZvJMfjsmAGcNGEw1z/xLqHG1p0eU1yQu31WZryCDc1KsBLIzBY45yp3dZxeYRERkQzTUemMhuYWnl+ynoffXM2LS9fT3OoYO7gPPzxpLKdMHMKefQoBGFlenLAuxuKCPJXJSAO1pImIiGSQaKUzwDHjxHEsXrOFOQur2VLfzJ6lBZx2wFBOP2AoYwf3iXoutYBlpnhb0pSkiYiIZIi2gfrRZl4C9O6Vyxf2G8gZk4dx2N7l5OZorctspO5OERGRLNNR6Yz8XOP7J+zDhYeOTG1QkjY5qb6gmR1vZkvNbJmZXRtl/wgze87MFpnZi2Y2LNUxioiIpEOgNhizdEZji2PN5tjFZaX7SWmSZma5wB+BE4BxwHlmNi7isF8Df3HOTQBuAH6RyhhFRETSpaKsmIK86B/Nqs7f86S6JW0qsMw597FzrhGYDZwaccw44Lnw9y9E2S8iItItHTyqjMbmnctmgKrz90SpTtKGAqva3a8Kb2tvIXBm+PvTgVIzK4s8kZldbmbzzWz+hg0bkhKsiIhIqrS0OmY8+g698nLo3SsnPKvTt6AVF+SqOn8PlOqfdrRpKJFDJK8B/mBmFwMvA6uB5p0e5NwdwB3gZ3cmNkwREZHU+sPzy3j1o1puPnMCJ00YrNIZkvIkrQrYq939YUB1+wOcc9XAGQBmVgKc6ZzbnLIIRUREUuy1j2q59bkPOP2AoZxVOQwzU/FYSXl35zxgjJmNNLN84FzgifYHmFm5mbXF9X1gZopjFBERSZmaugb+3+y3qCgv5qenjcdMtc/ES2mS5pxrBr4BPAssAR50zr1nZjeY2Snhw44ElprZB8BA4GepjFFERCRVWlsdVz3wNpu3NfHHL05Wl6Z8SsrfDc65p4GnI7b9uN33DwEPpTouERGRVLv9pY/4z4c1/Pz0/WMu7SQ9V8qL2YqIiAjMXb6R3/xzKSdPHMJ5U/fa9QOkx1GSJiIikmIbg438z/1vMbx/ET8/XePQJDp1fouIiKRQa6vj6gffZmOwkUe+diilhb3SHZJkKLWkiYiIpNAd//mYF5du4EfTxjJ+aN90hyMZTEmaiIhIiixYsZFfPbuUE/cfxPkHj0h3OJLhlKSJiIikwKZgI9/821sM7debm86coHFosksakyYiIpJkzjmu+ftCauoaefirh9JH49AkDmpJExERSbK7/rOc595fz4wT92X/YRqHJvFRkiYiIpJEb67cxC+feZ/j9xvERYdWpDscySJK0kRERJJkc6iJb/7tLQb1LeSX0zUOTTpHY9JERESSwDnHNQ8tZP3Wev5+5aH07a1xaNI5StJEREQSpK6hmTkLqwnUBqnatI1/LV7HD08ay6S9+qU7NMlCStJERKTHa59cVZQVM23iEEoKOvcROS+wkYtnzcU5CDW2AJCbAxM0UUC6yJxz6Y5ht1VWVrr58+enOwwREclCkclVUX4uZnD3JVOZUtE/rnPUNTRz0M//TbChZad9xQW5zJ1xLMWdTPqk+zKzBc65yl0dp3eMiIj0WHUNzVw8a+6nkqu2VrDz73qdH03bj1bn2NbYQqixhfqmFrY1tbCt0d/WN/ntVZtC2x8XyTmYs6iac6YMT8lzku5DSZqIiPRYcxZW09oavUepodnxw8fe/dS2grwcivJz6d0rl8Lwbe9eubQ6n4xFE2psIVATSnTo0gMoSRMRkR5p/ZZ67p+7km1NrTGPOf+g4Vz1+c/QOz+XwrxccnKil9CYPXclN8xZHLU1rSg/l4ryooTFLT2H6qSJiEiPsmpjiB8+9g6H3/wCi6o2kxsj8SrKz2X/YX0pKymgKD8vZoIGMG3iEGKVQDODaROGJCJ06WHUkiYiIimXiNmUnbVs/VZue+EjHl9YTa4ZZx44jAsOGcFZf3o16oD/ziRXJQV53H3J1JgTEDRpQLpCsztFRCSlEjGbsjPeqdrMH19YxrOL11KYl8sXDxrOZZ8dxaC+hQmPJ9jQzJxF1QRqQlSUFzFtwhAlaLKTeGd3KkkTEZGUSXSpio5a5N74uJY/vvgRL3+wgdLCPC4+tIJLDhtJ/+L8nc6j5EpSSSU4REQk48xZWB1zFmR9Ywtfv+9NvjB+EMP7FzG8fxGD+xaSlxt9+HS0FrAb5rzHt479DP9avI55gU2Ul+TzveP35fyDh1NaGHtZpuKCPJXIkIyjJE1ERFImUBuMWU+sxcHLH27gxQ82bN+Wl2MM26M3e/UvYkRZESP6F7NX/yIGlBZw0cw3CDXumJnZdt6fP/0+g/oW8pNT9uPsyr3onZ+b3CclkiRK0kREJGUqyoopyMuhoXnnshdF+bn8aNpYPjtmACs3hlhZG2LlxhArNoZYtTHEkwvXsHlb0y6vkZ9rfPPo0XzpoIokPAOR1FGSJiIiKVNRXhw1QQM/m/KUiUMpLshj2B5FHDp652M2h5pYuTHErc99wL+XrI96nsYWR9XG+kSGLZIWqpMmIiIpsbwmyNfve5MBJQX0zveV+8G3oBUX5MZVqqJvUS/2H9aXY8cO3P74SCoeK92FWtJERCTp1m+t58KZb9DqHA9eeQiD+hTu1mzKaROHcONTi6PuU/FY6S6UpImISFJtrW/i4pnzqNnayN8uO4jRA0oAdms2pYrHSk+gd7GIiCRNQ3MLV9y7gA/WbeXOiyo5YPgeCTv3lIr+zJ1xrOqbSbeld7KIiCRFa6vj6gcX8upHtfzmrIkctc+eCb+G6ptJd6aJAyIiknDOOW6Ys5inFq3h2hP25cwDh6U7JJGso5Y0EZEeIpWLmt/24kfc/WqALx8+kis+Nyop1xDp7pQi7iC2AAAgAElEQVSkiYj0ANGWULrxqcVJWdT8wfmr+NWzSzl10hB+cOJYzCyh5xfpKdTdKSLSzdU1NHPxrLkEG1q2L50Uamwh2NAS3t6csGs9t2Qd33/kHT47ppxfTZ9ITo4SNJGuUpImItLNdbSouXMwZ1F1Qq6zYMUmvv63Nxk3uA+3n38g+Xn6iOmyhq2w4B7413X+tmFruiOSNFB3p4hIN9fRouahxhYCNaHdvsay9Vv58j3zGNinkFmXTEnaWLceYcVrcN90cK3QFIJeRfDsDPjSQzDikHRHJymkf3NERLq5koI8Oup0fOzt1dz24jJWbexasrZ2cz0X/t9c8nKMv1w6lfKSgq4FKr7F7L7p0FjnEzTwt411fntDXXrjk5TSvzoiIt3Yi0vX8+eXPyZGbyf5ucbAPoXc/MxSbn5mKZP26sfJE4dw0v6DGdS3MOpj2s8S3bO0kPvnrmTztiYeuOIQRpQVJ+/J9ATvPuJb0KJxrfDeIzD5wtTGJGmjJE1EpBtyzvHnlz/ml8+8zz4DS/nG0XvzvYcXRV1CaUpFf1ZtDPHUO2t4cmE1N85ZzE+fWszUiv6cPHEIJ4wfRFm4dSxylmiOQauDH00by/ihfdP8rLuBjR/vaEGL1BSCwKsw6XzIUUdYT2Au1mjSLFJZWenmz5+f7jBERDLCtsYWvvfwIp5YWM1JEwbzq+kTKMrPI9jQHNcSSh9tqOPJhdU8ubCajzYEyc0xDtu7nOPGDeTnTy8m1LhzS09xQS5zZxyrJZl2139vged+Ers1DaB4AIw5zn+NPgoKO0iOG7b61rmNH0P/UTD+DCgoTXzc0ilmtsA5V7nL45SkiYh0H1WbQlz+lwUsWbuFa47bh68dObrLdcqccyxZs5UnF/mErWrTtpjHFuXnct3J47RE0+5Y8So8cD6EaqPv71UMx98Ey1+CZf+G+k8gJw+GHwKf+QKM+QKUj4G2n3e0CQiWowkIGUBJmohID/P6x7V87b43aWpu5dbzJnH0vgMTdm7nHFc/+DaPvhW7XMdXjxjN907YN2HX7DGcg/n/B//4HuxRAZ/7Hjx1VcfJVUszVM2DD56BD/8J6xf77XtUwGeOh4rPwiOXQ1Nw5+vll8C3l0JBSaqeoUSIN0lTu7SISJZzznHv6yu44cnFjCgr4o4LKxk9ILEfwGbGQSPLePa9dVHLeRTl51JRXpTQa/YIzQ3w9DXw5l989+UZd0LvfrDviX6SQO3HUDYK9jvj00lVbp5P2EYcAp//CXyy0idrHzwLC+6GN/4U+5qagOBlQVewkjQRkSzW0NzCjx57lwfnV3Hs2D357TmT6FPYKynXmjZxCDc+tTjqPjOYNmFIUq7bbW1dCw9cAFVz4bPXwFEzICfX7yso6VwS1W84TPmK/2oMwaNXwpLHox/bFPLJX0+WJbXoND1ERCRLrdtSz7l3vM6D86v45tF7c8cFlUlL0MDXW7v7kqkUF+RSlO+TiaL8XIoLcsPb9X9/3FbNgz8fAeveg7PugWN+tCNB2135RbD3MT7xiCYnD/oMSsy1slEW1aLTb5SISBZ6a+Umrrh3AXUNzdz+pcmcsP/glFx3SkV/5s44Nq5ZohLDm/fCU1dD6WC44BEYuF/irzH+DN8yFE1rMzz/MwhtgoO/6rtXs0lXuymbtsHG5TDvLt/NHE2GdQXrt0pEJMO1Lx5bUVZMU6vjxicXM7BvAX/58qHsO6hPSuMpLsjTLM6uaGmCZ74P8+6EUUfC9FlQ1D851yoo9V130WZ3Hv9L+OAf8NJN8PrtPlHLlmRtV92UrS2weRXULoOaZf62dhnUfuS3xyzrHJZhXcGa3SkiksEii8fm5RjNrY79h/blL5dOZY/i/HSHKPGo2wB/vwhWvAKHfAOO/Ykf/J9sDXWxJyCsWQQv/RLenwMFfTM/WWvYCr/Z13dLRrJcKBsNmwLQ0rhje0EfKNu73ddoWP8+vPYHaI5SUqZXEZzwy6S3pGl2p4hIlqtraObiWXMJNuyYTdnc6v+x/mjDVvLzsnhYcRbMrEuY6rdg9vkQqvGzNyecnbprdzQBYfAEOPe+HclaW8vaIV+Dg678dLKWCT+vXS2ZlVfok8ztCdkYKC7fUTeuTcNWeOP26OexHJ/IZgglaSIiGWrOwmpid3YYcxZVZ2e3Y5bMrOuSyGQG4B/fhaJyuPQZGHJAeuOLJjJZe/EX8NptO5K19Usy4+dVNT/2klk4GH0MfP76XZ+no67gLz2UUfXjlKSJiGSoQG0wak0y8F2fgZpYH1gZrP3MujZtH7z3Tc/uIquRyWdOnh+kP3B/uOBRKBmQ7gg7FjVZ+yM013+6CzHVP6/QRnjpZnj7r7GP6VXku3PjNeIQH3tHtegyQBa3lYuIdG977VFETowVnbK2eOyuuqzeeyS18SRKtLIOrc3+duPH0Kt3+mLrrLZk7Yr/+BUM2ido7SX759XSBG/8GX5/AMz9M0w8L3ZZka50U7Z1BX/+en+bYQkaKEkTEclIzjnmBjbSGqO7M2uLx278OHaXVVMINnyQ2ngS5d1HdiRlO3HZmXwOngCjj469P1kzIZ3zKyfcfqjvKh48Ea78L5x2G5z/iF/Wqi1Z61Xk72dYN2WiqLtTRCQD/fZfH/D429WcXTmMp95Zs312Z1F+LmZkb/HY/qP8B2usRG3enVC3DsadAnsfm/ktUE31sPgx3zXYXB/jmMwq69Apu/p5ffw8LD7AL+7eq3D3r7duMfzzB/DR837w/3mz/VqkbYP/s6SbMlFUgkNEJMPc98YKfvDou5w7ZS9+ccb+hBpbuk/x2IatcPNoaIlSTDSvEMadBh8+C9s2Qa9iGPN5n7CNOS76bMJ0zTqsWQYLZsHb9/lYiwfAtk+gtWnnY1NU1iEpOip7kdPLzwANbvClLsaeDPtPh5FHdH71hGANvPAzv+5oQSkcca1f4iqve5aYibcEh5I0EZEM8q/F67ji3vkc8ZkB3HlhJXm5CRyVkgllFJY8CQ9eCBjk5vtaVe1n1o04xI9FCvwXljwBS+ZAcD3kFviljsaeAvuc4JODaLNE258n0Vqa4P2nYP5MWP6Snxiw7zSovNTP2vzt2OjJTH5J95oQ0f51HjYFAv+Bdx7yP6+GLVC8J4w/E/Y/C4ZO3tEKFu39l5vvx529/CtoDMKUL8OR309ekd8MoSRNRCTLvLlyE1+883X2GVjK/ZcfTFF+AlvLUp3QRBN4Be49PTww/X5f9X5XXVatLbDqDVj8hE8Ctqz2LTgVh/nnFK1FritJUUcJ7CcrYcE98Na9viu273A48CI44AIoHbjjHJnwGidLR0Vx2zRtgw//Ce/83Y8pa2mEPUb6ZK1sDDx11Y7XJq834KD3HrB1jW8pPe6nMGCftDy9VFOSJiKSRZbXBDnz9lcpLczj4a8eSnlJQeJO3lGXVapaeda9BzNP8EnNpc92raWktRWq34TFj/tuxlBt9OPyCuGY63yNr5w4WiKjJVcYHPFdWPGqTzzM/Lirykt9i16s7rx4kpmeYNsnfiWDRQ/C8peJuRyT5cDZ98LYaSkNL92UpImIZIkNWxs48/ZXqWto5pGvHkpFeXFiL7DgHvjH99K3DM4nq+D/jgMcfPlf0G+v3T/nP38Mr97a8TGW6yvOF+/pb0v29GPH2r5K9oT8YvjrmbEHxhfv6VvNJl+UmLh7old+D8/dEH28Xl5vOPHm7Byvtxu0LJSISBYINjTz5XvmsWFrA/dffnDiEzSA9YujJ2iQ/JmHoY3w1zP8eKNL/5G4RKdsdOxZh7n5fuzaHhV+PFuwBurW+67M4IYOqtZHOc9RM6DyksTE3FOFaqMnaODfl9k68zUFlKSJiKRJU0srX//bm7y7ejN3XljJpL2SsLB11QLf5RRLbn7nKrV3RmMI/nY2bFrhK+4P3C9x5x5/hl+aKJrcfDj51tjdjI1Bn7QFa+CVW+H9J6Mf19LoY5fd01EZj86uFNDDqJitiEgaOOf4waPv8OLSDfzs9P05ZuzAXT+ocxeAV/8AM4/ztcbyYtQba2n0Y7KaYtT46qqWZnjoEli9AM68yw/0T6S29Re7Utg0vxj6j4S9pvgSH7Gq2CuBSIzxZ/ixZ9Fk2ILmmUYtaSIiaXDLvz/kwflV/M/Re3Pe1AQvkh7aCI99FT54xpeIOPUPsP796IPjx54MC/8G696Bs/+yY1Hw3eEczPl//von/dbXOUuGRBQ27ahFTglEYmTRguaZRhMHRERSbPbclVz7yDucdeAwbp4+AbMYC3R2xYrX4OEv+7FXx/0Mpl7Wrk5VjJmHHzwLj1zuk6vTbtv9mXbP3Qj/+TV87rtw9A92/zklW3cunZFJNPN1O83uFBHJQM+/v47L/rKAw/cu566LKumVqGK1ra3w39/CCz+HPUbA9FkwZFL8j9+0Av5+EVS/BYd+05ewyO3V+Tjm3glPX+NnQ558644EMdMpgZAU0uxOEZEMs3DVJ3z9vrcYN7gPt31pcuIStK3r4NHL4eMXfaX3abdAYZ/OnWOPEb5+2bMz4NX/har5MH0m9OnEIu7vPQZPfwf2OdF3c2ZLggY+IethZSAk8ylJExFJkrqGZuYsrCZQG6SkII+ZrwQoL81n5sVTErf25kcv+K7Khq1w8u99otHV5CivAE76DQw/BJ74H/jz5/yg/1FH7vqxy/8Dj1wGe02FM/8PcvXxIrK79FskIpIE8wIbuXjWXJyDUGMLhq+5ft3JkxhQmoDVBFqa4aWb4OVf+6V0LnwcBo7b/fOCXyR70P7wwAV+GacjZ8Bnvx27ev/ad2H2F/2kg/NmQ36M2ZIi0ilK0kREEqyuoZmLZ80l2NCyfVvb6N8Zj77DsWMHdq4lLXJdyeEHw5PfgpWvwgHnwwk3+7ISiTRgH7jseZhzFbzwU79+5hl3+OWc2sdT2Bfe+JMvfXH+w91+YWyRVFKSJiKSYHMWVhNrTpZzMGdRNedMibPsRuTMw9wCv6h4XiGcfgdMPCdxgUcqKPGJ2fCD4ZlrfffnZ78N//zhjnja2ghP+xP0HZa8WER6IBWzFRFJsEBtkFBjS9R9ocYWAjVxLkvUsNUnaI11O6q1tzT4W8uBfU9KQLS7YAZTvuwnFQDM+dan42lrI3z6Gj9DUkQSRkmaiEiCVZQVU5SfG3VfUX4uFeVxjtl69xHfYhXLe490IbouGjoZDv5a7MrxrjW18Yj0AErSREQSbNrEITEnWJrBtAlxlrVY8VrsxcCTvTB6NHXrYieN6YhHpJtLeZJmZseb2VIzW2Zm10bZP9zMXjCzt8xskZmdmOoYRUR2R0lBHndfMpWCvB2ZWlF+LsUFudx9ydRdTxrY8AE8eCEsuj/2MelYV7JtoexMiUekm4uZpJnZXma2oKMkycxOCB8T12+mmeUCfwROAMYB55lZ5JzxHwIPOucOAM4Fbovn3CIimWRKRX+u+NxoAL5y+EiuO3kcc2ccy5SKDmY/bq6Cx78Btx0Ey56Dw6+GXjFmbaZjXUktlC2SUh39O3cNUOecezrWAc65f5jZd4HvAF+N43pTgWXOuY8BzGw2cCqwuP1pgbZS2X2B6jjOKyKScao31zOwTwE/nLaL+mXBWr+k09w7AQcHXelnURaXw5jjMmdhai2ULZJSHSVppwDXx3GOmcBP4rzeUGBVu/tVwEERx1wP/NPMvgkUA8dGO5GZXQ5cDjB8eJxT2UVEUihQE2REWQf1yxq2wmu3+WWYmoIw8Ytw5LXQb68dx4w4BL69NHPWlcy0eES6sY6StCHAR3GcYzk++YpHtKG0kdWEzgPuds79xswOAe41s/HOfXq0qnPuDuAO8Ausx3l9EZGUCdTUccyeW+Ff1/nxXOPP8K1RzQ0wf6ZfLSBUA2NPhqN/5AvIRpNp60pmWjwi3VRHSdoWYEAc5ygPHxuPKqDdv4gMY+fuzC8DxwM4514zs8LwNdbHeQ0RkbTb+uGr1ASbqFg9B9Y84rsFn/k+HHQFvPN32LwKRh4Bx1wHww5Md7gikoE6mt35OhBPKetzw8fGYx4wxsxGmll++LFPRByzEjgGwMzGAoXAhjjPLyKSfg1bWTH72wBUuPAIj6aQ79L872+hd3+44DG46AklaCISU0ctab8BnjOzJcBPnXOfKp9tZjn4mZjTCSdVu+KcazazbwDPArnATOfce2Z2AzDfOfcE8G3gTjO7Ct8VerFzsRZYERHJQO8+wvIW3xGxr6389L7cfF/Bf/RRaQhMRLJJzCTNOfdiOFH6DXCFmT2Hb+VywHB8YjYAuMo591K8FwzPFn06YtuP232/GDisM09CRCSjbPyYFc39ABhhESM1Whph4/I0BCUi2abDiorOud+b2X+AbwKfw7eaAawG/gH8r3NuYXJDFBHJMv1HsZwmhrCBHIvoCFDRVxGJ0y7KXoNz7i3g0hTEIiLSPYw/g8AjM6m0pTvvU9FXEYmT1u4UEUm0glJW5I/hgLzAjm29iiC/REVfRSRuMVvSzGw5O9cwa9OML4nxMr7Lc00SYhMRyUpb65uo2eb4zNgxvpLk1Ctg0HgVfRWRTumou/NhYidpucBg4CvAV8zscOfcB4kOTkQkGwVqQgDsZeuhsC+c8EuwaLW8RURi62h25zW7erCZFQMvAT8DzkpgXCIiWStQGwSgrGEllI1RgiYiXbJbY9Kcc0Hgt4AK/oiIhAVqfJLWe8tyKNs7zdGISLZKxMSBdUBpAs4jItItLK8NUlEKOVurlaSJSJclIkk7EF/kVkREgBW1Iab2+8TfKRud3mBEJGt1OUkzs0Izmw78ALgvcSGJiGS3QE2Qib1r/B21pIlIF3VUgmMDHc/u7BO+fQD4aeJDExHJPlvqm6gNNjImd63foJY0Eemijkpw/JGO66RtAP7jnFuS8KhERLLUinD5jaEtq6HPUMgvTnNEIpKtOirBcX28JzGzXs65poREJCKSxZaHy2/0r1+pVjQR2S27MybNzOxoM7sTWJvAmEREspYvv+Eo3PKxxqOJyG7Z5QLrkczsIOA84GxgILARmJ3guEREslKgNsi+fZqw+s1K0kRkt8SVpJnZeHxidi5QATQC+cDVwB+dc83JClBEJJsEaoIcVLoRavGrDYiIdFHM7k4zG2VmM8zsHWAhcA2wBLgQGAMY8JYSNBGRHQK1Ifbvvd7f0Zg0EdkNHbWkLcPP7nwDuAJ42Dm3CcDM+qYgNhGRrLJ5WxMbg42MzlkHOXnQb0S6QxKRLNbRxIEV+Nay8cCRwKFm1ukxbCIiPcWK8MzOIS2rYY+RkKs/mSLSdTGTNOfcSOAw4B7gGOBJYF14NucxxK6hJiLSIy0PL6zeb9sKTRoQkd3WYQkO59xrzrlvAkOBLwCPA2cCD4UPuczMKpMboohIdgjUhMihlfzNAShXkiYiuyeuOmnOuVbn3L+cc5cCg4AzgL8DpwNvmJlWHRCRHm9FbZBJfeqwlga1pInIbut0MVvnXKNz7jHn3Ln4OmkX4icZiIj0aMtrg1SWbPR3lKSJyG7q8ooDAM65oHPuPufcyYkKSEQkWwVqgowv3ODvKEkTkd20W0maiIh4m0NNbAo1MSpnDeSXQMnAdIckIllOSZqISAIEwuU3BjVV+VY0szRHJCLZTkmaiEgCtCVpfUMqvyEiiaEkTUQkAQI1IQqskbytVUrSRCQhdjtJM7O9zez2RAQjIpKtArVBKks+wXBK0kQkITpM0swsz8wONrOzIovWmtkUM3sIv+j62ckMUkQk0y2vCXJgabj8hgrZikgCxEzSzGwE8DbwCvAAvmjtHDPrY2b3A68DnwN+DFSkIFYRkYy1ojbIuPx1/k7/0ekNRkS6hY5W//0F0B/4IrAQGBHetiD8/XXAb5xz25IdpIhIJmsrvzHS1vjSG4V90h2SiHQDHSVpnwVmOOceCN9/38zW4FvXvuOc+03SoxMRyQLLwzM792zUpAERSZyOxqQNAd6P2LY4fPvf5IQjIpJ9VoSTtNLgCihTV6eIJEZHSZoBLRHbXPi2MTnhiIhkn+U1QfpaHXn1tVA2Jt3hiEg30VF3J8DdZhaMsv1eMwu13+Ccm5q4sEREskegJsjU0k3+31d1d4pIgnSUpN0TY/t7yQhERCRbBWpDfKG4RkmaiCRUzCTNOXdJKgMREclWgdog+5SvB8uBPSrSHY6IdBO76u7EzA7E10FbCyxwztUnOygRkWzxSaiRT0JNjGAN9BsBefnpDklEuomYSZqZ7Qk8DkzFTyIAWGFmZznn5qciOBGRTBeo9cNzBzSuggGaNCAiidPR7M6bgFHAhcA44CSgDrgrBXGJiGSFQE0Qo5WSuoDGo4lIQnXU3Xk0vpjtfeH7bcVsF5hZuXOuJvnhiYhktuU1QQbZJnKat6lGmogkVEctaXux80zO9/Bdn0OTFpGISBZZURuksiS8sLpa0kQkgTpbzLY1jseJiPQYy2tDHFAc7lhQkiYiCaRitiIiuyFQE+QzZeuhVxGUDkl3OCLSjaiYrYhIF30SamTztiaGu9XQfzTkqJNBRBJHxWxFRLpoeY3vaChvWAUDD0hzNCLS3ejfPhGRLgrUBulFM72DVRqPJiIJpyRNRKSLAjUhhuesx1yLkjQRSTglaSIiXRSoDXJgcbj8RrlWGxCRxFKSJiLSRYGaIJOKNvg7/UelNxgR6XaUpImIdFGgNsSYvLVQVAZF/dMdjoh0M0rSRES6YFPQl98Y1lKt8WgikhS7Kma7nZlNB84AhgGFkftVzFZEepLltb78Rv/6VVBxXJqjEZHuKK4kzcyuB34MLAQWA41JjElEJOOtqA1SzDYK6tdrYXURSYp4W9K+DNzknJuRzGBERLLF8poQo3PW+Dvq7hSRJIh3TFop8FwyAxERySaBmiCT28pvKEkTkSSIN0mbDRyfzEBERLLJitogE3pvAAz6j0x3OCLSDcXb3fkc8EszKwf+BXwSeYBz7ulEBiYikqmccyyvCTJ6j7XQby/o1TvdIYlINxRvkvZA+LYCuCjKfgfkJiIgEZFMtynUxJb6ZoY0r4Y91dUpIskRb5KmtnwRkbBAbRBw7LFtBZQdnu5wRKSbiitJc86tSHYgIiLZIlATZACbyWsOatKAiCRN3MVsAcwsDxhO9GK2ixMVlIhIJgvUBBm1vfyGaqSJSHLEW8y2F/B7/Hi0ghiHaUyaiPQIgdoQk0tqfVnvsjHpDkdEuql4S3D8GJiGL2prwDeAS/CzPgPAyckITkQkEwVqg4wv2AC5BdB3WLrDEZFuKt4k7WzgeuDB8P25zrm/OOeOA/4LnJqE2EREMk5b+Y1Rtgb6j4IcdSKISHLEm6TtBXzgnGsB6oE92u27Dzgz0YGJiGSiTaEmttY3M7h5tcajiUhSxZukrQH6hb9fDnyu3T79lRKRHmN5TZBcWuhTX6WZnSKSVPHO7nwR+CzwJHAn8Gsz2xtoAM4B7k9KdCIiGSZQE2So1ZDT2gTlmjQgIskTb5L2A6AcwDl3i5kZMB3oDfwvcENywhMRySwraoOM3l5+Qy1pIpI88RazXQusbXf/d8DvkhWUiEimWl4bYnJRDTShJE1EkireMWkAmNk4M7vAzGaY2aDwtr3NrDQ54YmIZJZATZCx+euhsC8UlaU7HBHpxuJK0sysxMweBN4B7gJuBIaEd/8cuC7eC5rZ8Wa21MyWmdm1Ufb/zszeDn99YGafxHtuEZFkcs4RqA0y0tb6IrZm6Q5JRLqxeFvSfgscChwLlOIL2rZ5Gjg+npOYWS7wR+AEYBxwnpmNa3+Mc+4q59wk59wk/Hi3R+KMUUQkqTYGG9la38zAJs3sFJHkizdJOwP4nnPuBaAlYt8KYESc55kKLHPOfeycawRm03Eh3PPQzFERyRCB2iCFNFDSsFZJmogkXbxJWm+gNsa+UnZO3GIZCqxqd78qvG0nZjYCGAk8H+e5RUSSKlAT8l2doEK2IpJ08SZp84ALY+ybDrwa53miDeBwMY49F3govMrBzicyu9zM5pvZ/A0bNsR5eRGRrgvUBhmd05akqSVNRJIr3iTth8AZZvZv4Cv4xOpEM7sXOIv4Jw5U4ZeYajMMqI5x7Ll00NXpnLvDOVfpnKscMGBAnJcXEem65TVBJhbV+DtqSRORJIsrSXPO/Rc4BigA/oBvEfsJMAo41jk3L87rzQPGmNlIM8vHJ2JPRB5kZvvg1wd9Lc7ziogk3YrakC+/0Wco5BenOxwR6ebiXXEA59wrwGfNrDc+gfrEORfqzMWcc81m9g3gWSAXmOmce8/MbgDmO+faErbzgNnOuVhdoSIiKeWcI1ATZERxtVrRRCQl4k7S2jjntgHbunpB59zT+LId7bf9OOL+9V09v4hIMtQGG9na0MSevVZB2UHpDkdEeoCYSZqZ/TjWviicc+7GBMQjIpKRVtQG2YOtFDRv0aQBEUmJjlrSrse3mAWJPiuzPYdfhUBEpFta/qnyG2PSG4yI9AgdJWkfA8OBBfiis48657akJCoRkQwTqAmyd65qpIlI6sSc3emc2xu/FNR7+FaytWb2iJmdFZ48ICLSYwRqg0zovQFy8qBfvIusiIh0XYclOJxz851z1zjnhuPX51yLL8Gx3szuM7PPpSJIEZF0C9QG2SdvHewxEnI7PedKRKTT4i1mi3PuZefc1/DFaP8EnAN8K1mBiYhkCl9+I8RwV61JAyKSMnH/O2hmh+GLz07Hr9f5EHB7kuISEckYtcFGgg2NlOVWQflJ6Q5HRHqIDpM0M5uMT8zOAQYCzwBXAU90tpCtiEi2CtQEGUItea2NakkTkZTpqE7aUmAk8Dx+bc5HNLtTRHqi5TVBRmphdRFJsY5a0sYA9cCBwGTgZrPY5dKcc3smNjQRkcywojbE3jlr/B0laSKSIh0laT9JWRQiIhlseW2Qows3QE4JlAxMdzgi0kPETNKcc0rSRETwY9LG5K2DPfaGDt1RA8wAACAASURBVHoUREQSKe4SHCIiPZFzjhW1IfZqVfkNEUktVWQUEelATV0jjQ3b6GdrlKSJSEqpJU1EpAOB2iAjbB2GU5ImIimlJE1EpAOBmiCjLDyzs1xJmoikjpI0EZEOBGqDjG6rkdZ/dHqDEZEeRUmaiEgHAjUh9itc70tvFPZJdzgi0oMoSRMR6UCgNsjeues0Hk1EUk5JmohIDM45AjVBhrVUQZm6OkUktVSCQ0Qkhg11DeQ2bqY45xMoG5PucESkh1FLmohIDCtqQ4w0LawuIumhljSRLFPX0MychdUEaoNUlBUzbeIQSgr0q5wMy2uCStJEJG30l10ki8wLbOTiWXNxDkKNLRTl53LjU4u5+5KpTKnon+7wup1ATZDRuWtxloPtUZHucESkh1F3p0iWqGto5uJZcwk2tBBqbAF8ohZsaAlvb+70+WbPXclN/1jC7Lkrqevk43uCFbUhxhesw/qNgLz8dIcjIj2MWtJEssQTb6+mucVF3dfY3MqNcxZz6qShDOpbyKA+hfTOz415LrXIxWd5TZBRthbKNWlARFJPSZpIBtsYbOTlDzbw/Pvrefa9tTQ0t0Y9rqnFMXveKmbPW7V9W2lhHoP6FDIw/DWobwED+xTSt3cvvvfwIuqbdpyrrWXu4llzmTvjWIo1xs2X36itY3Deaij7QrrDEZEeSH+JRVIkngH/ra2OxWu28ML763l+6XreXvUJzkF5ST77DenDu6s30xilNa13rxy+duRoDhzRn7Vb6lm7pZ71WxpYu9l///FHNazf2kBza/SWuDbNLY6/L1jFxYeOTNhzSuV5EmlDXQOljTXk59SrRpqIpIU51/Ef7WxQWVnp5s+fn+4wRGKK1r1oBndfMpV9BpXyyoc1PP/+el78YMP/b+/O46Oq7/2Pvz4zmSQkgbCEfV8EREG0gLhcENwVl1orbtelm2t/tj9/rVXr1epPa+2t7W1V1LaKOyriUlRQ61KtoODCIoICsiMhEEI2JrN87x9nAgEmyYRMMpPk/Xw88piZc86c+c6XQ3jz3Q5bS4OYwag+HZk0rCuThnVjZO98KkIRjrzrLcqDkf3On5vlr7cFLBp1FJUH+d3ry3nh0411lrd/lxwO7ZXPiF4dOLR3Pof06kBBXlbC36khXabJOk8ylQXD/OWfX7P4g3/wTOadVF4wi3bDjk9JWUSk9TGzT5xzY+o9TiFNpGmVBcO1hiufgc+McNTRPjuDCUO7MnlYNyYO67pfKILkBJoZH6/j9tnLdndx1pSV4eO4YV3xmbF0Uwnrt1fu3tejQzaH9u7AiF75DOmWxw0vLKKyav/u10QCY7W66qYh56l5vsa2yFXXcSgc5fu8yZ2BRzje3c/dl5+m8XoikhSJhjR1d4o0sZc+3VDrgH8HHHtQAVcfN4Qj+nUkw1/3hOuxAzrz8U0nMHvxJtYUVTCgIIcpo3o1KMhMOawXd7y6LO6+DL9x73mjd5+vpCLEF5tLWLZpJ0s3lrB0007eXl5IXb2mwVCUK5/8hKHd2xN1Due88V0Odr/23u/4ekvZXmPjagpFHNPeXclF4/tTkJdFoJ66ScZkiJozaAEGZmymwmWxOpiv8Xoi0uz020akHg1pnQmGIyzfXMqSjSUs3VjCko0lfLl5Z62hxjk4uEcHxg1MvIUml0qm+t6BjNXgGwScA7RP+P15WRlMv3xcrS1yNUNIfk6AowcXcPTggt3bKqrC3DRrCS99vinu+cNRx7xV2/hs3Q4MMAMzw1fjEbzHsmCYSC2VUxWOct87q7jvnVWYQUFeFt07ZNG9fTbdOmR7z2OP7bMDXPrIR1RU7T8Z4tJHPuLFq48hGI5SXBFiR0UVxeVVe55XhCiuqGJHRYgNxRV7teoNtG9Z43rg8OEczF68ialj+yVc1yIijaGQJlKHulpnRvbOZ8W3pSzeWMLSDV4g+2pL6e7B+fntAozsnc9/HNSVeau34Q9XcIZvHgN93j/8syPjcZl5DCjISbxAa+fBU+eCi0KoAgI5MPcmuGgm9D8q4dM0pkUuJzOD8YO68MayLXG7THMy/dx6xoiEwkxdXa/ZGT7OH9eXod07xCZC7GLLzl1sLtnFog07KCqrSui7VlRFOflP78fd1yE7g065mXTMyaRLXia7QhGKK0K79w+0zSxz/WPnibCmqCKhzxQRSQaFNJFa7Nv1BXtaZ6Y+NG/3WDKAjjleIPvxsEGM7J3PyN759OnUDjOjLBjmqrvuY5r/txiOXAtS7rK4JeMJruJGpow6IbECBUu9gFZVtmdbKBYanjoXrl8BWXkJf7/GtMjV1WVqBlNG9Wr0efx+4xcnD681OFaFoxSVBfl25y4efHcVbyzbUuvnTBrWlQuP7E+nnAAdczLplBMgv11gv+7lmqExgzD9rJBXo+MBL3w2KFCLiDSSQppILZ79eB2hWtYlA28s2Xlj+u4VyOLJo5LpmffgD+3avS3XggBMD9yDnx8B+4Qr5yC8C0KVUFXuPS5+FiIh4opGvP1jf5jYl2tki1xDukyb6jyZGT56dWxHr47tmDy8Gx+sLKq1Ze+UQ3tw4oju9ZanOjTmUsll/jlkWJRuFJNLJVhewuFTRCQZNLtTpIbVW8t468stvLWskAVrtlPX346rJg7mhlOH13/STx6D12+AcOX++8wH+X0gkAuhWBirqoi1kB3A382MdpCdD+06QnbHGs/z97wOZMPcm70QuK/MvAa1yJUHw42axLD7PDuLWfbmdCJFq/AXDGbEiZeR26FTwu+vniVKsIwp/vkMsD1dymTlNWjA/5cfzaXva5cQIEyWhdnlAoTxs/60xzn4SC1qKyKNp9md0uYlMuA/EnV8uq6Yt5Zt4c0vt7B6azkAI3p2YNLwbvx7ZVHcVf4T6voKlsGqt+GjB+MHNPBasszn3XYoMxcC7bxWrUCO97zmtnXz4dPHIRLc/zy+AAw9GToPhModsKsEdu2Asm9h63Lv+a6d1Bv8IiH47AkYf1Xdx8XkZmU0fiD92nnkPnUuY6tb9YpyYMV/N2icXV5WBjNPM/q+dg0+HDkWpCLWpbz+tMcTD47BUg7+5w/A9gTYbAsBIW/76IZ1KYuINIZa0qRVqms9sRE9O/D+11t5c1khby/fQnFFiIDfGD+oCyeO6M7k4d3o0ynnwNbw2rEevpoDK16HNe9DpMoLWeEqcPufh0AOnPo7OOKS+r9UsBT+MHzvMWnVEmkBi0YhuBPeug0+ebTuzyoYBv2OhL5HQt/x3or78bpzg6WwdBZsXw2dB8Gh50BW4jNNG/2dEjlPIBeuXeCVv6rcO6aqItaNXB7bFvtZN88L1tE4N5tvyJ+ViEgd1JImbVZdA/7Pf9gb8B+KOPLbBZg8vBsnHNydCUMLaJ8d2Os81eOlrn70PU5xH9IruplNvp7MsaN54PKJXkCLRmHTp14o+2oObFnqvbnzYBj3Exh6CnQ/BP40Mn6AMB8cck5iXyyrvde6tO9YMvN52+sLMz6f1/XZ63Bv/FoozkxFfxYMPckLlcte8VruAHIKvMDWLxbaeo2GjZ8e2Li2cBDKi6B8Kyx6xguy8YQqYfrp0KG3F3CjYW/sXfXj7m1h73xV5bWcpxz+OKLuutnNqLW1MVQB21YneB4RkcZTSJNWZ/aiTdTWQByNwtEHdeGaSUMY079T/YvH2go+zrqWcCRCIFJJyN+O231PYeuuh0Wr4as3oLwQzA/9xsOJd8CwU73uy5oaE65q6n+U17r0xSwvMHQZ5IW8hpzj0HO8MBWPPwBnP+idLxqFoq9g/XxY95H3uOLV2HFZEA1536dadeh74myYfIvXulW+1auf6lBWvtXrik2Ei8COdV4I8/m9OvZlxH784Mvcsy1YSp1duQMmwshzvFa1zH1+Ajleq11mLiyaAXNvjB9gAzlefYuINBN1d0qrEok6/s8zn/Lqkm9rPSbhAf91daEBZHWAISd4oWzICZBTz4K0wbLGhatkije7szo01tUKVlYI6z+CBY/AN+/uHdL2Y16d5HaN/RRAbrcaz7vChgWxMXtxJjE0pHvxk8dgzq9qD1fN1aUsIpIAdXdKm7K5pJLnFmzguYXr2bijlkH6NHCtq0XP1t4V58+CE2+HMZcnXsisvPQZz3SgLXJ53eDgM2DDQlj9du3Hjf0JnPJb8NfzK2bQRFjwt/j7GtIVXFfrYHN2KYuIJJFCmrRY4UiUt5cXMmPBet5d4d1P8tghBfz8xKH818tL9rpFULV6F1p1zhtr9flT8NmTtYe0SBCK1ybpm6RIY0Jj50FeeKmt5arnyPoDGiQvFCUzXCWjS1lEJAkU0iTt1Ld0xrptFTy7cB3PL9xAYWmQbu2zuPq4IUwd25e+nb1Wsv5dchq2QOrOzd5g+s+fhqIVkJENPUbCt0viL3nR1scnJavlCpIXipIZrtKp1VNE2iyNSZO0UtvSGX+9ZAzby6uY8fF6PlhZhM9g0rBunD+uH5OGdY07AaDeBVJDu2DFa14wW/VPrwWm73gYfSEccrYXNjQ+qXYHOq5NRKSNS3RMmkKapI261iWr1rtjO6aO7cv3x/ShZ3672k9WW4C48HnIyPK6M5fO9GYadugNh13ghbMugxM7j4KIJ50mQ4iItBCaOCAtzuxFm4hG4/+nwW9w2TEDuPm0Efh88e+RuVtdNyKffhrgvO7Mg8/0gtnACd6SDvFofFLd1C0oItJkFNIk5cqCYeat2sbj89dSGYq/pEPEQabfX39AA28F/NqWhjCDUVPh1Hu8+1gmQkFERERSQCFNkiaRe2UCOOf4cnMp7321lfe+KuSTtcWEIo5Mvw+/eYFsXwkvnVH6LXzxYvxZh+CFt7yeiQc0ERGRFFFIk6SIN+D/jleXMf3ycYwd0Jni8ireX1nEeyu28q+vt7K11JsxObxHe35w7EAmDu3KwT06cOw9b8cdk1bn0hnbVsHy2fDlbG9xVBy13t6nrc/KFBGRFkMhTRqtrntlXvjX+Qzv0YGlm0pwDjrmBDh2SAETh3ZlwtCudO+Qvde56r1XJnhrmX272Atly2dD4TJve8/DYNLNMHgSPHamd8/GfTV0eQgREZEUUUiTRqvrXpmhiKOksorrjj+IiUO7MqpPR/x1jCuLe69M/9OYexbWmBfKls/27uloPuh3FJxyNww/HTr223Oii1/QqvEiItKiKaRJo63ZVr675Sye00f24mcnDK3/RLFZmVZVRiC2KRCphAjw2BTAgT8TBk2CCb+AYad594CMR7MyRUSkhVNIk0bzmWHm9UIOsM2c6fuQLXRmdmQ8LjMvsQH/wTJ4/97ab8NkPm+G5Ul3eLcASoRmZYqISAumkCYHrLIqwr1vruDvH3xDwIX4U8Z9TPQvJteClLssbsl4gqu4kSmjTtjzpkgYtq2Ewi+g8EvYssx7Xrym7g9zEcjumHhAExERaeEU0uSAzF+9jV+9sJg12yq4cExPfrn8XDqGt+3en2ve7M3p/rvwz6/ygtmWZd59Matby8wPXYZAr8Nh9EVQugU+fxLCu/b/QM3KFBGRNkYhTRqkLBjmd68v54n5a+nXOYenf3wkR+94FVZUxj3eHwnCO3dC+17QfYQ387L7IdBtBBQMhUCN2Z3BUlg8I/4Ha1amiIi0MQppkrD3vtrKTbOWsKmkkh8eO5DrTxpKTmYGvLGy9sVjAY68Ck69u/4PyGrvzb7UrEwRERGFNKlfSUWIO15dxsxPNjC4ay4zrzya7/Tv5I0vWzQDFj1T+5sDOV4LWqI0K1NERARQSJN6zP3iW3790lK2l1dxzaTB/HTyQWT7nBfO3rsHtq/yui6DpfHHkh1IN6VmZYqIiCikiWff+24ePaSAe+YsZ/bizRzcswOPXjaWQ3vkwtKZ8K97vIkA3UfC1Ke89crWf6RuShERkSRSSJP97ruZ6TeqIo4Mn/H/ThrKFf8xgMCyWTCrOpwdClOfhGGng8/nnUTdlCIiIkmlkNbGVd93k2AZx/s+5Uvrz8pIH3xEyfL7+FH+QgIPnl97OKtJ3ZQiIiJJo5DWxs1etInR0eUcl/ERfw6fQxUZ/NL/NJf557LD8sn+RxF0OwTOewKGT4kfzkRERCTpFNLauE9XbaQ4HOBOdzHH+Bbzx4xpdPOVAJDltvGPwbdxxkXXKZyJiIg0M4W0Nqq4vIp75i7n+UXb6EoHng/cxhjfV5jtOSZIgO45poAmIiKSAgppbUw06nhu4Xp+N2c5O3eFuaX7h1y64wH85vY7NseqGJ23PQWlFBEREYW0NmTpxhJueXkpn63bwdl9Srkjbybt17yJq9l8VkMkI4fMrkOauZQiIiICCmltQklliHvfWMET89cyNKeMd4bOZcD6F7HSPJh4A/bhfRAq3+99fp/ulykiIpIqCmmtmHOOFz/byF2vLSdUXsyj/f7FhG3PYxvCMO4KmPALyO0CgyZpIVoREZE0o5DWSq34ttTr2vymkF92+YDL/c+TsaUYDj0XJv8aOg/cc7AWohUREUk7Cmkt3L63c5o0vBt/e381j/57Nd/LWsAjnZ4nr3wjDJwIJ/4Geh0e/0RaiFZERCStKKS1YAvWbOfqR9/jFPchPSObeZVR3DRrBON9X/Bu/gv0qVwBHUbC9/4Mg4+HWiYIiIiISPpp9pBmZqcA/wP4gb855+6Oc8x5wG2AAxY55y5s1kK2AGXBMH9+9HHe4bescd25213INpfHc5m3Mcb3NdFAHzjlIRh5ntY5ExERaYGaNaSZmR+4HzgR2AAsMLNXnHPLahxzEHAjcIxzrtjMujVnGVuKOZ98zTR+y5LoQG4M/YifB2Zypm8eAEGXwWvjnua7hx2W4lKKiIjIgWruJpZxwErn3GrnXBUwAzhrn2N+DNzvnCsGcM4VNnMZW4Scr1/mi2h/Ho+cyJysX3Gmbx5mXo9mGD/Zq+amuogiIiLSCM0d0noD62u83hDbVtNQYKiZ/dvM5se6R/djZj8xs4VmtnDr1q1NVNz05asq5dXIUfwlcB9+onsNN8u1IP1936aucCIiItJozR3S4o1c3/d+RBnAQcBxwAXA38ys435vcu5h59wY59yYrl27Jr2g6WxDcQVfbyzi9szphMggYJG99le4LAYPG5Wi0omIiEgyNHdI2wD0rfG6D7ApzjEvO+dCzrlvgBV4oU2ArTsr+fcDV/FTZlBCe9pZ1X7HZGVmkHXYuSkonYiIiCRLc4e0BcBBZjbQzDKB84FX9jnmJWASgJkV4HV/rm7WUqapkrIKFt93AVNDL1F48CXkX/YsLjOPkL8dACF/O1xmHv6LX9BCtCIiIi1cs87udM6FzexaYC7eEhyPOOe+MLPbgYXOuVdi+04ys2VABPiFc25bc5YzHVWW7WTVX87h+KoFfDPq5wz87q1ghl2/gkDsTgEB3SlARESk1TDn9h0S1vKMGTPGLVy4MNXFaDJVO4tYd98UBgaX88URtzHqrJ+lukgiIiJygMzsE+fcmPqO0x0H0lykeD3bpp1O3+AmPvzOvfzHmT9IdZFERESkGWgp+jTmCpdT+sBkcoOFzD38fgU0ERGRNkQhLV2tX8Cuh08kVBVk5siHOfPsqakukYiIiDQjhbR09PWbhB6dwpaqdjw6/GEu/94ZqS6RiIiINDOFtHSz6FmiT5/PinAPHhzyANdPPRmzeGsAi4iISGumiQPp5MP74I2bmR8dwWP97uIvF07E71NAExERaYsU0lIlWApLZ8H21dBpIBQth/nTeD16JI/1uIlHLjmWzAw1dIqIiLRVCmmpsHYePHUuuCiEKsD84CI8Ez2BJzpdyzOXHUNOpv5oRERE2jIlgeYWLPUCWlXZnm0uwgOhM3jOTeK5/zyM/JxA6sonIiIiaUH9ac1t6Swi0ejul87B3aGpPBY5mcey/0i3dbNTWDgRERFJF2pJa2ZVW1eSGa7Y/frO8IXMjEzk+czb6R/d6O1PYflEREQkPaglrZl9Xt6ZShfAOXgnchjPRI5neuY9HOTbSIXL4vOyzqkuooiIiKQBhbRmNi9wJJmEqSCL60LXcG9gGqN9qwCIYnyQNSHFJRQREZF0oJDWzE4qfha/Oa4PXcVI+4aTfAspd1mUuWyudDfSu3tBqosoIiIiaUBj0prT2g8Z/s1jPB05njnRsZzs+5hpkTNZ47ozO3IUlpXLQ6N6pbqUIiIikgYU0ppLsAxeuopQ+77cWXQRGX4f7/uPZm7VkeRk+rEMmH75OHKz9EciIiIiCmnN581bcMVruavL7wm0a88bPz2WD1YWsaaoggEFOUwZ1UsBTURERHZTKmgOK9+ChY+w5qDLmb6kF7efNZTenXKYOrZfqksmIiIiaUohralVFsPL1xItGMaPNpzC0O7tuHCcwpmIiIjUTbM7m9prv4TyrbzQ/xZWFUf4rymHkOFXtYuIiEjdlBaa0rKXYclzlB35c25dEODEEd059iAtsSEiIiL1U0hrKmWFMPvn0HM0t5ecSigS5ebTDk51qURERKSFUEhrCs7BP34GwTKWH/V7nvv0W35wzEAGFOSmumQiIiLSQiikNYVFz8CKV3GTf83N/w5TkJfJtZOHpLpUIiIi0oIopCVbyQZ4/QbodzSv5JzNJ2uL+cXJw2ifHUh1yURERKQFUUhLpmgUXr4GohF2nX4fd8/5mkN6deDc7/RNdclERESkhdE6acm08O+w+l2Y8kemLY6wuWQX/3P+4fh9luqSiYiISAujlrRk2bYK3rgFhpzApsHn89C/VnH6qJ6MG9g51SUTERGRFkghLRmiEXjxSsjIhDP/wt1zVuAc3Hjq8FSXTERERFoohbRk+PDPsOFjOO0PLNyezSuLNnHFhEH06ZST6pKJiIhIC6WQ1lhbvoB37oIRZxE95Hv85h/L6NEhmyuPG5zqkomIiEgLppDWGOEqmHUFZOfD6ffywmcbWbKxhBtOHUZOpuZkiIiIyIFTkmioYCksnQXbV0PhMtiyBM5/hrKMjtwz93MO79eRsw7rnepSioiISAunkNYQa+fBU+eCi0Kowtvmy4B2nbj/nZVsLQ3y10vG4NOSGyIiItJI6u5MVLDUC2hVZXsCGkA0zLonruTv76/mnMN7M7pvx9SVUURERFoNhbRELZ3ltaDFceeuc8mwKL88RUtuiIiISHIopCVq++q9W9BiPoyMYG74CK7ut5Ee+dkpKJiIiIi0Rgppieo8CAJ7r3sWdj5uD/8nva2IH41ul6KCiYiISGukkJaoQ88B27u6ZkQmsdz15+bsmWQfdk6KCiYiIiKtkUJaorLaw0UzcZl5hPztKHE5/CF8HuP8X3HqpTdCVl6qSygiIiKtiJbgaIAFbhhXB+/nZPchy0PdKSaPr3wDWchwxqa6cCIiItKqqCUtQWXBMJc9+jFbgwGerJrIQjccMHaEAlz26MeUB8OpLqKIiIi0IgppCZq9aBPOxd/nHMxevKl5CyQiIiKtmkJagtZsK6eiKhJ3X0VVhDVF+y/PISIiInKgFNISNKBLLjmZ/rj7cjL9DCjIibtPRERE5EAopCVoymG9sFpuyWkGU0b1at4CiYiISKumkJagvKwMpl8+jtws/+4WtZxMP7lZ/th2TZQVERGR5FGyaICxAzrz8U0nMHvxJtYUVTCgIIcpo3opoImIiEjSKV00UG5WBlPH9kt1MURERKSVU3eniIiISBpSSBMRERFJQwppIiIiImlIIU1EREQkDSmkiYiIiKQhhTQRERGRNKSQJiIiIpKGFNJERERE0pBCmoiIiEgaUkgTERERSUMKaSIiIiJpSCFNREREJA0ppImIiIikIYU0ERERkTSkkCYiIiKShhTSRERERNKQQpqIiIhIGlJIExEREUlDCmkiIiIiacicc6kuQ6OZ2VZgbYKHFwBFTVgcUR03F9Vz01MdNz3VcfNQPTe9htRxf+dc1/oOahUhrSHMbKFzbkyqy9GaqY6bh+q56amOm57quHmonpteU9SxujtFRERE0pBCmoiIiEgaaosh7eFUF6ANUB03D9Vz01MdNz3VcfNQPTe9pNdxmxuTJiIiItIStMWWNBEREZG0p5AmIiIikobaVEgzs1PMbIWZrTSzX6W6PK2Rma0xsyVm9rmZLUx1eVoLM3vEzArNbGmNbZ3N7E0z+zr22CmVZWzpaqnj28xsY+x6/tzMTktlGVs6M+trZu+Y2Zdm9oWZXRfbrms5SeqoY13LSWRm2Wb2sZktitXzb2LbB5rZR7Fr+Vkzy2zU57SVMWlm5ge+Ak4ENgALgAucc8tSWrBWxszWAGOcc1o0MYnMbAJQBjzunDs0tu0eYLtz7u7Yfzo6OeduSGU5W7Ja6vg2oMw599+pLFtrYWY9gZ7OuU/NrD3wCXA2cBm6lpOijjo+D13LSWNmBuQ658rMLAB8AFwH/F9glnNuhpk9CCxyzk070M9pSy1p44CVzrnVzrkqYAZwVorLJJIQ59y/gO37bD4LeCz2/DG8X8RygGqpY0ki59xm59ynseelwJdAb3QtJ00ddSxJ5DxlsZeB2I8DJgMzY9sbfS23pZDWG1hf4/UGdOE2BQe8YWafmNlPUl2YVq67c24zeL+YgW4pLk9rda2ZLY51h6obLknMbABwOPARupabxD51DLqWk8rM/Gb2OVAIvAmsAnY458KxQxqdM9pSSLM429pGX2/zOsY5dwRwKnBNrAtJpKWaBgwGRgObgT+ktjitg5nlAS8AP3PO7Ux1eVqjOHWsaznJnHMR59xooA9eb93B8Q5rzGe0pZC2Aehb43UfYFOKytJqOec2xR4LgRfxLlxpGlti40+qx6EUprg8rY5zbkvsF3EU+Cu6nhstNn7nBeAp59ys2GZdy0kUr451LTcd59wO4F1gPNDRzDJiuxqdM9pSSFsAHBSbeZEJnA+8kuIytSpmlhsbqIqZ5QInAUvrfpc0wivApbHnlwIvp7AsrVJ1cIj59KnRJAAAA7lJREFULrqeGyU22PrvwJfOuXtr7NK1nCS11bGu5eQys65m1jH2vB1wAt74v3eAc2OHNfpabjOzOwFiU47/BPiBR5xzd6a4SK2KmQ3Caz0DyACeVh0nh5k9AxwHFABbgFuBl4DngH7AOuD7zjkNfD9AtdTxcXjdQw5YA1xRPXZKGs7MjgXeB5YA0djmm/DGTOlaToI66vgCdC0njZmNwpsY4Mdr8HrOOXd77N/BGUBn4DPgYudc8IA/py2FNBEREZGWoi11d4qIiIi0GAppIiIiImlIIU1EREQkDSmkiYiIiKQhhTQRERGRNKSQJiItnpndZmaulp+LU1AeZ2bXNvfnikjrklH/ISIiLUIJcEqc7SubuyAiIsmgkCYirUXYOTc/1YUQEUkWdXeKSKtnZgNiXZAXmtkTZlZqZoVmdmucYyeb2UdmtsvMtpjZA7GbVdc8pouZPWRmm2PHrTCzn+1zKr+Z3WVmW2Ofdb+ZZTXpFxWRVkUtaSLSatS4sfFuzrlwjZe/B2bj3VtvAnCrmRU55+6PvX8EMAd4E/ge0Be4GxhErCs1dp++d4FuwG+A5cCQ2E9N1wNvAxcDo4DfAmuBexr/TUWkLdBtoUSkxTOz2/DutRnPwNjjN8CbzrmTarzvr8BpQF/nXNTMZgDfAYY75yKxY84DngWOds7NM7MrgGnAEc65z2spjwPed85NqLHtJaCHc258I76qiLQh6u4UkdaiBBgb52dTjWNe3Oc9s4BeQJ/Y63HAi9UBLeYFIAwcG3s9GfistoBWwxv7vF5W43NEROql7k4RaS3CzrmF8XaYWfXTwn12Vb/uCayLPW6peYBzLmJm24DOsU1dgM0JlGfHPq+rgOwE3iciAqglTUTalm61vN5c43GvY8zMjxfMtsc2bcMLcyIiTUohTUTaku/u8/ocvGC2Ifb6I+C7sWBW85gM4IPY638Ch5vZqKYsqIiIujtFpLXIMLN4g/LX13h+iJk9hDfObALwQ+A651w0tv//A58BL5nZNLwxZL8D5jrn5sWOeRy4BngjNmFhBd7khKHOuV8l+TuJSBumkCYirUU+MC/O9luAJ2PPfwlMwQtpu4A7gPuqD3TOfWFmpwJ34U0q2Ak8E3tf9TG7zGwy3tIctwMdgDXAA8n9OiLS1mkJDhFp9cxsAN4SHGc452antjQiIonRmDQRERGRNKSQJiIiIpKG1N0pIiIikobUkiYiIiKShhTSRERERNKQQpqIiIhIGlJIExEREUlDCmkiIiIiaeh/AXpsGB8yegdfAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "pr_auc_history_train = train_metric_callback.get_pr_auc_history()\n",
    "pr_auc_history_val = val_callback.get_pr_auc_history()\n",
    "\n",
    "plt.figure(figsize=(10, 7))\n",
    "plot_with_dots(plt, pr_auc_history_train[-1])\n",
    "plot_with_dots(plt, pr_auc_history_val[-1])\n",
    "\n",
    "plt.xlabel('Epoch', fontsize=15)\n",
    "plt.ylabel('Mean PR AUC', fontsize=15)\n",
    "plt.legend(['Train', 'Val'])\n",
    "plt.title('Training and Validation PR AUC', fontsize=20)\n",
    "plt.savefig('pr_auc_hist.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10, 7))\n",
    "plot_with_dots(plt, history_0.history['loss']+history_1.history['loss'])\n",
    "plot_with_dots(plt, history_0.history['val_loss']+history_1.history['val_loss'])\n",
    "\n",
    "plt.xlabel('Epoch', fontsize=15)\n",
    "plt.ylabel('Binary Crossentropy', fontsize=15)\n",
    "plt.legend(['Train', 'Val'])\n",
    "plt.title('Training and Validation Loss', fontsize=20)\n",
    "plt.savefig('loss_hist.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-09T06:29:56.355960Z",
     "start_time": "2019-11-09T06:29:56.350884Z"
    }
   },
   "outputs": [],
   "source": [
    "import efficientnet\n",
    "efficientnet.sq"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I left the model to train longer on my local GPU. I then upload the best model and plots from the model training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:19:38.367460Z",
     "start_time": "2019-11-16T15:19:06.134736Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "W1116 18:19:15.930072 140150228264768 deprecation_wrapper.py:119] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
      "\n",
      "W1116 18:19:16.040837 140150228264768 deprecation.py:323] From /home/dex/anaconda3/envs/keras_env/lib/python3.7/site-packages/tensorflow/python/ops/math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
     ]
    }
   ],
   "source": [
    "from keras_radam import RAdam\n",
    "model = load_model('/home/dex/Downloads/classifier_densenet169_epoch_9_val_pr_auc_0.8459108536857672.h5', custom_objects={'RAdam':RAdam})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Selecting postprocessing thresholds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-12T18:41:54.222291Z",
     "start_time": "2019-11-12T18:41:40.710783Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7ca23eed123e432cbbde8752283db58c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "class_names = ['Fish', 'Flower', 'Sugar', 'Gravel']\n",
    "def get_threshold_for_recall(y_true, y_pred, class_i, recall_threshold=0.94, precision_threshold=0.90, plot=False):\n",
    "    precision, recall, thresholds = precision_recall_curve(y_true[:, class_i], y_pred[:, class_i])\n",
    "    i = len(thresholds) - 1\n",
    "    best_recall_threshold = None\n",
    "    while best_recall_threshold is None:\n",
    "        next_threshold = thresholds[i]\n",
    "        next_recall = recall[i]\n",
    "        if next_recall >= recall_threshold:\n",
    "            best_recall_threshold = next_threshold\n",
    "        i -= 1\n",
    "        \n",
    "    # consice, even though unnecessary passing through all the values\n",
    "    best_precision_threshold = [thres for prec, thres in zip(precision, thresholds) if prec >= precision_threshold][0]\n",
    "    \n",
    "    if plot:\n",
    "        plt.figure(figsize=(10, 7))\n",
    "        plt.step(recall, precision, color='r', alpha=0.3, where='post')\n",
    "        plt.fill_between(recall, precision, alpha=0.3, color='r')\n",
    "        plt.axhline(y=precision[i + 1])\n",
    "        recall_for_prec_thres = [rec for rec, thres in zip(recall, thresholds) \n",
    "                                 if thres == best_precision_threshold][0]\n",
    "        plt.axvline(x=recall_for_prec_thres, color='g')\n",
    "        plt.xlabel('Recall')\n",
    "        plt.ylabel('Precision')\n",
    "        plt.ylim([0.0, 1.05])\n",
    "        plt.xlim([0.0, 1.0])\n",
    "        plt.legend(['PR curve', \n",
    "                    f'Precision {precision[i + 1]: .2f} corresponding to selected recall threshold',\n",
    "                    f'Recall {recall_for_prec_thres: .2f} corresponding to selected precision threshold'])\n",
    "        plt.title(f'Precision-Recall curve for Class {class_names[class_i]}')\n",
    "    return best_recall_threshold, best_precision_threshold\n",
    "\n",
    "y_pred = model.predict_generator(data_generator_val, workers=num_cores)\n",
    "y_true = data_generator_val.get_labels()\n",
    "recall_thresholds = dict()\n",
    "precision_thresholds = dict()\n",
    "for i, class_name in tqdm(enumerate(class_names)):\n",
    "    recall_thresholds[class_name], precision_thresholds[class_name] = get_threshold_for_recall(y_true, y_pred, i, plot=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Post-processing segmentation submission"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Predicting cloud classes for test."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:21:16.198176Z",
     "start_time": "2019-11-16T15:20:47.922803Z"
    }
   },
   "outputs": [],
   "source": [
    "data_generator_test = DataGenenerator(folder_imgs=test_imgs_folder, shuffle=False)\n",
    "y_pred_test = model.predict_generator(data_generator_test, workers=num_cores)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Estimating set of images without masks."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:44:26.919122Z",
     "start_time": "2019-11-16T15:44:26.915305Z"
    }
   },
   "outputs": [],
   "source": [
    "recall_thresholds = {'Fish': 0.29,\n",
    " 'Flower': 0.4,\n",
    " 'Sugar': 0.29,\n",
    " 'Gravel': 0.29}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:44:27.869289Z",
     "start_time": "2019-11-16T15:44:27.819770Z"
    }
   },
   "outputs": [],
   "source": [
    "image_labels_empty = set()\n",
    "class_names = ['Fish', 'Flower', 'Sugar', 'Gravel']\n",
    "for i, (img, predictions) in enumerate(zip(os.listdir(test_imgs_folder), y_pred_test)):\n",
    "    for class_i, class_name in enumerate(class_names):\n",
    "        if predictions[class_i] < recall_thresholds[class_name]:\n",
    "            image_labels_empty.add(f'{img}_{class_name}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:25:56.591591Z",
     "start_time": "2019-11-16T15:25:56.573879Z"
    }
   },
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'precision_thresholds' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-20-14df8bfaf484>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mrecall_thresholds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mprecision_thresholds\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'precision_thresholds' is not defined"
     ]
    }
   ],
   "source": [
    "recall_thresholds, precision_thresholds"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Segmentation results:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:44:31.269807Z",
     "start_time": "2019-11-16T15:44:31.068930Z"
    }
   },
   "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>Image_Label</th>\n",
       "      <th>EncodedPixels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>877691a.jpg_Fish</td>\n",
       "      <td>108 238 449 249 796 253 1142 257 1489 260 1836...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>877691a.jpg_Flower</td>\n",
       "      <td>113830 27 114176 79 114523 107 114870 113 1152...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>877691a.jpg_Gravel</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>877691a.jpg_Sugar</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>e810552.jpg_Fish</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Image_Label                                      EncodedPixels\n",
       "0    877691a.jpg_Fish  108 238 449 249 796 253 1142 257 1489 260 1836...\n",
       "1  877691a.jpg_Flower  113830 27 114176 79 114523 107 114870 113 1152...\n",
       "2  877691a.jpg_Gravel                                                NaN\n",
       "3   877691a.jpg_Sugar                                                NaN\n",
       "4    e810552.jpg_Fish                                                NaN"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission = pd.read_csv('sub_convex_0.6586.csv')\n",
    "#pd.read_csv('/home/dex/Desktop/ml/cloud artgor/submissions/submission_Model_segmentation_segm_resnet152_bs_5_2019-11-05.csv')\n",
    "#submission = pd.read_csv('../input/densenet201cloudy/densenet201.csv')\n",
    "submission.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:34:15.438788Z",
     "start_time": "2019-11-16T15:34:15.430760Z"
    }
   },
   "outputs": [],
   "source": [
    "predictions_nonempty = set(submission.loc[~submission['EncodedPixels'].isnull(), 'Image_Label'].values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:34:16.067608Z",
     "start_time": "2019-11-16T15:34:16.062552Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "92 masks would be removed\n"
     ]
    }
   ],
   "source": [
    "print(f'{len(image_labels_empty.intersection(predictions_nonempty))} masks would be removed')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-16T15:34:16.955446Z",
     "start_time": "2019-11-16T15:34:16.686423Z"
    }
   },
   "outputs": [],
   "source": [
    "#removing masks\n",
    "submission.loc[submission['Image_Label'].isin(image_labels_empty), 'EncodedPixels'] = np.nan\n",
    "submission.to_csv('submission_segmentation_and_classifier_628.csv', index=None)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Future work\n",
    "1. estimate distribution of classes in test set using the classifier. Then, if necessary and doable, modify val set accordingly,\n",
    "2. use the classifier with explainability technique [Gradient-weighted Class Activation Mapping](http://gradcam.cloudcv.org/) to generate a baseline, (please see [GradCAM: extracting masks from classifier](https://www.kaggle.com/samusram/gradcam-extracting-masks-from-classifier)),\n",
    "3. improve the classifier,\n",
    "4. use the classifier as backbone for UNet-like solution."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-12T18:30:02.939807Z",
     "start_time": "2019-11-12T18:30:02.706581Z"
    }
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import os\n",
    "from tqdm import tqdm\n",
    "test_imgs_folder = '/home/dex/Desktop/ml/cloud data//test_images/'\n",
    "folder_images = '/home/dex/Desktop/ml/clouds/input/understanding-clouds-resized/test_images_525/test_images_525'\n",
    "model_class_names=['Fish', 'Flower', 'Gravel', 'Sugar']\n",
    "sub = pd.read_csv('/home/dex/Downloads/submission_segmentation_and_classifier (1).csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-12T18:30:04.061839Z",
     "start_time": "2019-11-12T18:30:04.049202Z"
    }
   },
   "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>Image_Label</th>\n",
       "      <th>EncodedPixels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>002f507.jpg_Fish</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>002f507.jpg_Flower</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>002f507.jpg_Gravel</td>\n",
       "      <td>5 337 353 343 702 346 1052 347 1401 348 1751 3...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>002f507.jpg_Sugar</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0035ae9.jpg_Fish</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          Image_Label                                      EncodedPixels\n",
       "0    002f507.jpg_Fish                                                NaN\n",
       "1  002f507.jpg_Flower                                                NaN\n",
       "2  002f507.jpg_Gravel  5 337 353 343 702 346 1052 347 1401 348 1751 3...\n",
       "3   002f507.jpg_Sugar                                                NaN\n",
       "4    0035ae9.jpg_Fish                                                NaN"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-11-12T18:32:40.283747Z",
     "start_time": "2019-11-12T18:30:20.228515Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 3698/3698 [02:19<00:00, 27.25it/s]\n"
     ]
    }
   ],
   "source": [
    "def rle_decode(mask_rle: str = '', shape = (1400, 2100)):\n",
    "    '''\n",
    "    Decode rle encoded mask.\n",
    "    \n",
    "    :param mask_rle: run-length as string formatted (start length)\n",
    "    :param shape: (height, width) of array to return \n",
    "    Returns numpy array, 1 - mask, 0 - background\n",
    "    '''\n",
    "    s = mask_rle.split()\n",
    "    starts, lengths = [np.asarray(x, dtype=int) for x in (s[0:][::2], s[1:][::2])]\n",
    "    starts -= 1\n",
    "    ends = starts + lengths\n",
    "    img = np.zeros(shape[0] * shape[1], dtype=np.uint8)\n",
    "    for lo, hi in zip(starts, ends):\n",
    "        img[lo:hi] = 1\n",
    "    \n",
    "    return img.reshape(shape, order='F')\n",
    "\n",
    "def mask2rle(img):\n",
    "    '''\n",
    "    Convert mask to rle.\n",
    "    img: numpy array, 1 - mask, 0 - background\n",
    "    Returns run length as string formated\n",
    "    '''\n",
    "    pixels= img.T.flatten()\n",
    "    pixels = np.concatenate([[0], pixels, [0]])\n",
    "    runs = np.where(pixels[1:] != pixels[:-1])[0] + 1\n",
    "    runs[1::2] -= runs[::2]\n",
    "    return ' '.join(str(x) for x in runs)\n",
    "\n",
    "\n",
    "\n",
    "\n",
    "min_size = [10000 ,10000, 10000, 10000]\n",
    "def post_process_minsize(mask, min_size):\n",
    "    \"\"\"\n",
    "    Post processing of each predicted mask, components with lesser number of pixels\n",
    "    than `min_size` are ignored\n",
    "    \"\"\"\n",
    "    \n",
    "    num_component, component = cv2.connectedComponents(mask.astype(np.uint8))\n",
    "    predictions = np.zeros(mask.shape)\n",
    "    num = 0\n",
    "    for c in range(1, num_component):\n",
    "        p = (component == c)\n",
    "        if p.sum() > min_size:\n",
    "            predictions[p] = 1\n",
    "            num += 1\n",
    "    return predictions #, num\n",
    "def make_mask(df, image_label, shape = (1400, 2100), cv_shape = (525, 350),debug=False):\n",
    "    \"\"\"\n",
    "    Create mask based on df, image name and shape.\n",
    "    \"\"\"\n",
    "    if debug:\n",
    "        print(shape,cv_shape)\n",
    "    df = df.set_index('Image_Label')\n",
    "    encoded_mask = df.loc[image_label, 'EncodedPixels']\n",
    "#     print('encode: ',encoded_mask[:10])\n",
    "    mask = np.zeros((shape[0], shape[1]), dtype=np.float32)\n",
    "    if encoded_mask is not np.nan:\n",
    "        mask = rle_decode(encoded_mask,shape=shape) # original size\n",
    "            \n",
    "    return cv2.resize(mask, cv_shape)\n",
    "def draw_convex_hull(mask, mode='convex'):\n",
    "    \n",
    "    img = np.zeros(mask.shape)\n",
    "    contours, hier = cv2.findContours(mask, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)\n",
    "    \n",
    "    for c in contours:\n",
    "        if mode=='rect': # simple rectangle\n",
    "            x, y, w, h = cv2.boundingRect(c)\n",
    "            cv2.rectangle(img, (x, y), (x+w, y+h), (255, 255, 255), -1)\n",
    "        elif mode=='convex': # minimum convex hull\n",
    "            hull = cv2.convexHull(c)\n",
    "            cv2.drawContours(img, [hull], 0, (255, 255, 255),-1)\n",
    "        elif mode=='approx':\n",
    "            epsilon = 0.02*cv2.arcLength(c,True)\n",
    "            approx = cv2.approxPolyDP(c,epsilon,True)\n",
    "            cv2.drawContours(img, [approx], 0, (255, 255, 255),-1)\n",
    "        else: # minimum area rectangle\n",
    "            rect = cv2.minAreaRect(c)\n",
    "            box = cv2.boxPoints(rect)\n",
    "            box = np.int0(box)\n",
    "            cv2.drawContours(img, [box], 0, (255, 255, 255),-1)\n",
    "    return img/255.\n",
    "\n",
    "mode='convex' # choose from 'rect', 'min', 'convex' and 'approx'\n",
    "\n",
    "img_label_list = []\n",
    "enc_pixels_list = []\n",
    "test_imgs = os.listdir(folder_images)\n",
    "for test_img_i, test_img in enumerate(tqdm(test_imgs)):\n",
    "    for class_i, class_name in enumerate(model_class_names):\n",
    "        \n",
    "        path = os.path.join(folder_images, test_img)\n",
    "        img = cv2.imread(path).astype(np.float32) # use already-resized ryches' dataset\n",
    "        img = img/255.\n",
    "        img2 = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n",
    "        \n",
    "        img_label_list.append(f'{test_img}_{class_name}')\n",
    "        \n",
    "        mask = make_mask(sub, test_img + '_' + class_name,shape=(350,525))\n",
    "        if True:\n",
    "        #if class_name == 'Flower' or class_name =='Sugar': # you can decide to post-process for some certain classes \n",
    "            mask = draw_convex_hull(mask.astype(np.uint8), mode=mode)\n",
    "        mask[img2<=2/255.] = 0\n",
    "        mask = post_process_minsize(mask, min_size[class_i])\n",
    "        \n",
    "        if mask.sum() == 0:\n",
    "            enc_pixels_list.append(np.nan)\n",
    "        else:\n",
    "            mask = np.where(mask > 0.5, 1.0, 0.0)\n",
    "            enc_pixels_list.append(mask2rle(mask))\n",
    "            \n",
    "submission_df = pd.DataFrame({'Image_Label': img_label_list, 'EncodedPixels': enc_pixels_list})\n",
    "submission_df.to_csv('sub_convex_0.6586.csv', index=None)\n",
    "class_names = ['Fish', 'Flower', 'Sugar', 'Gravel']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.3"
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {
     "0a53fa055d2b43358bafff0a453d2c8b": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "25064ebbb12a483690e0fd0feb52560b": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "ProgressStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "ProgressStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "bar_color": null,
       "description_width": "initial"
      }
     },
     "38ff85bb09144a9aaebf4dd59f000c69": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HBoxModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HBoxModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HBoxView",
       "box_style": "",
       "children": [
        "IPY_MODEL_7640235bcae6457c80e41c51a06677d5",
        "IPY_MODEL_dd1866cdda2742debd15613963dd4759"
       ],
       "layout": "IPY_MODEL_0a53fa055d2b43358bafff0a453d2c8b"
      }
     },
     "4fcf62cbaf71492f85b8e7e6f69a9105": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "DescriptionStyleModel",
      "state": {
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "DescriptionStyleModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "StyleView",
       "description_width": ""
      }
     },
     "7640235bcae6457c80e41c51a06677d5": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "IntProgressModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "IntProgressModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "ProgressView",
       "bar_style": "success",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_eac971125b6e4705b4db95f6c2ab6318",
       "max": 1,
       "min": 0,
       "orientation": "horizontal",
       "style": "IPY_MODEL_25064ebbb12a483690e0fd0feb52560b",
       "value": 1
      }
     },
     "ca8b02376bfc46859708c5f3069f7ca5": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     },
     "dd1866cdda2742debd15613963dd4759": {
      "model_module": "@jupyter-widgets/controls",
      "model_module_version": "1.5.0",
      "model_name": "HTMLModel",
      "state": {
       "_dom_classes": [],
       "_model_module": "@jupyter-widgets/controls",
       "_model_module_version": "1.5.0",
       "_model_name": "HTMLModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/controls",
       "_view_module_version": "1.5.0",
       "_view_name": "HTMLView",
       "description": "",
       "description_tooltip": null,
       "layout": "IPY_MODEL_ca8b02376bfc46859708c5f3069f7ca5",
       "placeholder": "​",
       "style": "IPY_MODEL_4fcf62cbaf71492f85b8e7e6f69a9105",
       "value": " 4/? [00:00&lt;00:00, 13.35it/s]"
      }
     },
     "eac971125b6e4705b4db95f6c2ab6318": {
      "model_module": "@jupyter-widgets/base",
      "model_module_version": "1.2.0",
      "model_name": "LayoutModel",
      "state": {
       "_model_module": "@jupyter-widgets/base",
       "_model_module_version": "1.2.0",
       "_model_name": "LayoutModel",
       "_view_count": null,
       "_view_module": "@jupyter-widgets/base",
       "_view_module_version": "1.2.0",
       "_view_name": "LayoutView",
       "align_content": null,
       "align_items": null,
       "align_self": null,
       "border": null,
       "bottom": null,
       "display": null,
       "flex": null,
       "flex_flow": null,
       "grid_area": null,
       "grid_auto_columns": null,
       "grid_auto_flow": null,
       "grid_auto_rows": null,
       "grid_column": null,
       "grid_gap": null,
       "grid_row": null,
       "grid_template_areas": null,
       "grid_template_columns": null,
       "grid_template_rows": null,
       "height": null,
       "justify_content": null,
       "justify_items": null,
       "left": null,
       "margin": null,
       "max_height": null,
       "max_width": null,
       "min_height": null,
       "min_width": null,
       "object_fit": null,
       "object_position": null,
       "order": null,
       "overflow": null,
       "overflow_x": null,
       "overflow_y": null,
       "padding": null,
       "right": null,
       "top": null,
       "visibility": null,
       "width": null
      }
     }
    },
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}