Nuclie Segmentation - UNet.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Nuclie Semantic Segmentation - UNet using Tensorflow 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "e61ef2d8-f315-4f7f-b07e-1de0f4e8441a",
    "_uuid": "1677fddbb95f7545b6540e9201f3339a0fdbfc5d"
   },
   "source": [
    "# Intro\n",
    "- Dataset used is from Kaggle's Data Science Bowl 2018 - Nuclei Segmentation\n",
    "- The architecture used is [U-Net](https://arxiv.org/abs/1505.04597), which is very common for image segmentation problems such as this."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "_cell_guid": "c332549b-8d23-4bb5-8497-e7a8eb8b21d2",
    "_uuid": "5c38504af3a84bee68c66d3cde74443c58df422f",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import random\n",
    "import warnings\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from tqdm import tqdm\n",
    "from itertools import chain\n",
    "from skimage.io import imread, imshow, imread_collection, concatenate_images\n",
    "from skimage.transform import resize\n",
    "from skimage.morphology import label\n",
    "\n",
    "from tensorflow.keras.models import Model, load_model\n",
    "from tensorflow.keras.layers import Input\n",
    "from tensorflow.keras.layers import Dropout, Lambda\n",
    "from tensorflow.keras.layers import Conv2D, Conv2DTranspose\n",
    "from tensorflow.keras.layers import MaxPooling2D\n",
    "from tensorflow.keras.layers import concatenate\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from tensorflow.keras import backend as K\n",
    "\n",
    "import tensorflow as tf\n",
    "\n",
    "# Set some parameters\n",
    "IMG_WIDTH = 128\n",
    "IMG_HEIGHT = 128\n",
    "IMG_CHANNELS = 3\n",
    "#TRAIN_PATH = '../input/stage1_train/'\n",
    "#TEST_PATH = '../input/stage1_test/'\n",
    "\n",
    "TRAIN_PATH = 'data/stage1_train/'\n",
    "TEST_PATH = 'data/stage1_test/'\n",
    "\n",
    "dir_path = 'C:\\\\Users\\\\Advait\\\\Downloads\\\\Data Science\\\\Kaggle\\\\MULTIC~1\\\\MEDICA~1\\\\MEDICA~1\\\\'\n",
    "\n",
    "warnings.filterwarnings('ignore', category=UserWarning, module='skimage')\n",
    "seed = 42\n",
    "random.seed = seed\n",
    "np.random.seed = seed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'data/stage1_train/'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "TRAIN_PATH"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "_cell_guid": "ffa0caf0-2d1b-40f2-865b-8e6db88526b6",
    "_uuid": "3fb9d6530fbbd0e22e41fc4fd9fd9fc0bff027ac",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Get train and test IDs\n",
    "train_ids = next(os.walk(TRAIN_PATH))[1]\n",
    "test_ids = next(os.walk(TEST_PATH))[1]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "59c4a25d-645f-4b74-9c53-145ac78cc481",
    "_uuid": "875af74f980236825de3a650825b46e25632422c"
   },
   "source": [
    "# Get the data\n",
    "- Downsample both the training and test images to reduce computations\n",
    "- Retain record of the original sizes of the test images to upsample predicted masks and create correct run-length encodings "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Getting and resizing train images and masks ... \n"
     ]
    }
   ],
   "source": [
    "# Get and resize train images and masks\n",
    "X_train = np.zeros((len(train_ids), IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS), dtype=np.uint8)\n",
    "Y_train = np.zeros((len(train_ids), IMG_HEIGHT, IMG_WIDTH, 1), dtype=np.bool)\n",
    "print('Getting and resizing train images and masks ... ')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "_cell_guid": "ca0cc34b-c26f-41ee-88d7-975aebdb634e",
    "_uuid": "9e389ba8bdb5b6fc03b231b6a6c84a8bde634053"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████| 670/670 [04:29<00:00,  6.16it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Getting and resizing test images ... \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████████| 65/65 [00:01<00:00, 44.24it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done!\n"
     ]
    }
   ],
   "source": [
    "\n",
    "sys.stdout.flush()\n",
    "for n, id_ in tqdm(enumerate(train_ids), total=len(train_ids)):\n",
    "    \n",
    "    #Read image files iteratively\n",
    "    path = TRAIN_PATH + id_\n",
    "    img = imread(dir_path + path + '/images/' + id_ + '.png')[:,:,:IMG_CHANNELS]\n",
    "    img = resize(img, (IMG_HEIGHT, IMG_WIDTH), mode='constant', preserve_range=True)\n",
    "    \n",
    "    #Append image to numpy array for train dataset\n",
    "    X_train[n] = img\n",
    "    \n",
    "    #Read corresponding mask files iteratively\n",
    "    mask = np.zeros((IMG_HEIGHT, IMG_WIDTH, 1), dtype=np.bool)\n",
    "    \n",
    "    #Looping through masks\n",
    "    for mask_file in next(os.walk(path + '/masks/'))[2]:\n",
    "        \n",
    "        #Read individual masks\n",
    "        mask_ = imread(dir_path + path + '/masks/' + mask_file)\n",
    "        \n",
    "        #Expand individual mask dimensions\n",
    "        mask_ = np.expand_dims(resize(mask_, (IMG_HEIGHT, IMG_WIDTH), mode='constant', \n",
    "                                      preserve_range=True), axis=-1)\n",
    "        \n",
    "        #Overlay individual masks to create a final mask for corresponding image\n",
    "        mask = np.maximum(mask, mask_)\n",
    "    \n",
    "    #Append mask to numpy array for train dataset\n",
    "    Y_train[n] = mask\n",
    "\n",
    "# Get and resize test images\n",
    "X_test = np.zeros((len(test_ids), IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS), dtype=np.uint8)\n",
    "sizes_test = []\n",
    "print('Getting and resizing test images ... ')\n",
    "sys.stdout.flush()\n",
    "for n, id_ in tqdm(enumerate(test_ids), total=len(test_ids)):\n",
    "    path = TEST_PATH + id_\n",
    "    \n",
    "    #Read images iteratively\n",
    "    img = imread(dir_path + path + '/images/' + id_ + '.png')[:,:,:IMG_CHANNELS]\n",
    "    \n",
    "    #Get test size\n",
    "    sizes_test.append([img.shape[0], img.shape[1]])\n",
    "    \n",
    "    #Resize image to match training data\n",
    "    img = resize(img, (IMG_HEIGHT, IMG_WIDTH), mode='constant', preserve_range=True)\n",
    "    \n",
    "    #Append image to numpy array for test dataset\n",
    "    X_test[n] = img\n",
    "\n",
    "print('Done!')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "c0523b03-1fc5-4505-a1b8-eb35ee617c8a",
    "_uuid": "d4f8327802a1ec6139ce0585953986272ba62ce1"
   },
   "source": [
    "## Visualize imported data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "_cell_guid": "88829b53-50ce-45d9-9540-77dd7384ad4c",
    "_uuid": "283af26f0860b7069bdfd133c746e5d20971542c"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Check if training data looks all right\n",
    "ix = random.randint(0, len(train_ids))\n",
    "imshow(X_train[ix])\n",
    "plt.show()\n",
    "imshow(np.squeeze(Y_train[ix]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "c3b9f148-1dba-4b6a-981b-6cdbf394fc3c",
    "_uuid": "986488a4c5223576be370e224426a30431911eb2"
   },
   "source": [
    "# Build and train our neural network\n",
    "Next we build our U-Net model, loosely based on [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://arxiv.org/pdf/1505.04597.pdf) and very similar to [this repo](https://github.com/jocicmarko/ultrasound-nerve-segmentation) from the Kaggle Ultrasound Nerve Segmentation competition.\n",
    "\n",
    "![](https://lmb.informatik.uni-freiburg.de/people/ronneber/u-net/u-net-architecture.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "_cell_guid": "c1dbc57c-b497-4ccb-b077-2053203ab7ed",
    "_uuid": "0aa97d66c29f45dfac9b0f45fcf74ba0e778ba5d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_3\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_4 (InputLayer)            [(None, 128, 128, 3) 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "lambda_3 (Lambda)               (None, 128, 128, 3)  0           input_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_57 (Conv2D)              (None, 128, 128, 16) 448         lambda_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dropout_27 (Dropout)            (None, 128, 128, 16) 0           conv2d_57[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_58 (Conv2D)              (None, 128, 128, 16) 2320        dropout_27[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_12 (MaxPooling2D) (None, 64, 64, 16)   0           conv2d_58[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_59 (Conv2D)              (None, 64, 64, 32)   4640        max_pooling2d_12[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "dropout_28 (Dropout)            (None, 64, 64, 32)   0           conv2d_59[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_60 (Conv2D)              (None, 64, 64, 32)   9248        dropout_28[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_13 (MaxPooling2D) (None, 32, 32, 32)   0           conv2d_60[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_61 (Conv2D)              (None, 32, 32, 64)   18496       max_pooling2d_13[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "dropout_29 (Dropout)            (None, 32, 32, 64)   0           conv2d_61[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_62 (Conv2D)              (None, 32, 32, 64)   36928       dropout_29[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_14 (MaxPooling2D) (None, 16, 16, 64)   0           conv2d_62[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_63 (Conv2D)              (None, 16, 16, 128)  73856       max_pooling2d_14[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "dropout_30 (Dropout)            (None, 16, 16, 128)  0           conv2d_63[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_64 (Conv2D)              (None, 16, 16, 128)  147584      dropout_30[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_15 (MaxPooling2D) (None, 8, 8, 128)    0           conv2d_64[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_65 (Conv2D)              (None, 8, 8, 256)    295168      max_pooling2d_15[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "dropout_31 (Dropout)            (None, 8, 8, 256)    0           conv2d_65[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_66 (Conv2D)              (None, 8, 8, 256)    590080      dropout_31[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_12 (Conv2DTran (None, 16, 16, 128)  131200      conv2d_66[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_12 (Concatenate)    (None, 16, 16, 256)  0           conv2d_transpose_12[0][0]        \n",
      "                                                                 conv2d_64[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_67 (Conv2D)              (None, 16, 16, 128)  295040      concatenate_12[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "dropout_32 (Dropout)            (None, 16, 16, 128)  0           conv2d_67[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_68 (Conv2D)              (None, 16, 16, 128)  147584      dropout_32[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_13 (Conv2DTran (None, 32, 32, 64)   32832       conv2d_68[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_13 (Concatenate)    (None, 32, 32, 128)  0           conv2d_transpose_13[0][0]        \n",
      "                                                                 conv2d_62[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_69 (Conv2D)              (None, 32, 32, 64)   73792       concatenate_13[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "dropout_33 (Dropout)            (None, 32, 32, 64)   0           conv2d_69[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_70 (Conv2D)              (None, 32, 32, 64)   36928       dropout_33[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_14 (Conv2DTran (None, 64, 64, 32)   8224        conv2d_70[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_14 (Concatenate)    (None, 64, 64, 64)   0           conv2d_transpose_14[0][0]        \n",
      "                                                                 conv2d_60[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_71 (Conv2D)              (None, 64, 64, 32)   18464       concatenate_14[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "dropout_34 (Dropout)            (None, 64, 64, 32)   0           conv2d_71[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_72 (Conv2D)              (None, 64, 64, 32)   9248        dropout_34[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_transpose_15 (Conv2DTran (None, 128, 128, 16) 2064        conv2d_72[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_15 (Concatenate)    (None, 128, 128, 32) 0           conv2d_transpose_15[0][0]        \n",
      "                                                                 conv2d_58[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_73 (Conv2D)              (None, 128, 128, 16) 4624        concatenate_15[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "dropout_35 (Dropout)            (None, 128, 128, 16) 0           conv2d_73[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_74 (Conv2D)              (None, 128, 128, 16) 2320        dropout_35[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_75 (Conv2D)              (None, 128, 128, 1)  17          conv2d_74[0][0]                  \n",
      "==================================================================================================\n",
      "Total params: 1,941,105\n",
      "Trainable params: 1,941,105\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# Build U-Net model\n",
    "inputs = Input((IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS))\n",
    "s = Lambda(lambda x: x / 255) (inputs)\n",
    "\n",
    "c1 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (s)\n",
    "c1 = Dropout(0.1) (c1)\n",
    "c1 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c1)\n",
    "p1 = MaxPooling2D((2, 2)) (c1)\n",
    "\n",
    "c2 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p1)\n",
    "c2 = Dropout(0.1) (c2)\n",
    "c2 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c2)\n",
    "p2 = MaxPooling2D((2, 2)) (c2)\n",
    "\n",
    "c3 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p2)\n",
    "c3 = Dropout(0.2) (c3)\n",
    "c3 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c3)\n",
    "p3 = MaxPooling2D((2, 2)) (c3)\n",
    "\n",
    "c4 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p3)\n",
    "c4 = Dropout(0.2) (c4)\n",
    "c4 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c4)\n",
    "p4 = MaxPooling2D(pool_size=(2, 2)) (c4)\n",
    "\n",
    "c5 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (p4)\n",
    "c5 = Dropout(0.3) (c5)\n",
    "c5 = Conv2D(256, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c5)\n",
    "\n",
    "u6 = Conv2DTranspose(128, (2, 2), strides=(2, 2), padding='same') (c5)\n",
    "u6 = concatenate([u6, c4])\n",
    "c6 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u6)\n",
    "c6 = Dropout(0.2) (c6)\n",
    "c6 = Conv2D(128, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c6)\n",
    "\n",
    "u7 = Conv2DTranspose(64, (2, 2), strides=(2, 2), padding='same') (c6)\n",
    "u7 = concatenate([u7, c3])\n",
    "c7 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u7)\n",
    "c7 = Dropout(0.2) (c7)\n",
    "c7 = Conv2D(64, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c7)\n",
    "\n",
    "u8 = Conv2DTranspose(32, (2, 2), strides=(2, 2), padding='same') (c7)\n",
    "u8 = concatenate([u8, c2])\n",
    "c8 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u8)\n",
    "c8 = Dropout(0.1) (c8)\n",
    "c8 = Conv2D(32, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c8)\n",
    "\n",
    "u9 = Conv2DTranspose(16, (2, 2), strides=(2, 2), padding='same') (c8)\n",
    "u9 = concatenate([u9, c1], axis=3)\n",
    "c9 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (u9)\n",
    "c9 = Dropout(0.1) (c9)\n",
    "c9 = Conv2D(16, (3, 3), activation='elu', kernel_initializer='he_normal', padding='same') (c9)\n",
    "\n",
    "outputs = Conv2D(1, (1, 1), activation='sigmoid') (c9)\n",
    "\n",
    "model = Model(inputs=[inputs], outputs=[outputs])\n",
    "model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "_cell_guid": "9415b1c4-aa69-41b9-a1e3-d6053dbd4f64",
    "_uuid": "c060db22daa2abf12b28240cd81bbcbf1ce1bf87"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 603 samples, validate on 67 samples\n",
      "Epoch 1/50\n",
      "592/603 [============================>.] - ETA: 1:57 - loss: 0.7497 - accuracy: 0.46 - ETA: 58s - loss: 0.6919 - accuracy: 0.5690 - ETA: 38s - loss: 0.6792 - accuracy: 0.613 - ETA: 28s - loss: 0.6664 - accuracy: 0.642 - ETA: 22s - loss: 0.6336 - accuracy: 0.679 - ETA: 18s - loss: 0.6149 - accuracy: 0.698 - ETA: 16s - loss: 0.6086 - accuracy: 0.705 - ETA: 13s - loss: 0.5999 - accuracy: 0.715 - ETA: 12s - loss: 0.5874 - accuracy: 0.724 - ETA: 10s - loss: 0.5707 - accuracy: 0.735 - ETA: 9s - loss: 0.5624 - accuracy: 0.739 - ETA: 8s - loss: 0.5552 - accuracy: 0.74 - ETA: 7s - loss: 0.5520 - accuracy: 0.74 - ETA: 7s - loss: 0.5397 - accuracy: 0.75 - ETA: 6s - loss: 0.5308 - accuracy: 0.75 - ETA: 5s - loss: 0.5255 - accuracy: 0.75 - ETA: 5s - loss: 0.5159 - accuracy: 0.76 - ETA: 4s - loss: 0.5489 - accuracy: 0.74 - ETA: 4s - loss: 0.5441 - accuracy: 0.74 - ETA: 4s - loss: 0.5321 - accuracy: 0.75 - ETA: 3s - loss: 0.5282 - accuracy: 0.75 - ETA: 3s - loss: 0.5209 - accuracy: 0.75 - ETA: 3s - loss: 0.5180 - accuracy: 0.75 - ETA: 2s - loss: 0.5135 - accuracy: 0.75 - ETA: 2s - loss: 0.5099 - accuracy: 0.75 - ETA: 2s - loss: 0.5059 - accuracy: 0.76 - ETA: 2s - loss: 0.5039 - accuracy: 0.76 - ETA: 1s - loss: 0.4977 - accuracy: 0.76 - ETA: 1s - loss: 0.4916 - accuracy: 0.77 - ETA: 1s - loss: 0.4853 - accuracy: 0.77 - ETA: 1s - loss: 0.4791 - accuracy: 0.77 - ETA: 0s - loss: 0.4734 - accuracy: 0.77 - ETA: 0s - loss: 0.4675 - accuracy: 0.78 - ETA: 0s - loss: 0.4641 - accuracy: 0.78 - ETA: 0s - loss: 0.4586 - accuracy: 0.78 - ETA: 0s - loss: 0.4538 - accuracy: 0.79 - ETA: 0s - loss: 0.4482 - accuracy: 0.7943\n",
      "Epoch 00001: val_loss improved from inf to 0.22165, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 7s 12ms/sample - loss: 0.4464 - accuracy: 0.7951 - val_loss: 0.2217 - val_accuracy: 0.9041\n",
      "Epoch 2/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.2755 - accuracy: 0.89 - ETA: 2s - loss: 0.2883 - accuracy: 0.88 - ETA: 2s - loss: 0.2724 - accuracy: 0.89 - ETA: 2s - loss: 0.2778 - accuracy: 0.89 - ETA: 2s - loss: 0.2657 - accuracy: 0.89 - ETA: 2s - loss: 0.2560 - accuracy: 0.90 - ETA: 2s - loss: 0.2471 - accuracy: 0.90 - ETA: 2s - loss: 0.2419 - accuracy: 0.90 - ETA: 2s - loss: 0.2415 - accuracy: 0.90 - ETA: 2s - loss: 0.2416 - accuracy: 0.90 - ETA: 2s - loss: 0.2390 - accuracy: 0.90 - ETA: 1s - loss: 0.2426 - accuracy: 0.90 - ETA: 1s - loss: 0.2383 - accuracy: 0.90 - ETA: 1s - loss: 0.2379 - accuracy: 0.90 - ETA: 1s - loss: 0.2344 - accuracy: 0.90 - ETA: 1s - loss: 0.2310 - accuracy: 0.90 - ETA: 1s - loss: 0.2311 - accuracy: 0.90 - ETA: 1s - loss: 0.2292 - accuracy: 0.90 - ETA: 1s - loss: 0.2274 - accuracy: 0.91 - ETA: 1s - loss: 0.2296 - accuracy: 0.90 - ETA: 1s - loss: 0.2311 - accuracy: 0.90 - ETA: 1s - loss: 0.2281 - accuracy: 0.90 - ETA: 1s - loss: 0.2272 - accuracy: 0.91 - ETA: 1s - loss: 0.2276 - accuracy: 0.90 - ETA: 0s - loss: 0.2257 - accuracy: 0.91 - ETA: 0s - loss: 0.2229 - accuracy: 0.91 - ETA: 0s - loss: 0.2222 - accuracy: 0.91 - ETA: 0s - loss: 0.2213 - accuracy: 0.91 - ETA: 0s - loss: 0.2210 - accuracy: 0.91 - ETA: 0s - loss: 0.2192 - accuracy: 0.91 - ETA: 0s - loss: 0.2185 - accuracy: 0.91 - ETA: 0s - loss: 0.2193 - accuracy: 0.91 - ETA: 0s - loss: 0.2184 - accuracy: 0.91 - ETA: 0s - loss: 0.2180 - accuracy: 0.91 - ETA: 0s - loss: 0.2168 - accuracy: 0.91 - ETA: 0s - loss: 0.2144 - accuracy: 0.91 - ETA: 0s - loss: 0.2142 - accuracy: 0.9152\n",
      "Epoch 00002: val_loss improved from 0.22165 to 0.18199, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.2144 - accuracy: 0.9150 - val_loss: 0.1820 - val_accuracy: 0.9308\n",
      "Epoch 3/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1727 - accuracy: 0.93 - ETA: 2s - loss: 0.1541 - accuracy: 0.93 - ETA: 2s - loss: 0.1635 - accuracy: 0.93 - ETA: 2s - loss: 0.1707 - accuracy: 0.93 - ETA: 2s - loss: 0.1592 - accuracy: 0.93 - ETA: 2s - loss: 0.1598 - accuracy: 0.93 - ETA: 2s - loss: 0.1571 - accuracy: 0.93 - ETA: 2s - loss: 0.1625 - accuracy: 0.93 - ETA: 2s - loss: 0.1588 - accuracy: 0.93 - ETA: 2s - loss: 0.1619 - accuracy: 0.93 - ETA: 2s - loss: 0.1581 - accuracy: 0.93 - ETA: 1s - loss: 0.1556 - accuracy: 0.93 - ETA: 1s - loss: 0.1550 - accuracy: 0.93 - ETA: 1s - loss: 0.1536 - accuracy: 0.93 - ETA: 1s - loss: 0.1563 - accuracy: 0.93 - ETA: 1s - loss: 0.1598 - accuracy: 0.93 - ETA: 1s - loss: 0.1607 - accuracy: 0.93 - ETA: 1s - loss: 0.1624 - accuracy: 0.93 - ETA: 1s - loss: 0.1608 - accuracy: 0.93 - ETA: 1s - loss: 0.1598 - accuracy: 0.93 - ETA: 1s - loss: 0.1583 - accuracy: 0.93 - ETA: 1s - loss: 0.1565 - accuracy: 0.93 - ETA: 1s - loss: 0.1563 - accuracy: 0.93 - ETA: 1s - loss: 0.1567 - accuracy: 0.93 - ETA: 0s - loss: 0.1543 - accuracy: 0.93 - ETA: 0s - loss: 0.1576 - accuracy: 0.93 - ETA: 0s - loss: 0.1589 - accuracy: 0.93 - ETA: 0s - loss: 0.1583 - accuracy: 0.93 - ETA: 0s - loss: 0.1574 - accuracy: 0.93 - ETA: 0s - loss: 0.1563 - accuracy: 0.93 - ETA: 0s - loss: 0.1539 - accuracy: 0.93 - ETA: 0s - loss: 0.1547 - accuracy: 0.93 - ETA: 0s - loss: 0.1558 - accuracy: 0.93 - ETA: 0s - loss: 0.1572 - accuracy: 0.93 - ETA: 0s - loss: 0.1561 - accuracy: 0.93 - ETA: 0s - loss: 0.1559 - accuracy: 0.93 - ETA: 0s - loss: 0.1571 - accuracy: 0.9379\n",
      "Epoch 00003: val_loss improved from 0.18199 to 0.13388, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.1562 - accuracy: 0.9383 - val_loss: 0.1339 - val_accuracy: 0.9449\n",
      "Epoch 4/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1729 - accuracy: 0.92 - ETA: 2s - loss: 0.1518 - accuracy: 0.93 - ETA: 2s - loss: 0.1503 - accuracy: 0.93 - ETA: 2s - loss: 0.1463 - accuracy: 0.94 - ETA: 2s - loss: 0.1401 - accuracy: 0.94 - ETA: 2s - loss: 0.1486 - accuracy: 0.93 - ETA: 2s - loss: 0.1440 - accuracy: 0.94 - ETA: 2s - loss: 0.1390 - accuracy: 0.94 - ETA: 2s - loss: 0.1351 - accuracy: 0.94 - ETA: 2s - loss: 0.1385 - accuracy: 0.94 - ETA: 2s - loss: 0.1384 - accuracy: 0.94 - ETA: 1s - loss: 0.1378 - accuracy: 0.94 - ETA: 1s - loss: 0.1354 - accuracy: 0.94 - ETA: 1s - loss: 0.1346 - accuracy: 0.94 - ETA: 1s - loss: 0.1398 - accuracy: 0.94 - ETA: 1s - loss: 0.1396 - accuracy: 0.94 - ETA: 1s - loss: 0.1378 - accuracy: 0.94 - ETA: 1s - loss: 0.1404 - accuracy: 0.94 - ETA: 1s - loss: 0.1405 - accuracy: 0.94 - ETA: 1s - loss: 0.1383 - accuracy: 0.94 - ETA: 1s - loss: 0.1378 - accuracy: 0.94 - ETA: 1s - loss: 0.1376 - accuracy: 0.94 - ETA: 1s - loss: 0.1385 - accuracy: 0.94 - ETA: 1s - loss: 0.1393 - accuracy: 0.94 - ETA: 0s - loss: 0.1431 - accuracy: 0.94 - ETA: 0s - loss: 0.1427 - accuracy: 0.94 - ETA: 0s - loss: 0.1418 - accuracy: 0.94 - ETA: 0s - loss: 0.1424 - accuracy: 0.94 - ETA: 0s - loss: 0.1431 - accuracy: 0.94 - ETA: 0s - loss: 0.1439 - accuracy: 0.94 - ETA: 0s - loss: 0.1446 - accuracy: 0.94 - ETA: 0s - loss: 0.1454 - accuracy: 0.94 - ETA: 0s - loss: 0.1437 - accuracy: 0.94 - ETA: 0s - loss: 0.1426 - accuracy: 0.94 - ETA: 0s - loss: 0.1428 - accuracy: 0.94 - ETA: 0s - loss: 0.1431 - accuracy: 0.94 - ETA: 0s - loss: 0.1435 - accuracy: 0.9437\n",
      "Epoch 00004: val_loss did not improve from 0.13388\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.1424 - accuracy: 0.9442 - val_loss: 0.1418 - val_accuracy: 0.9447\n",
      "Epoch 5/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1095 - accuracy: 0.95 - ETA: 2s - loss: 0.1446 - accuracy: 0.94 - ETA: 2s - loss: 0.1716 - accuracy: 0.93 - ETA: 2s - loss: 0.1563 - accuracy: 0.94 - ETA: 2s - loss: 0.1539 - accuracy: 0.94 - ETA: 2s - loss: 0.1489 - accuracy: 0.94 - ETA: 2s - loss: 0.1419 - accuracy: 0.94 - ETA: 2s - loss: 0.1516 - accuracy: 0.94 - ETA: 2s - loss: 0.1489 - accuracy: 0.94 - ETA: 2s - loss: 0.1432 - accuracy: 0.94 - ETA: 2s - loss: 0.1411 - accuracy: 0.94 - ETA: 1s - loss: 0.1380 - accuracy: 0.94 - ETA: 1s - loss: 0.1358 - accuracy: 0.94 - ETA: 1s - loss: 0.1362 - accuracy: 0.94 - ETA: 1s - loss: 0.1374 - accuracy: 0.94 - ETA: 1s - loss: 0.1386 - accuracy: 0.94 - ETA: 1s - loss: 0.1361 - accuracy: 0.94 - ETA: 1s - loss: 0.1370 - accuracy: 0.94 - ETA: 1s - loss: 0.1341 - accuracy: 0.94 - ETA: 1s - loss: 0.1351 - accuracy: 0.94 - ETA: 1s - loss: 0.1349 - accuracy: 0.94 - ETA: 1s - loss: 0.1366 - accuracy: 0.94 - ETA: 1s - loss: 0.1367 - accuracy: 0.94 - ETA: 1s - loss: 0.1381 - accuracy: 0.94 - ETA: 0s - loss: 0.1387 - accuracy: 0.94 - ETA: 0s - loss: 0.1377 - accuracy: 0.94 - ETA: 0s - loss: 0.1373 - accuracy: 0.94 - ETA: 0s - loss: 0.1365 - accuracy: 0.94 - ETA: 0s - loss: 0.1376 - accuracy: 0.94 - ETA: 0s - loss: 0.1389 - accuracy: 0.94 - ETA: 0s - loss: 0.1384 - accuracy: 0.94 - ETA: 0s - loss: 0.1368 - accuracy: 0.94 - ETA: 0s - loss: 0.1349 - accuracy: 0.94 - ETA: 0s - loss: 0.1334 - accuracy: 0.94 - ETA: 0s - loss: 0.1335 - accuracy: 0.94 - ETA: 0s - loss: 0.1337 - accuracy: 0.94 - ETA: 0s - loss: 0.1329 - accuracy: 0.9493\n",
      "Epoch 00005: val_loss improved from 0.13388 to 0.11671, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.1318 - accuracy: 0.9498 - val_loss: 0.1167 - val_accuracy: 0.9538\n",
      "Epoch 6/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1216 - accuracy: 0.95 - ETA: 2s - loss: 0.1225 - accuracy: 0.95 - ETA: 2s - loss: 0.1134 - accuracy: 0.95 - ETA: 2s - loss: 0.1249 - accuracy: 0.95 - ETA: 2s - loss: 0.1217 - accuracy: 0.95 - ETA: 2s - loss: 0.1240 - accuracy: 0.95 - ETA: 2s - loss: 0.1263 - accuracy: 0.95 - ETA: 2s - loss: 0.1213 - accuracy: 0.95 - ETA: 2s - loss: 0.1238 - accuracy: 0.95 - ETA: 2s - loss: 0.1250 - accuracy: 0.95 - ETA: 2s - loss: 0.1216 - accuracy: 0.95 - ETA: 1s - loss: 0.1240 - accuracy: 0.95 - ETA: 1s - loss: 0.1220 - accuracy: 0.95 - ETA: 1s - loss: 0.1238 - accuracy: 0.95 - ETA: 1s - loss: 0.1241 - accuracy: 0.95 - ETA: 1s - loss: 0.1244 - accuracy: 0.95 - ETA: 1s - loss: 0.1252 - accuracy: 0.95 - ETA: 1s - loss: 0.1237 - accuracy: 0.95 - ETA: 1s - loss: 0.1251 - accuracy: 0.95 - ETA: 1s - loss: 0.1281 - accuracy: 0.95 - ETA: 1s - loss: 0.1296 - accuracy: 0.95 - ETA: 1s - loss: 0.1285 - accuracy: 0.95 - ETA: 1s - loss: 0.1274 - accuracy: 0.95 - ETA: 1s - loss: 0.1266 - accuracy: 0.95 - ETA: 0s - loss: 0.1247 - accuracy: 0.95 - ETA: 0s - loss: 0.1243 - accuracy: 0.95 - ETA: 0s - loss: 0.1257 - accuracy: 0.95 - ETA: 0s - loss: 0.1259 - accuracy: 0.95 - ETA: 0s - loss: 0.1255 - accuracy: 0.95 - ETA: 0s - loss: 0.1249 - accuracy: 0.95 - ETA: 0s - loss: 0.1243 - accuracy: 0.95 - ETA: 0s - loss: 0.1239 - accuracy: 0.95 - ETA: 0s - loss: 0.1233 - accuracy: 0.95 - ETA: 0s - loss: 0.1234 - accuracy: 0.95 - ETA: 0s - loss: 0.1245 - accuracy: 0.95 - ETA: 0s - loss: 0.1239 - accuracy: 0.95 - ETA: 0s - loss: 0.1239 - accuracy: 0.9533\n",
      "Epoch 00006: val_loss did not improve from 0.11671\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.1235 - accuracy: 0.9535 - val_loss: 0.1474 - val_accuracy: 0.9467\n",
      "Epoch 7/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1605 - accuracy: 0.93 - ETA: 2s - loss: 0.1416 - accuracy: 0.94 - ETA: 2s - loss: 0.1532 - accuracy: 0.93 - ETA: 2s - loss: 0.1336 - accuracy: 0.94 - ETA: 2s - loss: 0.1414 - accuracy: 0.94 - ETA: 2s - loss: 0.1312 - accuracy: 0.94 - ETA: 2s - loss: 0.1348 - accuracy: 0.94 - ETA: 2s - loss: 0.1355 - accuracy: 0.94 - ETA: 2s - loss: 0.1361 - accuracy: 0.94 - ETA: 2s - loss: 0.1344 - accuracy: 0.94 - ETA: 2s - loss: 0.1313 - accuracy: 0.94 - ETA: 1s - loss: 0.1293 - accuracy: 0.95 - ETA: 1s - loss: 0.1293 - accuracy: 0.95 - ETA: 1s - loss: 0.1276 - accuracy: 0.95 - ETA: 1s - loss: 0.1281 - accuracy: 0.95 - ETA: 1s - loss: 0.1261 - accuracy: 0.95 - ETA: 1s - loss: 0.1241 - accuracy: 0.95 - ETA: 1s - loss: 0.1249 - accuracy: 0.95 - ETA: 1s - loss: 0.1244 - accuracy: 0.95 - ETA: 1s - loss: 0.1238 - accuracy: 0.95 - ETA: 1s - loss: 0.1231 - accuracy: 0.95 - ETA: 1s - loss: 0.1214 - accuracy: 0.95 - ETA: 1s - loss: 0.1212 - accuracy: 0.95 - ETA: 1s - loss: 0.1204 - accuracy: 0.95 - ETA: 0s - loss: 0.1186 - accuracy: 0.95 - ETA: 0s - loss: 0.1173 - accuracy: 0.95 - ETA: 0s - loss: 0.1182 - accuracy: 0.95 - ETA: 0s - loss: 0.1174 - accuracy: 0.95 - ETA: 0s - loss: 0.1183 - accuracy: 0.95 - ETA: 0s - loss: 0.1182 - accuracy: 0.95 - ETA: 0s - loss: 0.1190 - accuracy: 0.95 - ETA: 0s - loss: 0.1199 - accuracy: 0.95 - ETA: 0s - loss: 0.1197 - accuracy: 0.95 - ETA: 0s - loss: 0.1191 - accuracy: 0.95 - ETA: 0s - loss: 0.1183 - accuracy: 0.95 - ETA: 0s - loss: 0.1189 - accuracy: 0.95 - ETA: 0s - loss: 0.1179 - accuracy: 0.9551\n",
      "Epoch 00007: val_loss improved from 0.11671 to 0.11464, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.1183 - accuracy: 0.9549 - val_loss: 0.1146 - val_accuracy: 0.9544\n",
      "Epoch 8/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0964 - accuracy: 0.96 - ETA: 2s - loss: 0.1191 - accuracy: 0.95 - ETA: 2s - loss: 0.1217 - accuracy: 0.95 - ETA: 2s - loss: 0.1324 - accuracy: 0.94 - ETA: 2s - loss: 0.1329 - accuracy: 0.94 - ETA: 2s - loss: 0.1266 - accuracy: 0.95 - ETA: 2s - loss: 0.1250 - accuracy: 0.95 - ETA: 2s - loss: 0.1156 - accuracy: 0.95 - ETA: 2s - loss: 0.1164 - accuracy: 0.95 - ETA: 2s - loss: 0.1183 - accuracy: 0.95 - ETA: 2s - loss: 0.1171 - accuracy: 0.95 - ETA: 1s - loss: 0.1166 - accuracy: 0.95 - ETA: 1s - loss: 0.1182 - accuracy: 0.95 - ETA: 1s - loss: 0.1171 - accuracy: 0.95 - ETA: 1s - loss: 0.1174 - accuracy: 0.95 - ETA: 1s - loss: 0.1173 - accuracy: 0.95 - ETA: 1s - loss: 0.1154 - accuracy: 0.95 - ETA: 1s - loss: 0.1146 - accuracy: 0.95 - ETA: 1s - loss: 0.1163 - accuracy: 0.95 - ETA: 1s - loss: 0.1151 - accuracy: 0.95 - ETA: 1s - loss: 0.1172 - accuracy: 0.95 - ETA: 1s - loss: 0.1170 - accuracy: 0.95 - ETA: 1s - loss: 0.1154 - accuracy: 0.95 - ETA: 1s - loss: 0.1159 - accuracy: 0.95 - ETA: 0s - loss: 0.1154 - accuracy: 0.95 - ETA: 0s - loss: 0.1150 - accuracy: 0.95 - ETA: 0s - loss: 0.1136 - accuracy: 0.95 - ETA: 0s - loss: 0.1143 - accuracy: 0.95 - ETA: 0s - loss: 0.1137 - accuracy: 0.95 - ETA: 0s - loss: 0.1133 - accuracy: 0.95 - ETA: 0s - loss: 0.1137 - accuracy: 0.95 - ETA: 0s - loss: 0.1138 - accuracy: 0.95 - ETA: 0s - loss: 0.1140 - accuracy: 0.95 - ETA: 0s - loss: 0.1148 - accuracy: 0.95 - ETA: 0s - loss: 0.1164 - accuracy: 0.95 - ETA: 0s - loss: 0.1154 - accuracy: 0.95 - ETA: 0s - loss: 0.1150 - accuracy: 0.9564\n",
      "Epoch 00008: val_loss improved from 0.11464 to 0.10729, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.1145 - accuracy: 0.9567 - val_loss: 0.1073 - val_accuracy: 0.9589\n",
      "Epoch 9/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "592/603 [============================>.] - ETA: 2s - loss: 0.1406 - accuracy: 0.95 - ETA: 2s - loss: 0.1143 - accuracy: 0.96 - ETA: 2s - loss: 0.1130 - accuracy: 0.95 - ETA: 2s - loss: 0.1059 - accuracy: 0.96 - ETA: 2s - loss: 0.1035 - accuracy: 0.96 - ETA: 2s - loss: 0.1072 - accuracy: 0.96 - ETA: 2s - loss: 0.1141 - accuracy: 0.95 - ETA: 2s - loss: 0.1134 - accuracy: 0.95 - ETA: 2s - loss: 0.1136 - accuracy: 0.95 - ETA: 2s - loss: 0.1108 - accuracy: 0.95 - ETA: 2s - loss: 0.1116 - accuracy: 0.95 - ETA: 1s - loss: 0.1086 - accuracy: 0.95 - ETA: 1s - loss: 0.1059 - accuracy: 0.95 - ETA: 1s - loss: 0.1056 - accuracy: 0.95 - ETA: 1s - loss: 0.1054 - accuracy: 0.95 - ETA: 1s - loss: 0.1044 - accuracy: 0.96 - ETA: 1s - loss: 0.1030 - accuracy: 0.96 - ETA: 1s - loss: 0.1031 - accuracy: 0.96 - ETA: 1s - loss: 0.1030 - accuracy: 0.96 - ETA: 1s - loss: 0.1036 - accuracy: 0.96 - ETA: 1s - loss: 0.1032 - accuracy: 0.96 - ETA: 1s - loss: 0.1038 - accuracy: 0.96 - ETA: 1s - loss: 0.1050 - accuracy: 0.95 - ETA: 1s - loss: 0.1040 - accuracy: 0.96 - ETA: 0s - loss: 0.1060 - accuracy: 0.95 - ETA: 0s - loss: 0.1056 - accuracy: 0.95 - ETA: 0s - loss: 0.1039 - accuracy: 0.96 - ETA: 0s - loss: 0.1054 - accuracy: 0.95 - ETA: 0s - loss: 0.1038 - accuracy: 0.96 - ETA: 0s - loss: 0.1026 - accuracy: 0.96 - ETA: 0s - loss: 0.1036 - accuracy: 0.96 - ETA: 0s - loss: 0.1030 - accuracy: 0.96 - ETA: 0s - loss: 0.1027 - accuracy: 0.96 - ETA: 0s - loss: 0.1037 - accuracy: 0.96 - ETA: 0s - loss: 0.1056 - accuracy: 0.95 - ETA: 0s - loss: 0.1056 - accuracy: 0.95 - ETA: 0s - loss: 0.1047 - accuracy: 0.9601\n",
      "Epoch 00009: val_loss improved from 0.10729 to 0.09647, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.1040 - accuracy: 0.9603 - val_loss: 0.0965 - val_accuracy: 0.9625\n",
      "Epoch 10/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1100 - accuracy: 0.95 - ETA: 2s - loss: 0.1025 - accuracy: 0.95 - ETA: 2s - loss: 0.0985 - accuracy: 0.96 - ETA: 2s - loss: 0.1007 - accuracy: 0.96 - ETA: 2s - loss: 0.0976 - accuracy: 0.96 - ETA: 2s - loss: 0.0961 - accuracy: 0.96 - ETA: 2s - loss: 0.0976 - accuracy: 0.96 - ETA: 2s - loss: 0.0965 - accuracy: 0.96 - ETA: 2s - loss: 0.0964 - accuracy: 0.96 - ETA: 2s - loss: 0.0965 - accuracy: 0.96 - ETA: 2s - loss: 0.0968 - accuracy: 0.96 - ETA: 1s - loss: 0.0947 - accuracy: 0.96 - ETA: 1s - loss: 0.0948 - accuracy: 0.96 - ETA: 1s - loss: 0.1005 - accuracy: 0.96 - ETA: 1s - loss: 0.1007 - accuracy: 0.96 - ETA: 1s - loss: 0.1017 - accuracy: 0.96 - ETA: 1s - loss: 0.1031 - accuracy: 0.96 - ETA: 1s - loss: 0.1023 - accuracy: 0.96 - ETA: 1s - loss: 0.1042 - accuracy: 0.96 - ETA: 1s - loss: 0.1031 - accuracy: 0.96 - ETA: 1s - loss: 0.1023 - accuracy: 0.96 - ETA: 1s - loss: 0.1029 - accuracy: 0.96 - ETA: 1s - loss: 0.1029 - accuracy: 0.96 - ETA: 1s - loss: 0.1027 - accuracy: 0.96 - ETA: 0s - loss: 0.1021 - accuracy: 0.96 - ETA: 0s - loss: 0.1021 - accuracy: 0.96 - ETA: 0s - loss: 0.1016 - accuracy: 0.96 - ETA: 0s - loss: 0.1018 - accuracy: 0.96 - ETA: 0s - loss: 0.1011 - accuracy: 0.96 - ETA: 0s - loss: 0.1008 - accuracy: 0.96 - ETA: 0s - loss: 0.1015 - accuracy: 0.96 - ETA: 0s - loss: 0.1010 - accuracy: 0.96 - ETA: 0s - loss: 0.1011 - accuracy: 0.96 - ETA: 0s - loss: 0.1006 - accuracy: 0.96 - ETA: 0s - loss: 0.0999 - accuracy: 0.96 - ETA: 0s - loss: 0.0990 - accuracy: 0.96 - ETA: 0s - loss: 0.0993 - accuracy: 0.9615\n",
      "Epoch 00010: val_loss improved from 0.09647 to 0.09492, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0984 - accuracy: 0.9618 - val_loss: 0.0949 - val_accuracy: 0.9633\n",
      "Epoch 11/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1305 - accuracy: 0.95 - ETA: 2s - loss: 0.1203 - accuracy: 0.95 - ETA: 2s - loss: 0.1157 - accuracy: 0.95 - ETA: 2s - loss: 0.1072 - accuracy: 0.95 - ETA: 2s - loss: 0.1058 - accuracy: 0.95 - ETA: 2s - loss: 0.1018 - accuracy: 0.96 - ETA: 2s - loss: 0.0999 - accuracy: 0.96 - ETA: 2s - loss: 0.1022 - accuracy: 0.96 - ETA: 2s - loss: 0.1016 - accuracy: 0.96 - ETA: 2s - loss: 0.0962 - accuracy: 0.96 - ETA: 2s - loss: 0.1022 - accuracy: 0.96 - ETA: 1s - loss: 0.1008 - accuracy: 0.96 - ETA: 1s - loss: 0.1016 - accuracy: 0.96 - ETA: 1s - loss: 0.0997 - accuracy: 0.96 - ETA: 1s - loss: 0.0980 - accuracy: 0.96 - ETA: 1s - loss: 0.0994 - accuracy: 0.96 - ETA: 1s - loss: 0.0987 - accuracy: 0.96 - ETA: 1s - loss: 0.0992 - accuracy: 0.96 - ETA: 1s - loss: 0.0988 - accuracy: 0.96 - ETA: 1s - loss: 0.0994 - accuracy: 0.96 - ETA: 1s - loss: 0.1011 - accuracy: 0.96 - ETA: 1s - loss: 0.1009 - accuracy: 0.96 - ETA: 1s - loss: 0.1017 - accuracy: 0.96 - ETA: 1s - loss: 0.1024 - accuracy: 0.96 - ETA: 0s - loss: 0.1020 - accuracy: 0.96 - ETA: 0s - loss: 0.1015 - accuracy: 0.96 - ETA: 0s - loss: 0.1013 - accuracy: 0.96 - ETA: 0s - loss: 0.0999 - accuracy: 0.96 - ETA: 0s - loss: 0.0989 - accuracy: 0.96 - ETA: 0s - loss: 0.0990 - accuracy: 0.96 - ETA: 0s - loss: 0.0987 - accuracy: 0.96 - ETA: 0s - loss: 0.0985 - accuracy: 0.96 - ETA: 0s - loss: 0.0977 - accuracy: 0.96 - ETA: 0s - loss: 0.0966 - accuracy: 0.96 - ETA: 0s - loss: 0.0955 - accuracy: 0.96 - ETA: 0s - loss: 0.0952 - accuracy: 0.96 - ETA: 0s - loss: 0.0956 - accuracy: 0.9629\n",
      "Epoch 00011: val_loss improved from 0.09492 to 0.08880, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0953 - accuracy: 0.9630 - val_loss: 0.0888 - val_accuracy: 0.9660\n",
      "Epoch 12/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0846 - accuracy: 0.96 - ETA: 2s - loss: 0.1023 - accuracy: 0.95 - ETA: 2s - loss: 0.0934 - accuracy: 0.96 - ETA: 2s - loss: 0.0945 - accuracy: 0.96 - ETA: 2s - loss: 0.0957 - accuracy: 0.96 - ETA: 2s - loss: 0.0954 - accuracy: 0.96 - ETA: 2s - loss: 0.0935 - accuracy: 0.96 - ETA: 2s - loss: 0.0922 - accuracy: 0.96 - ETA: 2s - loss: 0.0932 - accuracy: 0.96 - ETA: 2s - loss: 0.0949 - accuracy: 0.96 - ETA: 2s - loss: 0.0931 - accuracy: 0.96 - ETA: 1s - loss: 0.0916 - accuracy: 0.96 - ETA: 1s - loss: 0.0923 - accuracy: 0.96 - ETA: 1s - loss: 0.0908 - accuracy: 0.96 - ETA: 1s - loss: 0.0928 - accuracy: 0.96 - ETA: 1s - loss: 0.0953 - accuracy: 0.96 - ETA: 1s - loss: 0.0944 - accuracy: 0.96 - ETA: 1s - loss: 0.0924 - accuracy: 0.96 - ETA: 1s - loss: 0.0942 - accuracy: 0.96 - ETA: 1s - loss: 0.0931 - accuracy: 0.96 - ETA: 1s - loss: 0.0928 - accuracy: 0.96 - ETA: 1s - loss: 0.0912 - accuracy: 0.96 - ETA: 1s - loss: 0.0904 - accuracy: 0.96 - ETA: 1s - loss: 0.0913 - accuracy: 0.96 - ETA: 0s - loss: 0.0910 - accuracy: 0.96 - ETA: 0s - loss: 0.0896 - accuracy: 0.96 - ETA: 0s - loss: 0.0901 - accuracy: 0.96 - ETA: 0s - loss: 0.0911 - accuracy: 0.96 - ETA: 0s - loss: 0.0906 - accuracy: 0.96 - ETA: 0s - loss: 0.0901 - accuracy: 0.96 - ETA: 0s - loss: 0.0909 - accuracy: 0.96 - ETA: 0s - loss: 0.0914 - accuracy: 0.96 - ETA: 0s - loss: 0.0906 - accuracy: 0.96 - ETA: 0s - loss: 0.0921 - accuracy: 0.96 - ETA: 0s - loss: 0.0929 - accuracy: 0.96 - ETA: 0s - loss: 0.0930 - accuracy: 0.96 - ETA: 0s - loss: 0.0932 - accuracy: 0.9640\n",
      "Epoch 00012: val_loss improved from 0.08880 to 0.08646, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0934 - accuracy: 0.9639 - val_loss: 0.0865 - val_accuracy: 0.9660\n",
      "Epoch 13/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1247 - accuracy: 0.95 - ETA: 2s - loss: 0.1011 - accuracy: 0.96 - ETA: 2s - loss: 0.0981 - accuracy: 0.96 - ETA: 2s - loss: 0.1063 - accuracy: 0.96 - ETA: 2s - loss: 0.1025 - accuracy: 0.96 - ETA: 2s - loss: 0.0947 - accuracy: 0.96 - ETA: 2s - loss: 0.0928 - accuracy: 0.96 - ETA: 2s - loss: 0.0917 - accuracy: 0.96 - ETA: 2s - loss: 0.0948 - accuracy: 0.96 - ETA: 2s - loss: 0.0922 - accuracy: 0.96 - ETA: 2s - loss: 0.0937 - accuracy: 0.96 - ETA: 1s - loss: 0.0921 - accuracy: 0.96 - ETA: 1s - loss: 0.0941 - accuracy: 0.96 - ETA: 1s - loss: 0.0945 - accuracy: 0.96 - ETA: 1s - loss: 0.0966 - accuracy: 0.96 - ETA: 1s - loss: 0.0958 - accuracy: 0.96 - ETA: 1s - loss: 0.0947 - accuracy: 0.96 - ETA: 1s - loss: 0.0926 - accuracy: 0.96 - ETA: 1s - loss: 0.0916 - accuracy: 0.96 - ETA: 1s - loss: 0.0895 - accuracy: 0.96 - ETA: 1s - loss: 0.0900 - accuracy: 0.96 - ETA: 1s - loss: 0.0896 - accuracy: 0.96 - ETA: 1s - loss: 0.0908 - accuracy: 0.96 - ETA: 1s - loss: 0.0902 - accuracy: 0.96 - ETA: 0s - loss: 0.0914 - accuracy: 0.96 - ETA: 0s - loss: 0.0901 - accuracy: 0.96 - ETA: 0s - loss: 0.0902 - accuracy: 0.96 - ETA: 0s - loss: 0.0897 - accuracy: 0.96 - ETA: 0s - loss: 0.0909 - accuracy: 0.96 - ETA: 0s - loss: 0.0916 - accuracy: 0.96 - ETA: 0s - loss: 0.0918 - accuracy: 0.96 - ETA: 0s - loss: 0.0911 - accuracy: 0.96 - ETA: 0s - loss: 0.0918 - accuracy: 0.96 - ETA: 0s - loss: 0.0931 - accuracy: 0.96 - ETA: 0s - loss: 0.0923 - accuracy: 0.96 - ETA: 0s - loss: 0.0926 - accuracy: 0.96 - ETA: 0s - loss: 0.0946 - accuracy: 0.9633\n",
      "Epoch 00013: val_loss improved from 0.08646 to 0.08643, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0946 - accuracy: 0.9632 - val_loss: 0.0864 - val_accuracy: 0.9660\n",
      "Epoch 14/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0692 - accuracy: 0.97 - ETA: 2s - loss: 0.0782 - accuracy: 0.97 - ETA: 2s - loss: 0.0847 - accuracy: 0.96 - ETA: 2s - loss: 0.0830 - accuracy: 0.96 - ETA: 2s - loss: 0.0889 - accuracy: 0.96 - ETA: 2s - loss: 0.0902 - accuracy: 0.96 - ETA: 2s - loss: 0.0931 - accuracy: 0.96 - ETA: 2s - loss: 0.0896 - accuracy: 0.96 - ETA: 2s - loss: 0.0951 - accuracy: 0.96 - ETA: 2s - loss: 0.0982 - accuracy: 0.96 - ETA: 2s - loss: 0.0982 - accuracy: 0.96 - ETA: 1s - loss: 0.0986 - accuracy: 0.96 - ETA: 1s - loss: 0.0987 - accuracy: 0.96 - ETA: 1s - loss: 0.0986 - accuracy: 0.96 - ETA: 1s - loss: 0.0995 - accuracy: 0.96 - ETA: 1s - loss: 0.0996 - accuracy: 0.96 - ETA: 1s - loss: 0.1000 - accuracy: 0.96 - ETA: 1s - loss: 0.0993 - accuracy: 0.96 - ETA: 1s - loss: 0.0978 - accuracy: 0.96 - ETA: 1s - loss: 0.0972 - accuracy: 0.96 - ETA: 1s - loss: 0.0954 - accuracy: 0.96 - ETA: 1s - loss: 0.0955 - accuracy: 0.96 - ETA: 1s - loss: 0.0944 - accuracy: 0.96 - ETA: 1s - loss: 0.0945 - accuracy: 0.96 - ETA: 0s - loss: 0.0943 - accuracy: 0.96 - ETA: 0s - loss: 0.0938 - accuracy: 0.96 - ETA: 0s - loss: 0.0934 - accuracy: 0.96 - ETA: 0s - loss: 0.0927 - accuracy: 0.96 - ETA: 0s - loss: 0.0921 - accuracy: 0.96 - ETA: 0s - loss: 0.0920 - accuracy: 0.96 - ETA: 0s - loss: 0.0917 - accuracy: 0.96 - ETA: 0s - loss: 0.0910 - accuracy: 0.96 - ETA: 0s - loss: 0.0911 - accuracy: 0.96 - ETA: 0s - loss: 0.0908 - accuracy: 0.96 - ETA: 0s - loss: 0.0919 - accuracy: 0.96 - ETA: 0s - loss: 0.0921 - accuracy: 0.96 - ETA: 0s - loss: 0.0920 - accuracy: 0.9641\n",
      "Epoch 00014: val_loss improved from 0.08643 to 0.08531, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0920 - accuracy: 0.9642 - val_loss: 0.0853 - val_accuracy: 0.9667\n",
      "Epoch 15/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0642 - accuracy: 0.97 - ETA: 2s - loss: 0.0711 - accuracy: 0.97 - ETA: 2s - loss: 0.0817 - accuracy: 0.96 - ETA: 2s - loss: 0.0766 - accuracy: 0.97 - ETA: 2s - loss: 0.0759 - accuracy: 0.97 - ETA: 2s - loss: 0.0782 - accuracy: 0.96 - ETA: 2s - loss: 0.0789 - accuracy: 0.96 - ETA: 2s - loss: 0.0805 - accuracy: 0.96 - ETA: 2s - loss: 0.0892 - accuracy: 0.96 - ETA: 2s - loss: 0.0874 - accuracy: 0.96 - ETA: 2s - loss: 0.0878 - accuracy: 0.96 - ETA: 1s - loss: 0.0890 - accuracy: 0.96 - ETA: 1s - loss: 0.0897 - accuracy: 0.96 - ETA: 1s - loss: 0.0889 - accuracy: 0.96 - ETA: 1s - loss: 0.0876 - accuracy: 0.96 - ETA: 1s - loss: 0.0878 - accuracy: 0.96 - ETA: 1s - loss: 0.0889 - accuracy: 0.96 - ETA: 1s - loss: 0.0896 - accuracy: 0.96 - ETA: 1s - loss: 0.0874 - accuracy: 0.96 - ETA: 1s - loss: 0.0884 - accuracy: 0.96 - ETA: 1s - loss: 0.0875 - accuracy: 0.96 - ETA: 1s - loss: 0.0881 - accuracy: 0.96 - ETA: 1s - loss: 0.0877 - accuracy: 0.96 - ETA: 1s - loss: 0.0880 - accuracy: 0.96 - ETA: 0s - loss: 0.0889 - accuracy: 0.96 - ETA: 0s - loss: 0.0890 - accuracy: 0.96 - ETA: 0s - loss: 0.0882 - accuracy: 0.96 - ETA: 0s - loss: 0.0873 - accuracy: 0.96 - ETA: 0s - loss: 0.0870 - accuracy: 0.96 - ETA: 0s - loss: 0.0864 - accuracy: 0.96 - ETA: 0s - loss: 0.0868 - accuracy: 0.96 - ETA: 0s - loss: 0.0876 - accuracy: 0.96 - ETA: 0s - loss: 0.0890 - accuracy: 0.96 - ETA: 0s - loss: 0.0888 - accuracy: 0.96 - ETA: 0s - loss: 0.0892 - accuracy: 0.96 - ETA: 0s - loss: 0.0879 - accuracy: 0.96 - ETA: 0s - loss: 0.0891 - accuracy: 0.9652\n",
      "Epoch 00015: val_loss improved from 0.08531 to 0.08236, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0887 - accuracy: 0.9654 - val_loss: 0.0824 - val_accuracy: 0.9681\n",
      "Epoch 16/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0856 - accuracy: 0.96 - ETA: 2s - loss: 0.0865 - accuracy: 0.96 - ETA: 2s - loss: 0.0804 - accuracy: 0.96 - ETA: 2s - loss: 0.0920 - accuracy: 0.96 - ETA: 2s - loss: 0.0968 - accuracy: 0.96 - ETA: 2s - loss: 0.0957 - accuracy: 0.96 - ETA: 2s - loss: 0.0963 - accuracy: 0.96 - ETA: 2s - loss: 0.0961 - accuracy: 0.96 - ETA: 2s - loss: 0.0952 - accuracy: 0.96 - ETA: 2s - loss: 0.0958 - accuracy: 0.96 - ETA: 2s - loss: 0.0935 - accuracy: 0.96 - ETA: 1s - loss: 0.0924 - accuracy: 0.96 - ETA: 1s - loss: 0.0929 - accuracy: 0.96 - ETA: 1s - loss: 0.0936 - accuracy: 0.96 - ETA: 1s - loss: 0.0925 - accuracy: 0.96 - ETA: 1s - loss: 0.0932 - accuracy: 0.96 - ETA: 1s - loss: 0.0923 - accuracy: 0.96 - ETA: 1s - loss: 0.0915 - accuracy: 0.96 - ETA: 1s - loss: 0.0917 - accuracy: 0.96 - ETA: 1s - loss: 0.0904 - accuracy: 0.96 - ETA: 1s - loss: 0.0899 - accuracy: 0.96 - ETA: 1s - loss: 0.0898 - accuracy: 0.96 - ETA: 1s - loss: 0.0907 - accuracy: 0.96 - ETA: 1s - loss: 0.0897 - accuracy: 0.96 - ETA: 0s - loss: 0.0899 - accuracy: 0.96 - ETA: 0s - loss: 0.0885 - accuracy: 0.96 - ETA: 0s - loss: 0.0878 - accuracy: 0.96 - ETA: 0s - loss: 0.0888 - accuracy: 0.96 - ETA: 0s - loss: 0.0880 - accuracy: 0.96 - ETA: 0s - loss: 0.0877 - accuracy: 0.96 - ETA: 0s - loss: 0.0867 - accuracy: 0.96 - ETA: 0s - loss: 0.0872 - accuracy: 0.96 - ETA: 0s - loss: 0.0881 - accuracy: 0.96 - ETA: 0s - loss: 0.0880 - accuracy: 0.96 - ETA: 0s - loss: 0.0883 - accuracy: 0.96 - ETA: 0s - loss: 0.0880 - accuracy: 0.96 - ETA: 0s - loss: 0.0884 - accuracy: 0.9655\n",
      "Epoch 00016: val_loss improved from 0.08236 to 0.07772, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0884 - accuracy: 0.9655 - val_loss: 0.0777 - val_accuracy: 0.9686\n",
      "Epoch 17/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "592/603 [============================>.] - ETA: 2s - loss: 0.0806 - accuracy: 0.96 - ETA: 2s - loss: 0.0884 - accuracy: 0.96 - ETA: 2s - loss: 0.0867 - accuracy: 0.96 - ETA: 2s - loss: 0.0889 - accuracy: 0.96 - ETA: 2s - loss: 0.0822 - accuracy: 0.96 - ETA: 2s - loss: 0.0800 - accuracy: 0.96 - ETA: 2s - loss: 0.0792 - accuracy: 0.96 - ETA: 2s - loss: 0.0790 - accuracy: 0.96 - ETA: 2s - loss: 0.0879 - accuracy: 0.96 - ETA: 2s - loss: 0.0873 - accuracy: 0.96 - ETA: 2s - loss: 0.0880 - accuracy: 0.96 - ETA: 1s - loss: 0.0885 - accuracy: 0.96 - ETA: 1s - loss: 0.0908 - accuracy: 0.96 - ETA: 1s - loss: 0.0892 - accuracy: 0.96 - ETA: 1s - loss: 0.0918 - accuracy: 0.96 - ETA: 1s - loss: 0.0927 - accuracy: 0.96 - ETA: 1s - loss: 0.0932 - accuracy: 0.96 - ETA: 1s - loss: 0.0927 - accuracy: 0.96 - ETA: 1s - loss: 0.0939 - accuracy: 0.96 - ETA: 1s - loss: 0.0927 - accuracy: 0.96 - ETA: 1s - loss: 0.0937 - accuracy: 0.96 - ETA: 1s - loss: 0.0938 - accuracy: 0.96 - ETA: 1s - loss: 0.0929 - accuracy: 0.96 - ETA: 1s - loss: 0.0935 - accuracy: 0.96 - ETA: 0s - loss: 0.0933 - accuracy: 0.96 - ETA: 0s - loss: 0.0933 - accuracy: 0.96 - ETA: 0s - loss: 0.0917 - accuracy: 0.96 - ETA: 0s - loss: 0.0916 - accuracy: 0.96 - ETA: 0s - loss: 0.0901 - accuracy: 0.96 - ETA: 0s - loss: 0.0891 - accuracy: 0.96 - ETA: 0s - loss: 0.0882 - accuracy: 0.96 - ETA: 0s - loss: 0.0881 - accuracy: 0.96 - ETA: 0s - loss: 0.0893 - accuracy: 0.96 - ETA: 0s - loss: 0.0890 - accuracy: 0.96 - ETA: 0s - loss: 0.0893 - accuracy: 0.96 - ETA: 0s - loss: 0.0888 - accuracy: 0.96 - ETA: 0s - loss: 0.0885 - accuracy: 0.9656\n",
      "Epoch 00017: val_loss did not improve from 0.07772\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0883 - accuracy: 0.9657 - val_loss: 0.0928 - val_accuracy: 0.9648\n",
      "Epoch 18/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0764 - accuracy: 0.97 - ETA: 2s - loss: 0.0863 - accuracy: 0.96 - ETA: 2s - loss: 0.0931 - accuracy: 0.96 - ETA: 2s - loss: 0.0902 - accuracy: 0.96 - ETA: 2s - loss: 0.0922 - accuracy: 0.96 - ETA: 2s - loss: 0.0909 - accuracy: 0.96 - ETA: 2s - loss: 0.0918 - accuracy: 0.96 - ETA: 2s - loss: 0.0899 - accuracy: 0.96 - ETA: 2s - loss: 0.0905 - accuracy: 0.96 - ETA: 2s - loss: 0.0905 - accuracy: 0.96 - ETA: 2s - loss: 0.0921 - accuracy: 0.96 - ETA: 1s - loss: 0.0914 - accuracy: 0.96 - ETA: 1s - loss: 0.0902 - accuracy: 0.96 - ETA: 1s - loss: 0.0899 - accuracy: 0.96 - ETA: 1s - loss: 0.0890 - accuracy: 0.96 - ETA: 1s - loss: 0.0885 - accuracy: 0.96 - ETA: 1s - loss: 0.0865 - accuracy: 0.96 - ETA: 1s - loss: 0.0856 - accuracy: 0.96 - ETA: 1s - loss: 0.0871 - accuracy: 0.96 - ETA: 1s - loss: 0.0860 - accuracy: 0.96 - ETA: 1s - loss: 0.0853 - accuracy: 0.96 - ETA: 1s - loss: 0.0846 - accuracy: 0.96 - ETA: 1s - loss: 0.0849 - accuracy: 0.96 - ETA: 1s - loss: 0.0839 - accuracy: 0.96 - ETA: 0s - loss: 0.0835 - accuracy: 0.96 - ETA: 0s - loss: 0.0842 - accuracy: 0.96 - ETA: 0s - loss: 0.0835 - accuracy: 0.96 - ETA: 0s - loss: 0.0846 - accuracy: 0.96 - ETA: 0s - loss: 0.0860 - accuracy: 0.96 - ETA: 0s - loss: 0.0865 - accuracy: 0.96 - ETA: 0s - loss: 0.0878 - accuracy: 0.96 - ETA: 0s - loss: 0.0875 - accuracy: 0.96 - ETA: 0s - loss: 0.0872 - accuracy: 0.96 - ETA: 0s - loss: 0.0870 - accuracy: 0.96 - ETA: 0s - loss: 0.0866 - accuracy: 0.96 - ETA: 0s - loss: 0.0877 - accuracy: 0.96 - ETA: 0s - loss: 0.0871 - accuracy: 0.9658\n",
      "Epoch 00018: val_loss did not improve from 0.07772\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0871 - accuracy: 0.9658 - val_loss: 0.0864 - val_accuracy: 0.9669\n",
      "Epoch 19/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0796 - accuracy: 0.96 - ETA: 2s - loss: 0.0682 - accuracy: 0.97 - ETA: 2s - loss: 0.0768 - accuracy: 0.97 - ETA: 2s - loss: 0.0733 - accuracy: 0.97 - ETA: 2s - loss: 0.0703 - accuracy: 0.97 - ETA: 2s - loss: 0.0779 - accuracy: 0.96 - ETA: 2s - loss: 0.0824 - accuracy: 0.96 - ETA: 2s - loss: 0.0827 - accuracy: 0.96 - ETA: 2s - loss: 0.0801 - accuracy: 0.96 - ETA: 2s - loss: 0.0821 - accuracy: 0.96 - ETA: 2s - loss: 0.0826 - accuracy: 0.96 - ETA: 1s - loss: 0.0835 - accuracy: 0.96 - ETA: 1s - loss: 0.0864 - accuracy: 0.96 - ETA: 1s - loss: 0.0841 - accuracy: 0.96 - ETA: 1s - loss: 0.0831 - accuracy: 0.96 - ETA: 1s - loss: 0.0836 - accuracy: 0.96 - ETA: 1s - loss: 0.0864 - accuracy: 0.96 - ETA: 1s - loss: 0.0858 - accuracy: 0.96 - ETA: 1s - loss: 0.0855 - accuracy: 0.96 - ETA: 1s - loss: 0.0849 - accuracy: 0.96 - ETA: 1s - loss: 0.0852 - accuracy: 0.96 - ETA: 1s - loss: 0.0851 - accuracy: 0.96 - ETA: 1s - loss: 0.0845 - accuracy: 0.96 - ETA: 1s - loss: 0.0846 - accuracy: 0.96 - ETA: 0s - loss: 0.0846 - accuracy: 0.96 - ETA: 0s - loss: 0.0871 - accuracy: 0.96 - ETA: 0s - loss: 0.0866 - accuracy: 0.96 - ETA: 0s - loss: 0.0871 - accuracy: 0.96 - ETA: 0s - loss: 0.0862 - accuracy: 0.96 - ETA: 0s - loss: 0.0858 - accuracy: 0.96 - ETA: 0s - loss: 0.0851 - accuracy: 0.96 - ETA: 0s - loss: 0.0853 - accuracy: 0.96 - ETA: 0s - loss: 0.0850 - accuracy: 0.96 - ETA: 0s - loss: 0.0849 - accuracy: 0.96 - ETA: 0s - loss: 0.0844 - accuracy: 0.96 - ETA: 0s - loss: 0.0850 - accuracy: 0.96 - ETA: 0s - loss: 0.0850 - accuracy: 0.9668\n",
      "Epoch 00019: val_loss did not improve from 0.07772\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0849 - accuracy: 0.9668 - val_loss: 0.0857 - val_accuracy: 0.9673\n",
      "Epoch 20/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0778 - accuracy: 0.97 - ETA: 2s - loss: 0.0875 - accuracy: 0.96 - ETA: 2s - loss: 0.0808 - accuracy: 0.96 - ETA: 2s - loss: 0.0828 - accuracy: 0.96 - ETA: 2s - loss: 0.0887 - accuracy: 0.96 - ETA: 2s - loss: 0.0844 - accuracy: 0.96 - ETA: 2s - loss: 0.0826 - accuracy: 0.96 - ETA: 2s - loss: 0.0829 - accuracy: 0.96 - ETA: 2s - loss: 0.0829 - accuracy: 0.96 - ETA: 2s - loss: 0.0850 - accuracy: 0.96 - ETA: 2s - loss: 0.0877 - accuracy: 0.96 - ETA: 2s - loss: 0.0848 - accuracy: 0.96 - ETA: 1s - loss: 0.0867 - accuracy: 0.96 - ETA: 1s - loss: 0.0903 - accuracy: 0.96 - ETA: 1s - loss: 0.0896 - accuracy: 0.96 - ETA: 1s - loss: 0.0892 - accuracy: 0.96 - ETA: 1s - loss: 0.0881 - accuracy: 0.96 - ETA: 1s - loss: 0.0895 - accuracy: 0.96 - ETA: 1s - loss: 0.0892 - accuracy: 0.96 - ETA: 1s - loss: 0.0875 - accuracy: 0.96 - ETA: 1s - loss: 0.0881 - accuracy: 0.96 - ETA: 1s - loss: 0.0886 - accuracy: 0.96 - ETA: 1s - loss: 0.0880 - accuracy: 0.96 - ETA: 1s - loss: 0.0874 - accuracy: 0.96 - ETA: 0s - loss: 0.0868 - accuracy: 0.96 - ETA: 0s - loss: 0.0865 - accuracy: 0.96 - ETA: 0s - loss: 0.0852 - accuracy: 0.96 - ETA: 0s - loss: 0.0838 - accuracy: 0.96 - ETA: 0s - loss: 0.0846 - accuracy: 0.96 - ETA: 0s - loss: 0.0851 - accuracy: 0.96 - ETA: 0s - loss: 0.0848 - accuracy: 0.96 - ETA: 0s - loss: 0.0851 - accuracy: 0.96 - ETA: 0s - loss: 0.0848 - accuracy: 0.96 - ETA: 0s - loss: 0.0841 - accuracy: 0.96 - ETA: 0s - loss: 0.0847 - accuracy: 0.96 - ETA: 0s - loss: 0.0841 - accuracy: 0.96 - ETA: 0s - loss: 0.0844 - accuracy: 0.9669\n",
      "Epoch 00020: val_loss did not improve from 0.07772\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0838 - accuracy: 0.9672 - val_loss: 0.0855 - val_accuracy: 0.9676\n",
      "Epoch 21/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0597 - accuracy: 0.97 - ETA: 2s - loss: 0.0620 - accuracy: 0.97 - ETA: 2s - loss: 0.0658 - accuracy: 0.97 - ETA: 2s - loss: 0.0694 - accuracy: 0.97 - ETA: 2s - loss: 0.0706 - accuracy: 0.97 - ETA: 2s - loss: 0.0712 - accuracy: 0.97 - ETA: 2s - loss: 0.0723 - accuracy: 0.97 - ETA: 2s - loss: 0.0724 - accuracy: 0.97 - ETA: 2s - loss: 0.0727 - accuracy: 0.97 - ETA: 2s - loss: 0.0779 - accuracy: 0.96 - ETA: 2s - loss: 0.0774 - accuracy: 0.96 - ETA: 1s - loss: 0.0771 - accuracy: 0.96 - ETA: 1s - loss: 0.0786 - accuracy: 0.96 - ETA: 1s - loss: 0.0799 - accuracy: 0.96 - ETA: 1s - loss: 0.0793 - accuracy: 0.96 - ETA: 1s - loss: 0.0811 - accuracy: 0.96 - ETA: 1s - loss: 0.0827 - accuracy: 0.96 - ETA: 1s - loss: 0.0816 - accuracy: 0.96 - ETA: 1s - loss: 0.0806 - accuracy: 0.96 - ETA: 1s - loss: 0.0798 - accuracy: 0.96 - ETA: 1s - loss: 0.0796 - accuracy: 0.96 - ETA: 1s - loss: 0.0823 - accuracy: 0.96 - ETA: 1s - loss: 0.0838 - accuracy: 0.96 - ETA: 1s - loss: 0.0830 - accuracy: 0.96 - ETA: 0s - loss: 0.0827 - accuracy: 0.96 - ETA: 0s - loss: 0.0819 - accuracy: 0.96 - ETA: 0s - loss: 0.0818 - accuracy: 0.96 - ETA: 0s - loss: 0.0815 - accuracy: 0.96 - ETA: 0s - loss: 0.0837 - accuracy: 0.96 - ETA: 0s - loss: 0.0839 - accuracy: 0.96 - ETA: 0s - loss: 0.0839 - accuracy: 0.96 - ETA: 0s - loss: 0.0837 - accuracy: 0.96 - ETA: 0s - loss: 0.0841 - accuracy: 0.96 - ETA: 0s - loss: 0.0839 - accuracy: 0.96 - ETA: 0s - loss: 0.0845 - accuracy: 0.96 - ETA: 0s - loss: 0.0848 - accuracy: 0.96 - ETA: 0s - loss: 0.0839 - accuracy: 0.9674\n",
      "Epoch 00021: val_loss improved from 0.07772 to 0.07234, saving model to model-dsbowl2018-1.h5\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0833 - accuracy: 0.9677 - val_loss: 0.0723 - val_accuracy: 0.9707\n",
      "Epoch 22/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0811 - accuracy: 0.96 - ETA: 2s - loss: 0.0821 - accuracy: 0.96 - ETA: 2s - loss: 0.0814 - accuracy: 0.96 - ETA: 2s - loss: 0.0797 - accuracy: 0.96 - ETA: 2s - loss: 0.0800 - accuracy: 0.96 - ETA: 2s - loss: 0.0900 - accuracy: 0.96 - ETA: 2s - loss: 0.0901 - accuracy: 0.96 - ETA: 2s - loss: 0.0881 - accuracy: 0.96 - ETA: 2s - loss: 0.0901 - accuracy: 0.96 - ETA: 2s - loss: 0.0879 - accuracy: 0.96 - ETA: 2s - loss: 0.0861 - accuracy: 0.96 - ETA: 1s - loss: 0.0856 - accuracy: 0.96 - ETA: 1s - loss: 0.0848 - accuracy: 0.96 - ETA: 1s - loss: 0.0837 - accuracy: 0.96 - ETA: 1s - loss: 0.0836 - accuracy: 0.96 - ETA: 1s - loss: 0.0858 - accuracy: 0.96 - ETA: 1s - loss: 0.0862 - accuracy: 0.96 - ETA: 1s - loss: 0.0844 - accuracy: 0.96 - ETA: 1s - loss: 0.0857 - accuracy: 0.96 - ETA: 1s - loss: 0.0856 - accuracy: 0.96 - ETA: 1s - loss: 0.0852 - accuracy: 0.96 - ETA: 1s - loss: 0.0853 - accuracy: 0.96 - ETA: 1s - loss: 0.0858 - accuracy: 0.96 - ETA: 1s - loss: 0.0872 - accuracy: 0.96 - ETA: 0s - loss: 0.0865 - accuracy: 0.96 - ETA: 0s - loss: 0.0866 - accuracy: 0.96 - ETA: 0s - loss: 0.0856 - accuracy: 0.96 - ETA: 0s - loss: 0.0859 - accuracy: 0.96 - ETA: 0s - loss: 0.0861 - accuracy: 0.96 - ETA: 0s - loss: 0.0867 - accuracy: 0.96 - ETA: 0s - loss: 0.0866 - accuracy: 0.96 - ETA: 0s - loss: 0.0864 - accuracy: 0.96 - ETA: 0s - loss: 0.0861 - accuracy: 0.96 - ETA: 0s - loss: 0.0856 - accuracy: 0.96 - ETA: 0s - loss: 0.0851 - accuracy: 0.96 - ETA: 0s - loss: 0.0849 - accuracy: 0.96 - ETA: 0s - loss: 0.0847 - accuracy: 0.9671\n",
      "Epoch 00022: val_loss did not improve from 0.07234\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0852 - accuracy: 0.9668 - val_loss: 0.0769 - val_accuracy: 0.9692\n",
      "Epoch 23/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0599 - accuracy: 0.97 - ETA: 2s - loss: 0.0741 - accuracy: 0.97 - ETA: 2s - loss: 0.0776 - accuracy: 0.96 - ETA: 2s - loss: 0.0838 - accuracy: 0.96 - ETA: 2s - loss: 0.0859 - accuracy: 0.96 - ETA: 2s - loss: 0.0835 - accuracy: 0.96 - ETA: 2s - loss: 0.0826 - accuracy: 0.96 - ETA: 2s - loss: 0.0833 - accuracy: 0.96 - ETA: 2s - loss: 0.0823 - accuracy: 0.96 - ETA: 2s - loss: 0.0864 - accuracy: 0.96 - ETA: 2s - loss: 0.0847 - accuracy: 0.96 - ETA: 1s - loss: 0.0861 - accuracy: 0.96 - ETA: 1s - loss: 0.0860 - accuracy: 0.96 - ETA: 1s - loss: 0.0830 - accuracy: 0.96 - ETA: 1s - loss: 0.0841 - accuracy: 0.96 - ETA: 1s - loss: 0.0826 - accuracy: 0.96 - ETA: 1s - loss: 0.0835 - accuracy: 0.96 - ETA: 1s - loss: 0.0825 - accuracy: 0.96 - ETA: 1s - loss: 0.0826 - accuracy: 0.96 - ETA: 1s - loss: 0.0811 - accuracy: 0.96 - ETA: 1s - loss: 0.0823 - accuracy: 0.96 - ETA: 1s - loss: 0.0817 - accuracy: 0.96 - ETA: 1s - loss: 0.0810 - accuracy: 0.96 - ETA: 1s - loss: 0.0799 - accuracy: 0.96 - ETA: 0s - loss: 0.0793 - accuracy: 0.96 - ETA: 0s - loss: 0.0784 - accuracy: 0.96 - ETA: 0s - loss: 0.0782 - accuracy: 0.96 - ETA: 0s - loss: 0.0790 - accuracy: 0.96 - ETA: 0s - loss: 0.0789 - accuracy: 0.96 - ETA: 0s - loss: 0.0797 - accuracy: 0.96 - ETA: 0s - loss: 0.0819 - accuracy: 0.96 - ETA: 0s - loss: 0.0828 - accuracy: 0.96 - ETA: 0s - loss: 0.0827 - accuracy: 0.96 - ETA: 0s - loss: 0.0825 - accuracy: 0.96 - ETA: 0s - loss: 0.0820 - accuracy: 0.96 - ETA: 0s - loss: 0.0821 - accuracy: 0.96 - ETA: 0s - loss: 0.0819 - accuracy: 0.9678\n",
      "Epoch 00023: val_loss did not improve from 0.07234\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0822 - accuracy: 0.9677 - val_loss: 0.0771 - val_accuracy: 0.9697\n",
      "Epoch 24/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.1074 - accuracy: 0.95 - ETA: 2s - loss: 0.0994 - accuracy: 0.95 - ETA: 2s - loss: 0.1044 - accuracy: 0.95 - ETA: 2s - loss: 0.0998 - accuracy: 0.95 - ETA: 2s - loss: 0.1050 - accuracy: 0.95 - ETA: 2s - loss: 0.0992 - accuracy: 0.95 - ETA: 2s - loss: 0.0987 - accuracy: 0.96 - ETA: 2s - loss: 0.0932 - accuracy: 0.96 - ETA: 2s - loss: 0.0911 - accuracy: 0.96 - ETA: 2s - loss: 0.0876 - accuracy: 0.96 - ETA: 2s - loss: 0.0849 - accuracy: 0.96 - ETA: 1s - loss: 0.0884 - accuracy: 0.96 - ETA: 1s - loss: 0.0878 - accuracy: 0.96 - ETA: 1s - loss: 0.0889 - accuracy: 0.96 - ETA: 1s - loss: 0.0869 - accuracy: 0.96 - ETA: 1s - loss: 0.0837 - accuracy: 0.96 - ETA: 1s - loss: 0.0826 - accuracy: 0.96 - ETA: 1s - loss: 0.0806 - accuracy: 0.96 - ETA: 1s - loss: 0.0810 - accuracy: 0.96 - ETA: 1s - loss: 0.0805 - accuracy: 0.96 - ETA: 1s - loss: 0.0802 - accuracy: 0.96 - ETA: 1s - loss: 0.0799 - accuracy: 0.96 - ETA: 1s - loss: 0.0787 - accuracy: 0.96 - ETA: 1s - loss: 0.0792 - accuracy: 0.96 - ETA: 0s - loss: 0.0793 - accuracy: 0.96 - ETA: 0s - loss: 0.0798 - accuracy: 0.96 - ETA: 0s - loss: 0.0797 - accuracy: 0.96 - ETA: 0s - loss: 0.0798 - accuracy: 0.96 - ETA: 0s - loss: 0.0798 - accuracy: 0.96 - ETA: 0s - loss: 0.0794 - accuracy: 0.96 - ETA: 0s - loss: 0.0796 - accuracy: 0.96 - ETA: 0s - loss: 0.0797 - accuracy: 0.96 - ETA: 0s - loss: 0.0791 - accuracy: 0.96 - ETA: 0s - loss: 0.0792 - accuracy: 0.96 - ETA: 0s - loss: 0.0806 - accuracy: 0.96 - ETA: 0s - loss: 0.0804 - accuracy: 0.96 - ETA: 0s - loss: 0.0805 - accuracy: 0.9682\n",
      "Epoch 00024: val_loss did not improve from 0.07234\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0800 - accuracy: 0.9683 - val_loss: 0.0737 - val_accuracy: 0.9703\n",
      "Epoch 25/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "592/603 [============================>.] - ETA: 2s - loss: 0.0965 - accuracy: 0.96 - ETA: 2s - loss: 0.1041 - accuracy: 0.96 - ETA: 2s - loss: 0.0913 - accuracy: 0.96 - ETA: 2s - loss: 0.0895 - accuracy: 0.96 - ETA: 2s - loss: 0.0927 - accuracy: 0.96 - ETA: 2s - loss: 0.0913 - accuracy: 0.96 - ETA: 2s - loss: 0.0917 - accuracy: 0.96 - ETA: 2s - loss: 0.0898 - accuracy: 0.96 - ETA: 2s - loss: 0.0898 - accuracy: 0.96 - ETA: 2s - loss: 0.0907 - accuracy: 0.96 - ETA: 2s - loss: 0.0897 - accuracy: 0.96 - ETA: 1s - loss: 0.0881 - accuracy: 0.96 - ETA: 1s - loss: 0.0895 - accuracy: 0.96 - ETA: 1s - loss: 0.0885 - accuracy: 0.96 - ETA: 1s - loss: 0.0873 - accuracy: 0.96 - ETA: 1s - loss: 0.0856 - accuracy: 0.96 - ETA: 1s - loss: 0.0845 - accuracy: 0.96 - ETA: 1s - loss: 0.0825 - accuracy: 0.96 - ETA: 1s - loss: 0.0816 - accuracy: 0.96 - ETA: 1s - loss: 0.0808 - accuracy: 0.96 - ETA: 1s - loss: 0.0796 - accuracy: 0.96 - ETA: 1s - loss: 0.0792 - accuracy: 0.96 - ETA: 1s - loss: 0.0798 - accuracy: 0.96 - ETA: 1s - loss: 0.0797 - accuracy: 0.96 - ETA: 0s - loss: 0.0791 - accuracy: 0.96 - ETA: 0s - loss: 0.0787 - accuracy: 0.96 - ETA: 0s - loss: 0.0773 - accuracy: 0.96 - ETA: 0s - loss: 0.0773 - accuracy: 0.96 - ETA: 0s - loss: 0.0780 - accuracy: 0.96 - ETA: 0s - loss: 0.0794 - accuracy: 0.96 - ETA: 0s - loss: 0.0794 - accuracy: 0.96 - ETA: 0s - loss: 0.0792 - accuracy: 0.96 - ETA: 0s - loss: 0.0791 - accuracy: 0.96 - ETA: 0s - loss: 0.0793 - accuracy: 0.96 - ETA: 0s - loss: 0.0790 - accuracy: 0.96 - ETA: 0s - loss: 0.0786 - accuracy: 0.96 - ETA: 0s - loss: 0.0788 - accuracy: 0.9691\n",
      "Epoch 00025: val_loss did not improve from 0.07234\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0793 - accuracy: 0.9689 - val_loss: 0.0764 - val_accuracy: 0.9695\n",
      "Epoch 26/50\n",
      "592/603 [============================>.] - ETA: 2s - loss: 0.0681 - accuracy: 0.97 - ETA: 2s - loss: 0.0630 - accuracy: 0.97 - ETA: 2s - loss: 0.0570 - accuracy: 0.97 - ETA: 2s - loss: 0.0604 - accuracy: 0.97 - ETA: 2s - loss: 0.0675 - accuracy: 0.97 - ETA: 2s - loss: 0.0723 - accuracy: 0.97 - ETA: 2s - loss: 0.0746 - accuracy: 0.97 - ETA: 2s - loss: 0.0844 - accuracy: 0.96 - ETA: 2s - loss: 0.0834 - accuracy: 0.96 - ETA: 2s - loss: 0.0848 - accuracy: 0.96 - ETA: 2s - loss: 0.0837 - accuracy: 0.96 - ETA: 1s - loss: 0.0822 - accuracy: 0.96 - ETA: 1s - loss: 0.0815 - accuracy: 0.96 - ETA: 1s - loss: 0.0827 - accuracy: 0.96 - ETA: 1s - loss: 0.0810 - accuracy: 0.96 - ETA: 1s - loss: 0.0801 - accuracy: 0.96 - ETA: 1s - loss: 0.0800 - accuracy: 0.96 - ETA: 1s - loss: 0.0802 - accuracy: 0.96 - ETA: 1s - loss: 0.0798 - accuracy: 0.96 - ETA: 1s - loss: 0.0799 - accuracy: 0.96 - ETA: 1s - loss: 0.0789 - accuracy: 0.96 - ETA: 1s - loss: 0.0792 - accuracy: 0.96 - ETA: 1s - loss: 0.0780 - accuracy: 0.96 - ETA: 1s - loss: 0.0771 - accuracy: 0.96 - ETA: 0s - loss: 0.0805 - accuracy: 0.96 - ETA: 0s - loss: 0.0803 - accuracy: 0.96 - ETA: 0s - loss: 0.0800 - accuracy: 0.96 - ETA: 0s - loss: 0.0796 - accuracy: 0.96 - ETA: 0s - loss: 0.0791 - accuracy: 0.96 - ETA: 0s - loss: 0.0799 - accuracy: 0.96 - ETA: 0s - loss: 0.0801 - accuracy: 0.96 - ETA: 0s - loss: 0.0804 - accuracy: 0.96 - ETA: 0s - loss: 0.0798 - accuracy: 0.96 - ETA: 0s - loss: 0.0799 - accuracy: 0.96 - ETA: 0s - loss: 0.0799 - accuracy: 0.96 - ETA: 0s - loss: 0.0801 - accuracy: 0.96 - ETA: 0s - loss: 0.0798 - accuracy: 0.9686\n",
      "Epoch 00026: val_loss did not improve from 0.07234\n",
      "603/603 [==============================] - 3s 5ms/sample - loss: 0.0796 - accuracy: 0.9687 - val_loss: 0.0753 - val_accuracy: 0.9700\n",
      "Epoch 00026: early stopping\n"
     ]
    }
   ],
   "source": [
    "# Fit model\n",
    "earlystopper = EarlyStopping(patience=5, verbose=1)\n",
    "checkpointer = ModelCheckpoint('model-dsbowl2018-1.h5', verbose=1, save_best_only=True)\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=16, epochs=50, \n",
    "                    callbacks=[earlystopper, checkpointer])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "1f381f5b-1b71-4daa-a417-e02f4894540b",
    "_uuid": "bb15226ea617cf91ed8f43179fccb5a15809e5a0"
   },
   "source": [
    "\n",
    "# Make predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "_cell_guid": "2daa48d5-ac98-4e18-af3f-a582baaa44f0",
    "_uuid": "f841760b4abca1a25cb750822f88268bd79bf2ce"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "603/603 [==============================] - ETA: 31 - ETA: 15 - ETA: 10 - ETA: 8 - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - ETA:  - 6s 9ms/sample\n",
      "67/67 [==============================] - ETA:  - ETA:  - 0s 5ms/sample\n",
      "65/65 [==============================] - ETA:  - ETA:  - 1s 12ms/sample\n"
     ]
    }
   ],
   "source": [
    "# Predict on train, val and test\n",
    "model = load_model('model-dsbowl2018-1.h5', custom_objects={'mean_iou': mean_iou})\n",
    "preds_train = model.predict(X_train[:int(X_train.shape[0]*0.9)], verbose=1)\n",
    "preds_val = model.predict(X_train[int(X_train.shape[0]*0.9):], verbose=1)\n",
    "preds_test = model.predict(X_test, verbose=1)\n",
    "\n",
    "# Threshold predictions\n",
    "preds_train_t = (preds_train > 0.5).astype(np.uint8)\n",
    "preds_val_t = (preds_val > 0.5).astype(np.uint8)\n",
    "preds_test_t = (preds_test > 0.5).astype(np.uint8)\n",
    "\n",
    "# Create list of upsampled test masks\n",
    "preds_test_upsampled = []\n",
    "for i in range(len(preds_test)):\n",
    "    preds_test_upsampled.append(resize(np.squeeze(preds_test[i]), \n",
    "                                       (sizes_test[i][0], sizes_test[i][1]), \n",
    "                                       mode='constant', preserve_range=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "_cell_guid": "649248cd-a1fb-4da6-ade2-4bebad44bcab",
    "_uuid": "7e06242a50870e07a080064a4912b761775990fa"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEYCAYAAAATaEB+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzsvVusJlmWHrR2RPzxX841T94q69ZdXT097RnE2DAyCEtopGEkCyyal7EMMhrjkfoFjMEgps2Ln5BaAgF+bRnQSFjgkbE0frDAoxbzAA+D22PL0z0905fqquqqysp7nut/i4jNw/6+tWOv+E9mnsoun1PtvaTMOHHbsWNH/LG+/a2b895LlixZslyWFJfdgSxZsvyLLfkjlCVLlkuV/BHKkiXLpUr+CGXJkuVSJX+EsmTJcqmSP0JZsmS5VMkfoSxZslyqfGofIefcn3XO/bFz7gfOua99WtfJkiXLZ1vcp+Gs6JwrReR7IvIrIvKBiPwjEfn3vfd/+BO/WJYsWT7TUn1K7f5pEfmB9/4dERHn3P8uIl8RkY0foe2tPX/92i0pyzLdwe+jk83bnyG+DcvVMhy8xlKwfb0O603bJZdwLvxV9C9fOOwL6wXwoytxLLo92Qrr5ch22NyHF/Ed+rVCP8+65JjxLFxkVIcN3jTJ85sl19PtXSdST9EftDHo1TndHIzvecf1j3XpodRt7Ne6CRta9Ldb8YC0GY+/4v30LoE/dPyrdPx58QLPZTQO6+c+D8qm3fbdO+edc/a5bDjfNnHe6+ywx9tBed41+8c5e4zb2LHn/YSeM2LPlLYLD++9D77/0Ht/83nHf1ofoddE5Me99Q9E5F/rH+Cc+6qIfFVE5GD/pnztP/4f5Nr+XtiJXxzuRSeNhYs/sA4/No62S1dldRQ2vP/DRkREPv7BOuw/Dvs/vhvWHzw5FRGRkQtv8ggfwmnppMOjKOrQgTF+1Fv4QFR7YTnbCcsv/elaRER2Xi3RX9wO7xmj7Rov69Ow9eP3w1fx/d+fh534YX3xF6ciInLzjXBSF5qWjj/mRVje+0E4v8Xpy0UYtNVpJ6//fPgV7r0R+ldikAr8ip358TYY8MLzQ5x+APnjiHP4It4c2nLm47MMwyv374Xn8PidsJx/FMa/w0dpjWY6/KLWZx73E9+DAstpGBqpr4WLbu+hn/jYTq6FHr72dhi7vVtlcr8NrlWxr4WIL3xyDa/vHtpGvzxOLjGG+oHDYk2l1hsXHUe8s86wIC00Y1WG/rZoo1UF6ZNrUlFyzJumk5L3UoZjav3wuqTfHN+299lP2sRgl+gjFbBIGKdkbPQSvK+w4eQ0/Mj+8l/7s+/JC8in9RF6lm4JK95/Q0S+ISLy1hs/62fjmczqLRyIj5BpzPXGtQWiETxY1bzcXIazp2UY/akLr/nRSdi+PA2NjTGwkyo80boKQzIqnTRduMjW7ijsG4dz6jocuzUL69vXQyPbO+FHP5uWaX/5geSPvvAyxoN78+2waWcWflljfIT2XwnXrLbCSY3jjyD0/3iORs8ClBrjC7dmn3dWcvON8FGfbBm0V/KrgnHGGHR46flCF3gRfcGPEtEaznNOP1Bdm36g1viqHD7BR/JeWE7X4XmU4zBWJ0CoHsioBTJslvghLlppXWh1dxtLfPxlnCKhEv2aYDkqwhhO6zA2BR42P7ZxPPrIDX9gWeLDoB/okSRS4nlxTEcYQ/3tVkVUkMJrpMuWyoGPxVPZdslxI3xgChyoP4HSi8M5DseMK9XcoQ0cTEXS4Fk2vM948bCOMdf7KHrPGh80vt9ONU84p9Mf54vJp0VMfyAib/TWXxeRjz6la2XJkuUzLJ8WEvpHIvIzzrm3RORDEfkLIvIfPOsE50Ran6IahTWF+Va6CFHXQAbUDuQJKvABW9thCaAjJycNmgjnb02JcsL+ySwsZ2Upo0kYnik073JOqBpkB9OwbUwLRmM7/473JiLiiNp6/RxheeOtMN/ilLMsCYcxBNSOaGw5J3IIB6xUlYflq1+YymQLbY2AABpCb+1ZOAXbOU/hNQm/B89Bz/eqYTlN7KAun94LY/UE066zR5hePQ3bTzHdOgMyZd9kFc5fY6yK0skMU9HJFu4D41xhyuHxFtd41ju3gAa2h1MK03103unfDRDdah4GY6RIB/0BwuBw4/WTitM5pQVIE/SmM3z+RELksAyRZh5lb2qIdUU32NH5eDOdefksCjPX5NIRvMTOoA9Am4XT/vNd5DSzsyTVBeVT+Qh57xvn3H8iIv+XhN/s/+y9/86nca0sWbJ8tuXTQkLivf8HIvIPXuRY50SKyklnidyOpCCPBOppJKprAzuo9WU7nLsLrTgBaimr8Mm/fi3cejkBkbcTjp9N4pza4wtP0nKKS40Aq4iAdm6mJCWRhRLRRhN3hY8aBtsq1bDWasQxgNbxKXKglWm1Ctv374R29l4plYOikou8QBBaBtdAI9SGIyAPGYcjR6reCcvQrvfaeANr4+nTsHzw/YCADu+GA04fp/tPF+CKgDxIEhcjoDcM9vZOIVszcHBApA7PtsZyCmQ0CZSi7N7AM9/i8wvbO5KwfB5kqHuj0gLanB2G/hdNWG7thodZH4z7h6uBhM+HCChavLyIIaa9Umt4B0Z8tkQ47AvaJLohBB+AnbjB9cGRRARH0pj7uW5sExE5EcnxN9iKkrSK7gww9uQt5WKSPaazZMlyqfKpIaGLSuEK5UEa+6XXSTS0vziduo5glaDZ0xktM4YGu/H5sGMxD2p+CuvMdBI4oeU67H/4UTDPLBYiI9hrqzpcd7Ib2j64HfTHtdfDcqZIiN00pjr11ejNr9UClXJZtH7xWEWBnHdjwwgqrABqKdHXa58L91VuuXjd1AVJOqKWJ2F5/3HQ9mNAhu0pfJQm4cQZXRLAuTi9P6dI4AwuEQ/eCQjn+INw7mNwQ4tjWGVg9aLlZ1U1uI+wf//NcEM3Xx1pX2qg0x2gVaLa8W6qQz3HBMii4ruhVkkeiGUfBmBbVRFVhZ2rOVw7iCh87xwRJYG8QRBJvwzX0xlLlfbTugOwTzRXGm7RKfLopIOJ0w0IpRSZ6SupBi0iOHN/ltcU3zMnbjwknntByUgoS5YslypXBAk58d6J5xyYmsA4x/VRgSocOnApEkqtLB38TkazcKsHbwTi4MGDoI7e/QDedE8AF87CoixEyhvgleB8+OqXgpae7QMpECHQ74ZaxWgEndsT/ZQ9Sov31lJNYgG04omu9OSwmO6Dw4I/6rVxOPDadTq0iZBkc2auvoZj45P3gFoeEpXA4rgDtEVrFJwBd0awDgJ9FUUPCWE8z4B8nj4MCGd5GJYduQUgn8lOaOP250O/vwAEd+NOlV67FdW8cNlRzop8nqLJFHBGL3MCCeVLUqRa9Pyd+CynO+Eikym4IDwHvYbxbragV3qPU/19uIu/OnJEazyzyidNuMJrG5vuh8+3WUfCpoSfG9GU8jcK1bhwybon36qW6BTueNnwznJpUdMFjWUZCWXJkuVS5YogIS/ivWoodW9XtBMW0d3dqXev/eo2sHgsjsP+B/fD+nvfC/EBTz6Ehr4fNPcaHrwTkCxTuP5P9wu5+eWg/l7/EviJA4Rj0HekSjkXVS7Qppz7zxFKQS/UqhYx+kaEFhEsmxPsh99S1JLhzOksrN/5HKx8sK6VCC2R1g84EIKpY1ioHv8YIRTgaxwMP46hH4uwTo/kGhzMjCEBhYv9pyfuCsgH62PySLB23Xhjgn6HRvdeSX2xHKybamFciwiRJnmLMuVMGiIl9MSpyjYIhIiQ705LdFZo2wVeqBrPttGYofR9I6qK14C3MCk9tWp6fbYqRbrs1P8GFquK23F/eKfprE2vaIGFcX26kmIU9nqMTWsYG2v9ipOKlEMiStPfVWJB7N/xBjGc0YvKlfgIeQmW3s5ARooz+Np73yMCUyZtDoL0w3fDA/re74ePz+N3w7Jo8QBBRHd4sqMJCMl9mN1vFxpzNAUhShO8mnhtP9kTdImhC9/5Z2GON8Y85u2fHclsJ+2//qAwDSORKyuS3iAx2U9MRaoDA2ZJoHZ9LjsluU8x9aSj4Bofm8qT0MX6mEoBY4WPqsNXyUv8IO/fDNvmb+CgOpyzfS3c843rYaDHcGso4UhZI5SiHKdvMH8MVefjr4/9ckX/0BhjqL8ntSVjgR9ml35oGOhaOK9j1VkTe8XnYwwFGs+VMtW+NR9M3zPbsx9N+p5rvBpPpULB2PIjpS4sfMYYo3JaimPwnfVsjDb3ZJWiHxs+dNMXPc736RBOQVOaIf5O5UKSp2NZsmS5VLkSSEg8iFkG09FRqrc/LKAR2k4DJhVBIBDyve8GlfBH/19AHyRK3ZokZhBC2prasErJ5p2blexdRzjGBAGpJqWHN0yhNxr36HG49rv/byC/T45Pcf6e/PwvYO6D6V8H7QbfOFkgtUfFIEfMOaowm1FnNCJz7ZNOB0RDUziQnBUs5i2OBRnJqURkasMCY1PV0bUg3CcO816RHMn61zB1vQGnSY+5hbNBl0uDRoB2SpjqiQ5cLxcGUVdhwnicTwM9eR9m1hM1teZk6R2PfRwjZ8fVBFXHcU/Tz1iHW+kR6xyDyEsTmSVNiAca5tjQAVffeQ4Og69nYw0ALtRBlo2xVYPQuFUzVqS/p0FKECfDkBCdyqU3cNEUZRkJZcmS5VLlaiAhEREfNUKFz28DtUlkQdK2XXlpwOnMEQj54TuB8/nBt8LyDGbnEVTCGNqcWQZ0HtulmmBrHzloblcyhcNaoWovVRN2mk1uggrjIZDQg0eBdFnC/P/hHxTypT8RkBCd44jymPSLAarU6iUDRck5gFylsxlzH7VAUq6NDoAc1yXQ4tkpzmWYhqbyCMs5QkA4ZnWdppDwuEbho3ZnqpECjTlwVy0QT7dInyERYFHStAytzwRsmg/Hq/uCumEwlQjJVHIufD5EDBwTTVRHJMXzeAXppfAwsMQiioHz3znEIKTrvBQdr4tDiKjZhjFsKA9PQwbQ7/KsRZt8//BulKUIjtGUHtb0obMJnssxJNTmNdP77tNvcWjS/ttQlYtKRkJZsmS5VLk6SEhiAN8C1oMVUmcQQayAelZLLwsghZPDcMw7/yx4JR7Bac7DVOzASRRg/zX9AK0ayNK3fzuQEgc3kFKj9TJH2okWZsoKxzLlKkWnxFiu0LePfhhgzdNj3ABCMparIqYt4blAOIujcLHVcarVqxvoA4NkaQ1E0yfo6zGcA2djJ9fgZDkeh2VLiyCuVRVpBkimvF3DRF+BC6sZ0EqlSoQqUVMugY74rE6f4paBvkZEs2yKrggw3XtaLamRyYdUEe1RfC8lSvKHCU+xqMSil9Q6TR7JJwfTFM+sGTZRXWqb7Vk7+/yJgpIUQatFTg/EPQMy2bYrtZZhO9qtimjeJ8fjNR8LGyFCItxK+UKbO9e6vnT9/puOxQBalyxeVDISypIly6XKlUBC3gctPD9Ogx1PD4NaPDwM6wv4AHUrkTl8XI6BfJ58EFAHk2tVJTmTsL/BhHaC1KuzbQZFwvIFTf7hD0I79czJDKERuwhY3YXFZ/cAPMaIaVyp/UIbJ+CC3v/Dw7Ad5Bad0A5ubakzHLUg+Rj6CVET1UBfMzgK0oq0RrTJg/cC8jn6CFZA3Pf2G7WsgUqcJr8Ki0J5F/R7jXVwSNtwIGTu5pp8A+6zacgniKpFOocyZQefoTPIpqOvj7HaFBrEyZQmPZ6HnBy0eUzPinOpefk2+1S3+vT2B0nznESepmMec6b9MClXvOEF40U2X8N1PqbTjSSKWWcTqdWOaKo1yeY0s0qPm/SGoFTEr5Zb01EbX1Kk11a/p56py1n4Z+/jE7JCGQllyZLlUuVKIKG28XJ8v5XlInzyTx8EfXKEJOlMEyEIvCx9TOp19gDcz5pWFhzDxPVAErtIYjZFOlf6yjyB9zBL0szg6TueFbI6SoNiaybXInrCtagd6Qn7AKjs0aM59iPJO0il7a1SfW9U0+AaY3hCu4OweQuVI+i1TZX2FNa/pz8K998CzUz3Qt9OH3v5+H7gyV55HQgOlr/t/dDUCqEhIyCbCfZfux3Wd/aJhGBlAsJYIijYN146pOY4xDgukR7EsSpFaaxn5DFw/0yJ6zQQ2RIKTjUwtTrplkpTYdAnCWOqxBOWuGZp/Lk0yZt3MfiYyK3nb5X0z0QnO+NiTG9sgrG26WIqEUVLPjm2SIFegtBCI+hvZQ5khQInPTSFXdaSdR4/tk6327JDm+gdb4mx3tnh4htOeoZkJJQlS5ZLlSuBhJqFyP0fNDJ/Gj7Lx0iKTmtOA2/QMT1OJ04qpgFl4i0t1QPOZyusb12Dlr8W/HKOHwQ1/uBe8N3pVtRcsBwhuX2zLLVgItOELoHE1kAdxZQRhtAa9PjGqE6qYFZawcln7wDexG+NpcXEXtOaYizGCORE9SNN7k5EscaYLB7BigaLIRPDL4Ei24+crCTc45ufD1GkrJt28w5QIkoWFUAtW+DJtlD+jfF0apUBAqJP1fLEy/IoXO8EvkcFxqwuU/3mNB4Nz4MpWomE6BpuXJULFxOsqzsNYErhUr8hdYi2ZiWuDqKGsb3x4poUMcR4L9PUIFiZ28nlpZxW5xs92BiqrNFsIOrXlFjx+tZCIruIHweUFRESfaO0dYN0TNK8zRZGwxudg3hyKo8sWbJ8puRqIKGVlwc/bqU5RpqN44Acalif6LnLcjzTayJTRLtPse3pI3ASQCtH8C49YzG9HyOKHvPoFRAF43SoQk6hEZtaBHnQZK1R5KlK0qk5/H9YQO7V1wKMefUgmLA8otS/9C9fExGR269NNKF6Zawa9EFiWg0tI4T+EQkxPUgBy1WB+5wfxgTyt74ENHiDnrehrT1Atek2PXDDuSy86BjHZerCaApQLFfzThbwTyowNgYYsKCsjIC6ZrC4MUVrpQnSMAwmtsl3Tr3aI3LgOhEQ+kcrmEvhS0y/kaIdogMpRb2yVbnrM6aVrEjuKyaoozc5+SqcR0RXOk2FovFn5v2x6Eu7wPPwTrbKJeFEes+XPnX/lmhZU95S2+rMWGyGLeewPWYjr+mSYy5qI8tIKEuWLJcqVwIJdZ2X1UkrcFORJRMzQbVt3UCyengA79woZAK/mf39oLa7Pwhw4xGsXrVyKGF9he2kKkApSaOR4WH/0gNStCOpwRdpHpoyXWrZFCIHE+vDhF5bYyAPxGCVrotWMRoUNHKd2k6wjv08Ts027IOafkREZJeJ1yaNvPlzwRpHj2eNiifiATKij9KIkeTGb0Wdf5lIHtubhZeWic9guXGmvyNwW+TmiGBptYzJwtIgrT5KYFsMHmc+IS26px1MuReKcirKg2DZr7Pe41dENN1ULwg99RuyieK4ucX7VtY9yypRIrMA8JqG63Empi1atnhtgxJ7hJUmyTf7bFT8wNVbDWwpr7YJz6hfkum35he6KARKu5AlS5YslyNXAgkVzsmkLrTkTAkL1wi+MbffDOsHr4fuznYLza9zhPSs9BGZjJl7JbRN/5m2YJL2sEMjssXM9Rmt3nlFBBV4p2qblhx0HNpSPaehjcbbYcfbfzKcONsJ6Gr/NqPyRckSoo2SCMJYHFqjuRTJIX/PI5Qo2pkFEqneCjdw7fWYvVE1sOGffJfqII0B0tgkWv2oAVNuom0jcmDsEnkzWv+291goEigR3JBqfSE6dP1Lqx4uC6ce0iqAh86yEEQKxgoWj0otXqrZxSlcUj8gpjUl+qPfj83iaOK4IrqJBzJzgDOm0JgdIO2XohNFWynnou8fEeIGCGKj4aNrvhmUCKuSa1lfIO+GFsJBOiGD2F5UMhLKkiXLpcrVQEKlyGTHyQhxT+MtllcOyxu3gID2SPWLPL4fvrbv/nHwLTo7JKEUFiv47sS4G2odxh0hTzI1tyYDCtu7zqlWrOG7swWfHZsPplBUFYSR7l/4+UB8vP4zgSQaw9xWjtxgLk5Nu0Z/mHye3uO0Is2AKG6/FTZUsHYw10+NYo+7NwsZs79Aaqqb4SXbdam2s9YNj/3McbQ4gVXsJGxfHno5egjOrRdZLyKyj36OkV9ojIKKsSRzkPOyaPa5oaHvi0U0WFo/IBMHRT8ub9CM8z0LG3bRObkxHJXySoY3i4ZT+GkRZZZFjPpXngvXKk3/eZ8sj6ROPuk1Fa303iH7LEu1aAJhDrIipsvBELO1DWYxr/F15Ljsgxie8yzJSChLliyXKlcCCZVjkb0vFLK3H7T7GHmeGfFeASY8QoT2xx+t5OQDFO6D/w/jmlgepSFCoGe1WlTwFe8QcwUNsQaJBKOGjKd1jENjvBPjuqiBqP4KciTU0OB5EGs2hoWo1Pgp13cOCQtoE3pjv/9H4b6OgISm4KPe/FNhjLZRkvr2zwYuSKs8jBj7VqrFSq9lyg9rmRvh9tQatkTM3sOPwomP3w8Qioho+bSTNeLPWpAlLB7YYOwewn9rhEKS9NKujNZsDe1DYFpEaBS9mDWDJfpvuRSTDdHHP8J+ooZKoUW8Bp8dLJ54TbQ/trqLZhSgVzdj5sj/lE7vjU7hNq9QvDibTi1dzjwnCrMztK6I/j58v9WXKunukEc7V8jvcC121hZOVIthNzz2ReQTIyHn3BvOuf/bOfdd59x3nHN/FdsPnHO/45z7PpbXPuk1smTJ8tMvL4OEGhH5L7z3v++c2xGRf+yc+x0R+Usi8k3v/dedc18Tka+JyG88q6HR2MlrXxzJzm7Q8hontQjfyLs/Chr48Tvgfx634pCzeGcn3AKj4JfIj+yhkVpq9xpWmleDitsFunr6gLmqg5WJ/iq7N2up4LMzhXcvyx835JGQr0ctK5zjG0sJc/w45lV2Rc/9F21Cox7DA/kx0Afjus4QI3aIuLpdICP6/Gi1CnAwRRnNGVpimhYsSaX1KVJYwev86d3Qh/lDcBqnQJPwbC+9SIt7ZpUQ6kDNs416b3uvBn5sPEutL+pfIymqUW3ru1jBAmOhvAu5ElxMq4L0rV4iQ9+XIkUJrfMRDGEfORVaEIl0WGhRzyXfo5auIJrayDlptJxzWNJHTZ+DqV3WY+WSNSsRGXrlPC30sYZF8kjFOZ7S8ewNME2fTYqSLsoBWfnEHyHv/V0RuYu/j51z3xWR10TkKyLySzjsN0Xkd+U5HyFXOKm3ndQgUpms/dHH4a179FFYNpgGtAuRCm/mFEGXu5hH8WO0RoqJFcI3JiBKd5GmQmvV42M2ReDqZBfm39JJiS/AtZuschrOWSBo8/HHgv4ikRhcCuikyAc+neAHxuqihQzeN7oUHD0i0Q6XAjrPnYYDbh4j/Sx/aCN+2ML5WkJHXITHmg4VLy0/lubaNCWf4EP3GB+h40eA/SCkG/hzJi8fp0ZMfYtfKxPTnT6F68BNPOMi/SU6+0vrIXxOuxge02B+w6DdkvfFAGdogRhDa6Y3Bv+7Ll7Ppl6NToxYZ6Ix3i9dLbhk4YRe8cOqSG+ObUa3BEvOp9Plga+BbqVBpOsd4tJTzTVjUnpDeqeHyzClSro3tIWx+YSENOUnQkw75z4vIn9KRH5PRG7jA8UP1a1zzvmqc+5bzrlvnTAhcZYsWf6Fk5cmpp1z2yLyf4jIf+a9PzovIM6K9/4bIvINEZG3Xv9ZX4jTAEQmRz9FvXh/yoTzvWkFpwHQMkz3eobpGJFQC9vxPsIZ9hEQ2mA/41fpZl8AWZS+k51bMM1fh8ah+RyI7DHN5zBht3OYo4EUNHEXU4ciVMH5TmLQX6oNl0A8NUI9SpaFhp18hCkhk6EP03JiiJoegatEbpoWVaeRNJszRQdq1Z99iIDiM2rRcCCrMredU5cCamXkpZM1pmw10KGJ/Y3BpkZDG79DER9dJTjlbEw+DbahdgKG2ZDcprnan4conF7YJhbTFB3GKVMd8wza4rUaRRheU45olIh1JbDX5Kqxj6u5XbsQTeMa2nEeHBlstuiMS0No94+zAba2bQvgXlBeCgk550YSPkB/23v/97D5nnPuDvbfEZH7L3ONLFmy/HTLJ0ZCLnyW/ycR+a73/r/v7fr7IvJrIvJ1LH/7+Y2JiHP65W+BZpZLhFjQhZ5fWucFQEaOngQOhcnlqbm0HAqCSVskPztACtOzraBWj+cw1SNR+xrX3nm1lFsoaayll4FKWiRZ03K97DfSazSYXXYIIVkDZTUw1bsqpn6wjndMY9KBFxvBLeD66wEC7d+A46GGYqRjoyEVK68aZt2SywGBi2uWuK8KCIM804iRxDDxN3gOcwRnKp3jRZZsE9ciSYxuayK32U6KWvoG3/66sa6LFDGMh1CngPm7LJliBcfyFaCjqinbrU36Dar7HFRS4NpqhWZFTkMeFQrOMLa9u9PUI9Y0n4K/nhi+hqu6GeNgttu/N7VuU6XEIbADsKFLg0DV1OChYSYXJIdeZjr2Z0TkPxSRP3DO/VNs+68lfHx+yzn36yLyvoj86ktcI0uWLD/l8jLWsf9HzufDf/mi7RWliylAwXtUM3AwMJXX4FjWSy9PYEU6fRLsyTCcxOJ60EXVJGjiPaSQ2LuOtBoMz4CV7CkCQQsEfV57eyRbt8K5BZObraiBw3JMXob2aE3FQLMtEAgQE5NTifQ5ECAhuhYw7QOtXFCxUzhwMj1qp1ooba/VsIFGyjZ0fAHO6gRWPZr9txEEy3SutKzNwIGNgB4ZIHn2mPldgfC8lxEdIoFOUC9RDl4JD/H6W0istp36LSjyoRY14RxaXkaKWL6G4IMRET4dIx5H58CG3FDfSVR6zou0zPVfYwObNJVKQdSL2zCOndGORfTCa7sY5mNCQyKiQL8sr2OcFPlu29CWpP8DIGPux6KYjUclvYz/e3OssSBqyxckhXLYRpYsWS5VrkTYhjhoN3wSJyAUXnkzIJEx5v4LlJNZrVtZL+E7hJPWII6WiFitYUpp1uEWl3RoQyDlFlT2BIm9rt8O21f4im/vlopGdAoP1UVrF1OxUuMi174uNWWDpEvnJGpcEw6wjQBUBq7ymr6iVgz9Xp/BMjflNdEe1GXZiiyXYdtpqMGoJbI7ICP6sjBMpgTKqpEQ/9proS/Hj4AScV/NIqDs2gWgAAAgAElEQVTQalTKDOEjM4zrFFbIA6Rf2bnukjEpjZLUZFxqFUv5DnE++tGYFBcadAq4ogngaUVlylabrd4kLKsKF1GS4V20n6a/is58Zw7wyXGdOH3UyuHYxGjO91d7bfE20uMLO4Zenm+5MtudIY8sl2TRTdpfcx/2sIyEsmTJ8lmSq4GEfPBpUZ8LqIxtpG6tQGIcI1nY4syp6qxh2mGAKi1rOq1mYTxOqIEYxrtAM9soaIj0o+R9XF1E7U2uQUvMQFvDiuQQdMoUHhOGd+AbTy0brSOup83ZVjiWKIQ+MeQztg4C5PnOt5E8H1r+S38ybN85SJFQ15YyPwvb5k9QIBJe2MpnIDUrwyGYXF/HFl7me0BEreY8RQjM/kjDMEpYAun1vo3ijUxipilxYbYkb+OI4AiRtBhf5My0ICGRpXIr1swEPg1bmZ5F0QpLURv/KSlERlWKptQdiEttokiOcyZcwxkLWOF6/fQWGqf8kfVi9oaEcQMCKiIRiz3OtU/ptVN0xZJXw/jWPo+5+VoWmV1UMhLKkiXLpcqVQEK+E1nNvRSMCUJg6GqeaogRUMJ0q5S9WYAKyyU0EwOi1FcEFhOsr+h1TY1GPxvyO+RDYMnqxInAj0YVDxENRo1pQhwQUTVOuRWmtZig30wNEpJo4dxSJ/5oC0iNFkIkJHv8NMCVQ6TTGOO8+zuB4JnsorNAEmenToNznyL2riZ3tU1khvujhY5xahp/F/bTk1rj6hgIW5ZyBA9vjxg9QT9b5CAdIyEdS+NEaCG4eLocOtx79fxWTkidiVI+iUKLFqkhywlprBwa9l30WCeajUnnUrOdltCht3mbIh9FQrhWeEyp9ctyhNZXKpYGR1cMx2gLMD4zdcYAIhk0NXAfP9/3x8eVjW3EwNZn9GeDZCSUJUuWS5UrgYSatcijDzsZYeLdIrK9BSrRBFMsBdQ52b8ZoMIxPZ1xzmQNzgHzb/VkhVZfAl2pzwkjyqleWYKmE1VvncInIAn0Z6Qlc1zST2HMmFqbcLpmp/UDr1JqkSnarMHPkIZ59H64gR2o4gY7OiAQJpg/Ae/z+F0va6S8nT9lKpFwj7N9XIOFFntpQvtCHmeEvszxPB4+DnBrvfJaJmhsUOBjeFkXs4CMZl8Oz2uyFe84vf+0D1Smjffqp0Q+qV+qJxycam/1MCrJ4YV1gk4tRriOJAf5o1gUsEiO5S+lRJsdyLdqEOFuOCXpVSQa3mxyzxZoKOBR7qpI7o/i+5cfBKANYIvpg08P092GFPM9BGzbsP5OcjHJSChLliyXKlcCCa2XXu7+aC0TFzRszUTf/KSqNgJXceQ11eo+kpUtgQicoyUnnatPoMFWQDXRHyU9ziag73eECrhiiWbwNgv44SzAO5XYX8DiQv4pjna0jllryhZirLaRd+dH7wQkUSCXT4mij0to7gli4NaLsP/hD5GP6JGLBidYDFf0MxmjpBJoJCbwogM474tIqYO/0RLJ9x0amk0rEVjYWPaIqPH4CdK6vhf6c3QLRQ9ZWxvq3Y6/zU0TrEvYR8DTnAMZeJxBJ+RxNAWrT4/zvlMrZNsAAQFZesbRafiahv+jg7iG8YaOfe5hFO22JcLMsdaCZTigTbROZ+LMrKuOzVJg49Zi6iLex5ArOg/hDCigC0KhjISyZMlyqXIlkJBvvCwfdlIgSXunScY5x0y10fqkizwMYpdqeP3uO5zMdKNATNWMcVCSLC0iKTjvdn4D+49zeQloy3tIut+CB3kV/MfWLhDUhgx/Nu8SV2sgnfF1IIvvBC6ohPXvFNYoFtIbwQ/nwbvhhh/9MNzn1lalx9ADnSilxjo9pZmdkb4vTF97iswER/cCGquQGXIb6MuVMZ/Q0REscIx5K8FVzcNYMEeUjaLvzHOgaGJ5iVCoM6lUKTS8eaPtCxNdz1VNydqL8yJK0X3w0i9NdDwb0ZLMpdHjBmGIxLbFWsdMKW9bGnwQbW9hSx+1G2vWAIxYw5Zpyw2gkTkxseVt5n7iz+ViUCgjoSxZslyqXAkkJBLy7VL7rddmfms8RdtlJy14I+Y71mhsIKMCFqnJFqw3O0RO4DlUA6c+HKKa2ktnCAr2h5TEQ/jsPH4vdILZGkugrlfe2kKfUu0vzvWCo9B/LEfguIpt8hdAcOCXGqzvEDEB1bz3w+BcdXQfqOXVSmbwTxrD4laAt5nsphH5TJLOa50oAtLAtnBfuFYBzst5H6P5D8FdASXNUO5ohnzbsz0iTJyrfAxuvEuX6uFeSC97IS2bPAj3ZTWzPlvjy2NKOiuV5GNWTy2SWZljhP02fygXkz5PdfVxvfd4QMgYAscgnMTE1r9B42vlbCc3nCLnrA9YLJtaUiFr7/7OI57OvcazJSOhLFmyXKpcGSTkvZfVGV1Ew2IC1EKNrXP/upCT4+jtKiLSQmeNcVCp54b9jGkaq/bXC4uIKOrpeh95naMbTucIRQF//EfBNMRigFRIC1rq6INUphqwrzUtD0C/FFZiYBZE5W1w/M51+s5gifJIW9sBiazmrXicw8oQB9fAEe3jWloaG8cBHi7gVzR/nOakZobFJTJIbm/XmuVwdg1xfhj3LUTXX3sT/UKeJk2vbBIUdm7zODjneggH70F0bAqbdX+qg2OoFjm9lKMrehySlsLRk9N+aWUS7GA8oQHQKjGvkKjv0bmezUXaiPoV0SrLNm1AWw8BWeQzjIrncemRcY2zjWeY4BjTZryq7VX+eWZW/InJei3y8d1WujO4/GM7K7DuISHWDgIq651CtjCdaVEji9Mbx3AHLGmGniDAk7XcWarFlnYpklCA9AVcweHxA5DFhyhJNEZbrL01qliWJ0h0wIsvplpjW/Ow8UKORunHMtasD/tv3Alf1yX65JGMfowpWDWN6SmWSL2x/1oYDE7HCnpCajqN9NqFMrqYCsJQwHS7i3knqzk+/pjyMRD41ttI6fF5JOyv0xeziIOjY9KXZN2SqMN6NmHRmUbo2uHPmVRxs3f6MecsmeNOP1X9SCpxjVO1YIHpas9kPzBu2/mKxs9gVahgLUucXmvTb/28aZi9dPzGmA+f9j9tcfNnxX50+N5cTPJ0LEuWLJcqVwIJdZ2Xs+NGPBKSNSgo3yIh2QhM6BiQfudGKVMgmgbhGr5hErBwTAlEsAtT9/U7NOXrfCecZzjBPrnZQbMyfOTpg7Drx98O07AK6pAObAtERjpM+UZG3fe1uy2qZ7VdXYOAPgBqOSWCY6qMMCb37gYotESa2imrkI4KWYIoH4EUPngF6WqJEFhAgAG5QHJbOH73djjw4x+EAxdwiOy6cEJ32EpJ4hzm/ptvhX7dRJGA8U6XXFMRgqbMSBEIA3kjwogJ+9ldi5401MISzjYvqiWTFUHJAEK4gmE/nIKIWbp03VvU8gKiHWDHzbrhjOP74+2GKOZ9jtdK0RYvxemlVpA1ripF73cRaQQzBzXvcCams2TJ8pmSK4GExHuRthEHu2yJINIG6yMEPe6/AnJ5x8ka4QIjmpvBy0yIQmCavwHtT06IQYwa0kD1ig1EKG0TE9STiH3/2+CCPkR6UzDPHXwL1i4QVOPtWWhTtSQ0BDRyVbrefD+IJmnH9Rk6sf9q+GONevBb12vsD+vHCFglWmPxweWyFZmElc9/MaCTGVLHDlOSog8M28B4jzFmxRjHh3xqspqDoJ63UsKBkaWKqnvwmQA3dAcuBwz6VbHmZmpkQ984Fw+J/C1REiECD0418yCRPAsZYqsmNeui24EOjYINY8gw2l4vabmhPr9r9kVC3cASYxWPxxuymNJbHaTZ0DYsQ81xxlITCab7Y/GGHoKyl3fDQzb14XmSkVCWLFkuVa4EEioKJ5NJLR0SlDWqmWneDd/Y668HdXVyfy2nD5jRHVqGAYa3g8q9fhPBozBZd8hh6kxwbAPzR6ku9UwT66WBxWmOBPsfvRO0fA03ACYrW4CP2Ude193tMKxMfEV0w1QUXWJKVpNIEGYNARK682Zoa30zXGMHKW+XKNq4APCgyVtNyrNOvvBzoZFbb9R6TyK9kAi6AaAL7Cfr+5ET24X5fY1rnsBUX5aO9RGlRMDwCUJYPoLZrluEZ/bGl+EouXUOd9IaixWtT2WwL4VNKQoZ2rwMZ8G0tedADE3R2m/EliQy1iE1wHEQB00bdOajG4U+Y14XpZPcYEg227gUbdnE+L0V21RctxwkuThzA5q2lkfF/cNcZSkvpltzUrMsWbJ8luRKICHv4ePBRFLgD27CF+ZzXwyavISKnj/uZA1LzvokbDtBMcTD+0FDffhu2L/7haAOX7kDKxnKKKvDnvI2WGIuvFp6LaszfxiusTpmcGjoINPNqoVnEoazhn+TLdPr+upqOPHHsbQmhfXZLUK8sNB0FEjOtg0/qAIcWAWu6OCLI7n9do2WgVKO6BcE3omGQiQkY+npBVJ2sDQQ1T8dKAuuF7GcDSkEJo2boo1T8GeHt2mdTLm5OPDpOBQDVNDTuJpci6gF42/OcEWKC9R6hv4TkZZeeoGpPIX9SBGaJts3ZYNsKpgE7CoCSq1yA8RgLFjeWL/skCU7uvTcgUXNbC/UISosrIuVNq3m44hIbePnJ/l4MclIKEuWLJcqVwIJiYh438oEHND+zdCtV94OSKg+COuHj4L6OTzqNMiVvE3ThnPnSLR+di9o4Ps/Charh68FVLD/hYBivvALoe3b1/k1BxcEf4lm2cmSbYNbmKFgYkekoHV5MbeHj05L45ixHvQNLLFMSpqyM7qbgKsi54U2eN/0u9lGyaIClqwprFLXXytU252chGNPmYwebbFg4RjBsmxbkdAqokIRkYZ+QuChWvHKmRBJ0AF8NQUSQluHSLZ/7QZKNI1xoHo1bza9OB+RHFEg+0ntTUTXqZ+QQQXn0FCuZyTU4FYbHGpQSfR6T/mpxKGm1/9C/+uhDX1v0n4NAI41N2kISdpH3z9kExILHZbYoXgxrz5W6WBpmpOE4yKvFJ9N79K9Py6GjDISypIly6XKlUBChQuevhNwKiwtMz+FZn4vaNGHD8Ly8ccrGXWpR+e6DeplArQyggWrxle7QVqKhyeB6LiOVB83/9VAqjis03O0W8XSQ0QU18DP3DsKUKCENY/xXMs2oK+yDm0yPs2mMC1Ezp3gR94oVd8MfmTS9wk8v3duh6OmSJmxA2/n6Z6TJXyp6Eu0ArKjJRD0jcxoMaRFhEnNjnD+47B+Bo5sCYti23oNJtV+w1J4yCyu8NvqjlnEAHFRDCQWI+qvFd2fVUejf0yQVjColwUJHBEq9tMviN7a1tJFbV+4YXIyI5pShddi7KF6eKcIqe/dbFHuBrorkcFmk7rVZvgQ/3zXHGsVs4G36qfWp4Akjm2/1PQgVay5yIbMIs+UjISyZMlyqfLSSMg5V4rIt0TkQ+/9n3POHYjI3xGRz4vIuyLy5733T57ZRuGknJQiKEnz0T34mLwf9vMrPp8H1d40XkYIl6+R7IvajuusBs05L3mEMbijE1i85ktamaDRoAratVcLHDmda2+kPjpzlNTpgITo0b1XGw1sNZdESwN5AluauNCS06mGYkmiCsnq92+xAfhFMSZORBaIqzs5DPfKODuWE+pw7y29xtkUEMeC1kF4Q69WaZGA9bqVooJ3OzksvFK89il5JI4F/J9sFHr0D0r5n87HRPxUmSwGWJr0E51RwYpKgK7oE+ONVcp1LiZRMwjAGy9l5ZEMBUT/M+W4FCy4c7kcywWdh5DM4Ul8I3fYbQpWLOJJh2xQWjtm+9vQFVs+6BzL20XlJ4GE/qqIfLe3/jUR+ab3/mdE5JtYz5IlS5aN8lJIyDn3uoj8OyLy34jIX8Pmr4jIL+Hv3xSR3xWR33hmQ15EGpHH4FoWywBBOP+uGVRMf5zKSdfScQbcT8GcN2HzWQwKExGRGfiDKaPr91CemLxBkS475xV9eKYHhf/SbfgtFUBXHi7H+/BF2tkBOtN+Y9kr6TJIcEXOgR6rSg1BI2N/bBNjQ5dlnzQjy6XXUj8sRVRhyOiDxFxE9A9iZDtfCvpMqXNwl6rVcV0oH7ZaI9k/bms6Cq3MUHJ69xpK/tSpio11C8lLAWXBMtd0XsZ4AViieaSolb2hJUck/QP3y6KJSe/7aMHHFKkmUG1gaTNIQXmSQUHGXl/U09lcNz1jEHdmEcbAgMguyRBNnJeldTAKA9RluLretc8FOqnx7p+7x/T/KCL/laRc1G3v/V0RESxvbTrROfdV59y3nHPfWnSHL9mNLFmyfFblEyMh59yfE5H73vt/7Jz7pYue773/hoh8Q0TkYPRF3zSdNAgBZ4njMRxACtXQyKUzLmWF/EFbU94IItmhHsfT0NbeHtAJE8MjBuvGF2iJI/eCjqm/SBwcciS0TBVbyPgI35y9fWj5bZsjlFzFBotJkXIf/IyT74iB1amWVy1TkOcgUiLfwT57jRVjzp8KMIqezywQ2SJzP1PhMp0u+RlzO8rRlOJ056xM730Gr/GbryGX0auwWlYpx0WvZQJXWiTn9O72XsoDwbmhDV+R98I9N8biY2PMOLbqpp2WfxLfA0DKqaVtWGSg6ybzZWcgiGGfkm3netMooti4eYCkNuWbt9Hz2qsBxLGNm2v22n4+9bOJSHq+vMx07M+IyL/rnPu3RWQiIrvOuf9VRO455+547+865+6IyP2XuEaWLFl+yuUTf4S8939dRP66iAiQ0H/pvf+Lzrn/VkR+TUS+juVvP7cxJ+JKJ6Uwoj0117Q0pdBiVFZy+9Xw9yuvI7+Osv3hjzFiqWY3gJ6QYXEM5MPMiyztYrVo1as/zAJ/jCyu0fYIHsejHbRZG34G6QDW0NSaXbD0qobVL8N6qFovWZZq1hihVJ9qHmta2RoRz5zXUNfMHW0D5ugBXcDTewzkQcvb4gE4JCAnIqnaFcpNdbBAjcEBHfD5IMf0BIhUI9dx7RbXProXBvcMSfYdy1zPnKISjR5n6mtGyTPzgHmb1THZeDlblFk6FzMJmmOceQ6af0cBKY9PLYdOybsIJSxA6AHjjTuGXs8+Pdx4Lvc2PddyFWPKzMVMsFmS/NFYBO01/ABWvZh8Gn5CXxeRX3HOfV9EfgXrWbJkybJRfiIe097735VgBRPv/SMR+eULdaJycu12KROQEk8eh+0sQdPAEqaler3IzvXQ9Ttvh2WLL/hZkyKDYhdL8Bz0YraT/KiVgFoK18tZjG24PhEFsznGXEDU1GE7K1E0cE0uYV0rtyLPRV8XW3IlevliafPssL9EQIYT8o0XNSBSuzO7IdcZfwc0QspkBm5r9xV4rqP0z5wlf07R99JJhVJK5XY45jbyH73yBri4/SrpF/knejXP4c19cj+0TdRYwuI4LV0cZ0jk0VKEY1Mu02VHU30TqPIdIVgpil75n5RvGjgEefPM1SqZHtcv+RMF6I/v3HnmsbS248ZU0sm1pQ90BuaujSfZMkjDgH6LtHs+Uz0rr0hEppfpJ5QlS5Ysn1iuROxYVTu58fpInjRBVR/BMrKGyWTdocQwvrzV2MkYMVIT+J8QCSyAojrETa1QmHBpvG1HQAWkSbQEFa5RliIFvHvVAkKU1KtogbOSxRr+N4zV6tpUK46ci/wGlUqbapc4tQdqMYUKORYsbKiOx4jrahZxDJgpsXJ6UmgLnArrpVE7ktvaBzd09jq8y/G6rHFfdenk+m34AyE/02yfkfnISoABXsA7u4TlkxbHs8d4xogTFFjoKMW45wVuOIiIZLBdQW6KNLS2FouI6fOIVS2Kc/xmorGLYydJ25Rn8j3GEHXuSRfbnQCSYnB0ejWLpmxxkGh5s8RPbM4ZR6eBf5YeejFIdCU+Qp0XOVt3coJUExVeuhLE7vY4fDG2d8Jy78ZIRnDSo/l7guTsE6StOMMPhT+YxQg/HKQXrfFlsbmg+DCricg2qohyekUitMa1ylH6cdH3lE6AnA4R8vKdL2NhwqbZ/CDj4GCppnhJOyzpeqeFCb0mMVvhRzfVUBAcjPFtQAJ7MOecsk7hgvB6FV6Tm7f5wY9TWOzq1V4nGY8lxqJCH8oinbLSUZJTXE5ZaUCY7RSqOPQjZCv16XQZY6QOp9jLIFOcp6b6XjyNJXQtvxtN8HauJMmRzp6YxFSkl7ViuWLLLQ9M/P12LJFuz7VOlmb/eSR5vw9x1zkfKp5zTo368yRPx7JkyXKpciWQUNt4ObnfKvk6Q5qKrkjZ2BYhGn4tsgaZqonqMYVgOeI5GqOZOrr42yURBhzYmH5jKrLFQEmmcV2lTomjOtVMqnlhria5TDRQcu5Xuh7ywqaYuQJ/+KTfWoue/daE7DSlAq1hXE7mcZqlhR6BkjyRBYldOvPhcE3whfveGqdTLroJNGsvK4SEcLq1WoKUZ3lqPEMi1dGY94FnibY49Z3u4Joo+jgaux5Jiv7xtjjFNsDG2+mCKbnD4GBNyytu6AAobCvdHqNow4LIKB6fPicnA/BxLoIYENH2ZTWoi1PIfnN0qLVRKOrCYU3vdF8w87JBn50/NxzDmw5fDAdlJJQlS5ZLlquBhNYihw+6OKenab6h2RnBkVh/euSkYIItzvOBYFg0kMGOvMMaGrg0XLKaeSVFL0VZaBjDiIXxGOhJkzH4CwdSmX0Y83iamzUdKra3TlEJOZyOSKdjiAUOpinfaHd1mrNOirzWMjKiPIemeB676lhsMgySM+SWIgj2Zcz7BZ/WdrJYhb9PmbQMyIgJ6XCotBij/f2wrDF2DWEM+LUSifsrPMei7HdAkn4yRELRLEtLKzFNL0aOUXo/yq8VQ6TQu1p6rKIt3B+NJQauKDXknfTN3OFcLO2VDNUyoJfYXWzonuWsaHmbAVmVXl0RkjPrsXfn+VRGM38cHLmIZCSUJUuWS5UrgYS6zsvpcSMjVKtrgQYazGwr9ZEnF9GKW4fv5xoOgUzKTt6lRrrTkoUKEU5AK4xTTzYsmOK0dynPgE/CElYJBOejebCgipjq1FOro+0RRxlqc3HcaTmdxTpFNkw4to3QkHJiuCChGD6KYRBMT9s4JXtIrVlLHNNqzGbpvdOaZNNTOGMjXy+9nCGx2/wROTisA6m2RIezVGuzuKMA+bBQYQxtwfYy8hx0Y9DUr+ADyXHR4l4Yx01KR0TkeZzT83R8DVCINAzHiGZAtKHJ8s0Y8rSklc1Qp482+gs9zNzHRtvgeeBDL+mT/kWkZg53G3ucmN0tJorHPttqdp5kJJQlS5ZLlSuBhMR78W0jrebYJLLAKhAIEcVs7KRk8q4FrTJhWaOQ3841ophwXD1JrTC+SD/XGr/XM7VoUT1ak1TbQVt05D1Yniccp1EC1NBAX+zr+lEnxwiFYBpUJlObHSAhGoo0bnEMRkRM6JMpyhf7H5blyiu6GCHdiaapIEKbICHcrEjatsnC6DvFIWNqjNVcZHECx9I5rq9cFpbqwJRCC/Zlgr4xooTFETl2RRHPiQUqJRFLd6hfENR+Zc1n5BETyxFRUmo9sg40iih4OwY5RAjFxv0Axsb+WovUOayLvT+DbkJni/TYAYdlmjSgJS69OaHXGwObzgU8xbl7Nh9+oaOzZMmS5ScsVwIJuSL49xRMHtZQY8P3h4nL4CVdjwo9llaYE1hlDqbhlsbwS6noy0NUwpIyqo7AK2BNk4QVEp0tmH5CNVHKrZRFmj5UtT34HfIG7WlYHj9q5fGPg6mtwrkjoJE5krUds2RRTX6JGheNkzOhGA7Dudh/JmGPpX/Rf1gMqykjO1Ntr+EPRnMTaaxWXtZAoAxNoQJlqEXFwoSwLJ7hedUzJK7DfRMt0ipGHs45p9Atcjy0eqHbTNJGayq5L32WhuewSr4T1d6D8Aaxko6Js+k1NrkkbyRxej46PSvdMy9uUFdyP6YsNdGrtfZpIPSACzKXMoSUlz76s91Mr+UtGfccyUgoS5YslypXAgmVhZPt7VrGbWAG1L2D8UT48rZM7FVE/xpaZR4VTNkJD11Yx7bAexSqNc3FNXAUnq+9pFbsR2FVj9o8UksJNTO1COOqGpOytZt7qYBO6FNEvxnyMR2KBBJBdEADWhYalxyUdMGYFePIM3Vrjhs0LxLFT0A4OYOqYpCs4VIMv9CJp4GqZ51L26LVcU0Ojwnx2/T+yxHvD+uaH871ygLRO9kn14zuwdTEfJZpkjm9P6vJnR/GXBnuRM8d/GEvsdm6FLbZk/jQJF0axDGwktl32LkhfaTnuvRcywUZJKevggkK7hehdANElM4MLioZCWXJkuVS5UogIREn0pXikca1K9MCfzoXLQkxvBRADgXKDp/eC9CoAoLYvg0+aTcQSkywruiGXBD8Wqh61RO5G3JAgwJ51nuZKoKITcvBUP3QEuZkMmFKWKIRl57Le7VWDfwRp/apJYhWp9lMZAmuh6iRvBh9pvb2gDRxrY4ex6otjRZNuyh1FTXpEtyQckToIH2RRkh+NoGVsjKe0bRiFr3ijeEPF1Oo0D2Iu0z+lU2llUSixdBb2Ni3aVk1zkNM0HzkDtP1Afeif0SUMrDqDdqy3KPtjE+O6yPTQTZZg4zsHVvWxgYSbISAtpFzoM9FEVFGQlmyZLlUuRJIaLX28uHHS6npRsvcP4hH2poERMQI68m0kBqR9rMdNAIr0gzH7N9CxDs0bicGWVDbUwNjc9+dRafD6kGMfXaCbeKL1B3FaIyit6SlSpOZgbepcD8VEENJC5tN66roDEvjLzTdK6SFGmd+o2KMMYIP1XSXSCjt/8Cd1uYywjiMp4VUQK2nT1Eq+jQsp0B6bYPYsgdAqNfDMy5h6SyjmSkcz9vDshSv19MEbmq1TLureZsIbhVZGIKHGQqEskl3u41rFqV4cw3LlyR4YkCxmXPMfcRS2OY9w1p0f3LnW/UGaFCSPwavspjjeo5FbvDjwUKtdueRVs+WjISyZMlyqXIlkFDnO35ADEYAACAASURBVFmsFrIGMbBGPlJakA5PwvbpMmjRm+OxbMPHZXoz7NtHnVcWQ6x3UqsYeQIb36Jfd2pVetMWhXjmJDIRxjq1Jw+jSMiaVNI5PpO3Sy3igNAI/raYbwcZBWsiIibop9nC8gVEQE26Xo8LuXaT/YPQEodrM/eSp8UNRE5rxqY1ieU1GbwvZHVGq2TYtzXluKOjgDYrpJpdnjG9Kzg4OUd6qEc5EKI+enJrgnsiB/JlYb3U+ECfjAOzP9J/qs/bbPS3kqFuj5xc+kAGVIrbxLOk74VeGrsZ/7dcszQ4boNIj5kX1Frs9BkS1aqVS0sS8Vpow6BDW3Z8U7xajDuzMM8goBxFnyVLls+SXAkk5ARfQ5ifPJxj1pwbIwiqgLftk486qWHVamGNmSD/9C74DhazU/8ZMyeOU2P6OLhkvfM+JlC3QID5k+nhDbUzpjuqo4ZKr10DCU1mTpAEQK1FW8izMzKlmAtjsYq52Y2FRNEallXkXSrmParN0sSSkZ9yEeolS/osnT4N63ffaaVFHuuajj1apxLPhzmkTUnpmH87OS0K49e8F1t+pzIJ7q3HrjOZEhQFELl26fnnJORJxG6OqGAzlutHq+sxg5SJRG7cCgtjQ0SaInD6YClKY7aJTsSpVzksovrucZDI26To0ZJJ2m+Oke/t0EMt8bWJAHtxyUgoS5YslypXAgmJhEyG/CaWMFnph9eQMWVXaCmfB98LZMPOnXDudZQfrhTygB8gomDlDMOtDPLB9OJf6BW8AhI7fhLWT5Zhxwzlk28iLzIj9l0/ElxEauRs3t4XWSHL4UjzH6HfzACpGR7P4Q+IeODjQ7+pUrWrj6iEXtSl0XrMNChsO9W8ehzj9GCtvPuDsOHJu62UzIfENjBWizVRFvqDfE+m6k70wSI1pzmSuYze0xpPZriH6K9lHqpeC+8APcFNsUSfDILhvcz2GH43MCWmx/fFmaWYbtKrvEkhd8l8SabfPgX54lsvHjzqskk9/wuNPUSMYi/qoN+GdtVWMnH9P1JEOuBV7Y29oGQklCVLlkuVq4GEXNDohVo37JcVkdbqMlvIGpqVFh7VkmVfvYl+lbVNTe2nlw5Lgi392HvVaiuUcf74g6BlHvxx4KgWx7DiwTep+IVAttx5HX1hkpyCaAe5giov1Sq12IzoSczc2MYiEi1S6B95ESKmKkUU3dJF3oUZBBTpEEEkl4qa1hRmXMKy9fDDcN9HH4fl6rCTCgO5xDkr5uFmlRNsP1qEc8bw0r715fTVM+Fh+hzqUe+Zsp8gPKLDdNpfMc+6Y+Aa83UjtD+26hQBRJ8cw3swptDwORb4WAzge9vUoZ6oCXCwAaJm5RjmvKIPFbk4ZmUYeIz3CLWOfBLfL1ghRzVryuG3NCh+5pLtm5zLh1Yxc0g6VC8sV+MjJOEF4odkxDSpHZPZIyQDL89o0kqF2uf7r4UnduPVcCt0/48F8uwUhGZOQtx4fRHpOY456QCP55x+fYB66SBmBU6ALaYe/JFev4VpJcJOWMyxQwhDVXoNlVDinP02gYMDK6hx1KNo4i682GfHrc5jlFxkShRr7mfThrTnfR0/RBHFY3xMcR/rSSmHSM52hjS7LaZnHDtOuxigu0TyMw0lMQifF69ZbqkoVIHw2UUXgnQsmOhOi0yqAyS/bOkN++TXkhK2NnZ34LxnxafnR8u9t/y+7uMHjwUHNCEdU49wymRKMrUs74TfR9mJdFRCnJpqyag0EHqtjpqbA6L7H2Zzg2J3DWakm39yz5U8HcuSJculykshIefcvoj8LRH5lyR8B/+yiPyxiPwdEfm8iLwrIn/ee//keW21rtWcHVPU7aFZt5sGTXHzVlDl114dyQ7Sn+7fCLewhRronXEaY711DSK1n+0u/WzrlEWcog4m0188QWgCpmdC8zOaOEFakaNjTBW3eW1oG00r6sWbaZeFBNFlYDOSGyAioJsGAb2Lo04KIBbeOpEBy1crIkAb0Scx/LEEEX//+4FFXx3htk9CO6cnnZwiTGMOpn1tpmPVCFMfhN60uL8GUyTfIWiZY8P76pX5VqSmsyqiFQNjjSOezrxZ0LIzaEGnRz46MA5CKcz46/OR5Lhzxbme8SM9iVM7BxRTmn6ztJQzaV5KdUCMU8TGhKJoShdNuMd7BjLS+0xnBAP807/BcxCODde4GC398kjob4rI/+m9/7KI/IKIfFdEviYi3/Te/4yIfBPrWbJkybJRPjEScs7tisi/KSJ/SUTEe78SkZVz7isi8ks47DdF5HdF5Dee1VZRiOzUpWxD+03hXTdCkOq1N0I3D95COohrpSbcOkPqjvlHQfVOEK4xBgqYTIFK6OhmEMhAkw2iT0VRR8O5OK5JM/sc5tEpNNdTBGse3Az7LfIQ7yLxTCpHevtEBqZhRS1YEvGw2KM+SRywOBMp4MipJYtAMFdwKbB8AGGIFkc8A6n8UbjRpg0XOQbiWx520i7pWAe+DBzEWELHKpKrLcef952iAabrVZKVTnfitT8DLa2lfVKoQfM5z6usWdqQ366UAQK1tvZYcBC72QeTUC36kMaxjbaScAzHSANtz0kPrDwVkSpdVIhosWzaLjqBsr8k4fmMfdoWEbWiMZtCRQcnQiRvnCctX3ae4+bz5GWQ0BdE5IGI/C/OuX/inPtbzrktEbntvb8bOufvisitTSc7577qnPuWc+5bq+74JbqRJUuWz7K8DCdUici/IiJ/xXv/e865vykXmHp5778hIt8QEdmr3vauHEk9C9pzOg3qkl9xmhVrpIeY1U6O7yOFBEvOgIMg58Bv8h6sZrdfC+u7KKlTqfkc/WnT9eAiRjQVluSdBKEKDaxkREQLFAI8/Rio4G1o/yrVMs451YoMVqRLvqZH1XtHGyYYk7xN0xARkR8I0q07KeCR2cxTxDBFkKxX0y60ZpH2kyEXNVKvHj0KnTs5DMtiHQsqNiailsna2Catfzs7pr9MxarWKYTd9FszaXO9uiukmleZIT0u5f2cIT6UI/KRN7PvAzkRG/ajSfWJZphqxXjBjoqIehua5D2DfjE2dEZUTgvdZolyg2J4G0RQ0nSaLpdPTx0ytV9YasoXIlOgZSXIDEeqlsQhpxUtgpKeIxeTl0FCH4jIB97738P635XwUbrnnLsTOufuiMj9l7hGlixZfsrlEyMh7/3HzrkfO+d+1nv/xyLyyyLyh/j3ayLydSx/+3ltORfSr7Kkj4cPDUMY6Gx29IgqrJQFrDJnsNjwS09L1vxJUOMeFh7yIQxVOLgDbUrN0Kbu7pgEi0gMt9jaC3vavdDPFc5ZPQUvgmsQKTDAddq7z9Cy00KDK/T34Ydh/QS+OKBf5PpBWL9xEwhhwrZoZQrro25oBemgOptlOpcfFM+j+PSAxVlYXSDadnkKUyPuu3UuJhpr6XcSDqFVZg1kdH0f6Vi0KCWSnkHba6lvi2LCTYVthn/RJ2ViWmLogbGIGktQxwDXLoaTaNJ/U8KI0g9wFumVOrJWNaK1keh4EvXask28hPo/6a1bzs4gOe4oo7diCwcsi/q8vuip1Uz5MU1/knKm2pNiaKl9Wf8gyss6K/4VEfnbzrlaRN4Rkf9IQvd/yzn36yLyvoj86kteI0uWLD/F8lIfIe/9PxWRX9yw65cv0k7hnIzHhWxPmHoCruawZK1hpbn/g/CVf/KxlxE+vyt46E5R/nlrFm6phpqhViRaOQVKObgN3okpJ/gV7yUjZ0oF+nFMkaz9CPNo8hqVzvnD8SfwpF7AirbDcA1yLs4pydAsA5S5+w402GHYPkdIyOEtcCi/GCDQHhOVGSuHpiylphsVsgZXpdCHPJhJ62DyvEkDl/+zo9CH08ehj5omVohyvKyJCMhfcCwwdkxNsnc7jMEYfBSDUtVj3Aaldr3dAw1rLDeQrsfxiPQ4F+OFLilIkM772H+mZzFIgEjJej+r/xm5oy5FhN7FwFSG0WhQsk/HrFdLKd1+TvrU2CenqKiilzzvrk07TKuZXtJYXRVJ22Bg6aeQTcfdWxPzBRFR9pjOkiXLpcqViB1zLqCesgbvAWsS/XIePwwoYQ6rVFk1WhiRpYw7JAWrb4VbaqugvQvGMsFCxNgm1cCMU6Pm6mnTZoWYHhAFFbghJp+n5Wd7FpbLdWqhWszD+UxlWuok3sdEVaBZToA2xqtgIWQA4tF74YB7uK/ZNhP4Mz4t5Q2YCmMkhayAXGi5KtG/FgipmKbWsA6OKAvwaAtawdD6GAhwic6fnHbC3B2jMdui31C4n4OdcD/7QEJTeJHT6lcaC52J1w1chJht1oucPkeuD5/6g5LCLrX+kQ9pI/ztDADQK9kyQtyuFqvUCsgD12svK6bmMBarQjkfi3TS/f4cxKoBx14UidESq6luGfyKizrjGW7vUy2PNs1LMeTHBknyjWXtRSUjoSxZslyqXAkkJCLimkIWK/hPtORLYAGDRqafReVK1doNLFL3YC07PQ3bp9DyE1iTilHYP5vRZTcs1OBCZUM/nFZ6HrhhOYGZa4bigctH0KwsYEhPVqIq+pKQb+ireWoaR2sRNCmQUQl0VcJa9uT9sOOAydtu875wH0qpgEdrRVNFaLJzGreWlnRIZQkua43zJ+DCRrSEIaHcuPLSAkKSGmmNX82tzwcktI3MAjOUntZih0x6Rs9dIoteMjflLdRiRW2foilaVzWi3KBeq8nbngVOSyqpdzKsjxgz+vAo4jHpFzQFBr2YwYmtGq/3SpSiKXtt+pAe+ut1O/af/B894JVC8tFwRu9rPiu8g0xsp5wbEWiPA+1fM8ZgRkQUS16l5wzqG+VE91myZPksyRVBQi4UcMOnnSWbHfxbWjUrYFFU6h/Tmqp/R/cC4XEEdLIN/5RXUEJnC9YZmyhLfTJ6kcwVLG7MU0O4MYOf0MkE/YPn9Bw8zjWcxxw6tKD0PWE9LW9oeraFKHNo721Endc4Ynkajnv0YRiLHfRhqveR8h2dkxgfx9g3jhk1qiao4UloCdq/5nMAEjpj60BKZVHouUwz66Eur78Sbn7vc2G5g9S3Y7RV0lJnsxgUZukjyurYQcAWY2DrJeICUtJyNql/TjxezWjqt0TRakAsF66vWcr7MZ4rogOfnF9WcRwLU2SS6I+AlSjL1piOlYKItlJoV/hoeY2ZkI3ljYiTJbW5uTb+QkS0mncL7ZQxL1ZhQ+4NELqoZCSUJUuWS5UrgYQKF0rfMPJ9C5aWFpnij1BieI7AqqbolLsR1VTgkaCaOmjrESw9xSSYtGgJoYe0t1pH6/R4zedCKwVRyfZ+OOToRujPGDFjTCp+8AosQeB1Ov3UR+1EfoMIYgKe6fiQZVzCOstBz1E08Ol9pJZ9M4wNbivmnsF9lJWIxziO1IkktFGP6KpL7cl+ga9hOlr6Mo1SRMGk9U0rsgZKHcG3q8Y1r78VuKBrt8P2bSDQ0ZhIlHxZikBVY2tagajdC631k+5QToXP0Pr0VEZV+zSKPfFrcekzFzOuNvqc68znoyiBr1HnojWPnBy7YS1tJl3reVRLzCpASOX1HB1PgkbGLWKDgmCNW8N5jDhoODY+WTrv1cpbwBdpZFIq6/KCVrKMhLJkyXKpciWQkDgJ3rz4JK749dXCgPBxQEa/1bpTS8NkHG5hwfLC+OIzj802LTsTY+oyPEHnU20fOCpJzqFvErX6AfIcFZq7Oew/QNmhySw1J8TYp6gtiBy2EaG/fAJNy8yDOLfCk1rB+jdfBNW13YI7Uk4A1rFVK45JzYHIavR/NE3n9GrMQHe3kJPpGBHvy3nKN6hl0UWvcXqF78Kf6cZrAaLV4BxclQZKFcrbpG1rtDeQYlmIlk4a6kw+K64StdAvB9cmOlDARG4o8nJaPNBkYaB3uOaOJi9CbkVLUBMZ8Rpe75NImOjOK/xI+2XziytXZPNGG/al7XyMvevi+ysSrZC2Lb2mklthw4rvHTbzsQXuFed2HDfkfOcYKa96MXYoI6EsWbJcqlwdJOQiT8AyxSPwJLdvBVPKyYMwqT79sZfDM3AR5Hp8cA8e4etcAwWM0cZ4N7TJzIpOY2gMQlJNvaGf+GTX8Pq9Ba/l/ZujpIkK24vUJak3+Y/5hByQwt710O/FESwmR5zvAzFguYQXdwuN1dKnCheh70vbeq1YwXl9CR8dV6fkgnXrmOyC1/kc7nea+p6cIYdTO3JSwUX4FnJ9778RbvoU3NbxYXhmryDT4q079FA3TidqMjJIxBWKUn0fSvYWhTXTFH7zdqOh+5yMzVVkWQ3drtRVijydY14rogW0XcXGfNrtuJ2oiqWt2M02Pc+WNtK78xLjzdQsFxbqP0b/M+U+OQNAI/ZxmFgz38XRVNTU9Ii7Xr/Vn+kFJSOhLFmyXKpcCSRUlCKzA5E9aGDNE73NyOuw/c7bYfvhG6Xcew/VHeApXZdpxsQOXEt9I3yVdw/CdsY4cW6vhhSjSbyPXqWqAahFWB8N3IoGp8MPx+jrXsE83K/zsXIC+jFDnp1r83DRx7AEeuT0YS7pSUU/m9hPkZ7vDy4+HTu1JhXqL4N+RccTrCe3p1p06wAZCSZdcs2SqGxeqC/VzqtAQIjRO/4oIFNa4s72YfW7Do5ri93tTN/SZSM9Lk07GH1XRCJvE/1VUhyjkeHKHWHRv6ZFKaY/trJxRI8YY/VvQryh+j9F/zd1rzHlnBWt0GKq5a157ZTn0TEihOiiVYx3kHj+S0QbzCNUABI1xkRH3lCthPRsb0URUGcGkpksOryztLC9qFyJj1A1Frn+Vin7O3jp8aGYYBpQzbgMb931z9Xy6pfCsQ9/HG74DHXSHczJBT5KWzAR7+AjpCZ4QyAq3Ob2zqt5lsd0JilV4c0vhtNJ3pipiR59vFwvVUQ4hgGpe0jdscK1Tu/iAwA4TQJ4QvMocpB4U/d+5/pIk+HzJR6Z6eGAENV7x358IEuQ+zu3cd94sZuzUkZQEO04NH7/9495crgPOJz6Vfoh1DS8mk0M19IfaLhmK0N4r1Oh1DIfCWpTYdaKTpV6ROogHYWZQ9lQBW9/ZySsOc1kkjzvBifrlfhemB91TKbPdzO5rd4Hm9fsHVOkZL0t90SSX90xOP21H/Do9RjacZ0aSdQCj31MdUvn0NU6naY9T/J0LEuWLJcqVwIJFYWT7VkpNe2J+NKuTXrLktC3EpndDvOTN2+EbSsEuy5hql+dhHNYh32FQNclyVqUt3GaBArHa1In37fThm0sHEft4k1Nc4XLmEox5apJR+qlpxyNpppgmnLjduioXwcIRLTNVBgsDKnF+DSVA9wDrhWyBijpgNDq7RSxRaSmUZthAdQlVdS04Xxc+82wwTchxauIyD2kHPFLPCPc7QLJ2Vgym2ECgyKVRYokvDotdr368JsRzmA6SUlPi2kqNBVrnP548xyGoQjoH7cAqTJFrB5PEKDouN9RtmnI95ST1x3qu2iQkAr6UDhn41Bj3XuX/oYiHE+neHQ5aNKZba+oQNHj9dP+6/utJaczMZ0lS5bPkFwJJOQ7kcXcaVJ6/apPEXoBvqSehf3jrZj6lQefnYK0foTNMHGvO5qyw3J/L2zfuQkiG8nryaXUIIDLOhKaREuMZFXOpDAqiprO5tbiHzEqUonDmNAqJStrOBjeQAhIVZHrQn/hfMkSwZq2E/vHM6epVT1CKjQDhJYR5rrha9hv3IcWD+TbshXvZ4EC3w/fD1DTIym+lgKiVuTYkOsBctX4W3ZOU+DiEt4PUIkNt4jkdi/qtbe/MIja8txd5xRtnJ+a1JDbmtBfku0t35n+YedwPPZ+Yi56n7QdSfMURfJOisKpsUADVw3X4wwiGpQXt8YKRWM91GPI+8hDhWWlHNW5g7hRMhLKkiXLpcqVQEJtI3L0wEvNIDtsHyMcYo2UGR34nsWhkyXmw4dIi7p6jCDXJ+GYJcMb8CUfIwShfSXccjMPbVe70DpAHlMct7vvtNSPBnDA4mMDJlWMio76wKgQGVpXIkWE/gAZzPYZ+Alkh9SsbkKUBusZua1e+AadJTVMoEuDdm2KT+2bhhykt6lWES0/7OTeu4HsWT0Ou9awhi00yBXaGgiTqWOPkLKXxQ3qV9AXuk7EvBwDLT5YKsRJyRQtn0x+RLmX9MZc0WtrMBbnrBvrZtekCEKfZ+c2lA1K/+iMSV5jqAdWPiIpy5H5Hs/EZ+r03vr91fcuzYCjlq9BqlmGLPX6oVe35bVxrpYdf0HJSChLliyXKlcCCXWtyOKoUx8GnRPDStOBX6B/ymrZyb0fIYQDIQSaXhQIaYVUHgv4LHhoqmNYZ5iqtMD5gjSw6+uw7vhCdgoGv6IfmHe3YE1GnpYg3khY0CFMrR0mh6zzPs7/mfDMcDqRV4J2QVI2pyljsb1kOaHkEtJ1PpZUNuWdNdWCOmjiWtg81mTpqTZke+zb8sTL8X0gHphVzlAgkelARixjPQ39fAS/p/vfDpDulTcDcbWLIo8eiKntQcXSIgKLVgz6UGc/LX0cDqBvklqd6IxaROvS0F0oRU2KHExhhA4OesZYhhNTdKKNEUHQ/0wRUIoN2KVGQ1p4LTwP8dJqSpSUp3TRAQh9SPk/BTwGrasjaEuE7XXg9L3me0FLbZeiqReVjISyZMlyqXIlkJB4EdeKlJxf49PIdKTUUx18YBZPWzl+GNQcy9qMcCdk85mknek312hrAc/qNSxvjmhrCWSBLs1HTiZT8jNhW2EcNlj4b+RSzdZ3pxfpWR7UCuKjrwq1tUktYtW7ohntA61lLOWSXCLR6IX9S/el2k99Spgo3pIA5rT1wsvykBZDoCemHsFAlvQih4Xu4Y/CA3t0NzyQbXjJa+An74Pot+j57gy6k3bQ9ziS/lKPNlxRkoDMmXvmrVoE5NNrEKUrouj7BYmIFD4+UxPPo68J1mlU4v3GYo64lkVM6lZXxnJHxuu6UI9ng1JonU3ppQHRpsi8/z5Zjovj6ofHvohkJJQlS5ZLlSuBhIpCpJ7Frzbj3xh8x4RMJdCMX/f8IJr0mJj2Aefg097NceAeiiMSSaAPyJAhI5TDWZ8V0gBlTQy8IN/BObpDh1mCN86zk1XtW9u5WHcaUE3RFjWtVQ9N2qZNoakWljZqTy0do+ApWs42XUNTgkIl03FavYCNFUcKkTVSisxhbVwehYdUbiHhG2Ld1vBYP0aq3gXG9vSU6/AJgwtyP7ncIBNHSW7HmO/OGXgb4hfdunrckT66lHfhDgakxvfLvmc8nPyIaF8sN+WUT0otWFomXH2SyBcS1WOMWNZHU9HKYIw6A0esJYsxh10KoHpxhen71fp4s7Zgpf4OJN3+opKRUJYsWS5VrgYSqkRme048tGUNHxOmCGgZtT2PE90SVpSKnA6+6Ato5hXLEG+FA0fT4DRTmrIrtKY1KNnsZmjYi6x1Tg4tQuVCrcJYMvIZZapqo9dpkDW0/3LhRZbkm6BdYIFjGteqTs9V7222HZ12wn0w9odlkpZeuR0tXYQzKnpSswW/WZORR2B0dGd4q9JFT/QK+mxrBo4H49ouwlWPHgAxPQLygVXy6CELXIb17eu8dtqXfv9bdYxOUccg2nzQwuYNrognOcOpbCTZeqvqqW76qyjGRYuVWpw0Ls4n5+ol7fPgtUjn0KrXM+mpVdSCQ4rho7yFTuegxv44xHcPmzQbAQ9JOasXlZdCQs65/9w59x3n3Ledc/+bc27inDtwzv2Oc+77WF57mWtkyZLlp1s+MRJyzr0mIv+piPyc937unPstEfkLIvJzIvJN7/3XnXNfE5GvichvPLOtUmS6V8gC3sCxei/hSlgukOh+XBeyhfI1SyHPAi4HsEljwBjXAnMN46AY58L57XLlk+PL0sdkWKom6ENEIiBFIURVBETWisBSOfOjToS5fqjVEPVPv5oJPLepPZlAjf3TxoEEV8fhvCdAHMcPO9V29W7YtrsbGtu/Qd8icAw2bkv5ASKh5PZj4vjO6b2TG6LCZUlvpHeSY5QqOnkYHvIcyHN+CP+iRwG5dm/gPuuoHxW1Yp3DqsnyDedS6P3IRlGOju24vjYnz1Ikx3iLQlj0Uf2z4LNjCyisvY6fyfUfiUxrLjMdJeIhF8mlaH6rQg+2Cd28MVnRX87yYl2R9iWiy8h9xfStXJ5jUSzNhufIy3JClYhMnXOViMxE5CMR+YqI/Cb2/6aI/HsveY0sWbL8FMsnRkLe+w+dc/+diLwvInMR+Yfe+3/onLvtvb+LY+465249ry3nglfyGb7sc2j1Ri1IgnXsLzpZ0iu5hQaFvwZTrVZt+EuVDTUXNAbLELOk83Q7QCeWx6lGTnP20GO3MOYV+i8xfaqWVTnHb6KlhWvlpYOlb35CTgr9JzLbDv1ion7ZR19muB+WYsF9PQUCOvwAKTZPRCqgwxN4NR9Owr6nyDZ56/XQyN61NIePjX9Saw76TK1bTEW2gNiePiUfALQKv6Axsh3MYTVroYmZoH9Nz3bE+jXYzhfTdX2rVvo+DDgfjC8tjbbcsmp9vY1oEop5c5JbjGjEWtYMzNV4Ls2UgDHyvWwFHEcW7PSM/jcQiVxckyIKr+98+l6KL/RYZnawVmIWP/RMpm8SWllPat/jm9g3tarSem3QUiw7fjFs84mRELier4jIWyLyqohsOef+4gXO/6pz7lvOuW+dzg8/aTeyZMnyGZeXsY79WyLyI+/9AxER59zfE5F/Q0TuOefuAAXdEZH7m0723n9DRL4hIvLGrS/5xjnlPSoQCasFUMLC8A7OqVZm8T1alea0DgH6tBpNHJYz8CK3kTS/3kMuZEVfoZ3xTsx1rf4bJiudWMSTupZEC4XJ3dw1Iqs5M0DiXuEnswKCaJmnFyREhSh6WgVHLdsKfVohOp0e4L710VLCmDeUqz48QTwdUEj55QALt/eI8CRZllqoEBs8rW4iN14LHTr8eIlzkkoGtAAAIABJREFUQn8Lx+h+tDFOg9uIMuljsgYSasGbCYtVFr2O8PL0Nsd2TchJTZ1SJsInYa1myn946UW6pxCIfmbqU2QHp0sWCnsU9Tg3KB+l/kJEPGzSFG/UAp32/vmM8e6vVp3MgaRrTgWIlHkuzqnU1JamcRiUuDKlwV0X+6N+T7QAmtzqrrGNPVtehhN6X0T+defczIXR/GUR+a6I/H0R+TUc82si8tsvcY0sWbL8lMvLcEK/55z7uyLy+xK+s/9EArLZFpHfcs79uoQP1a8+vy0n67VT0890G/NaVpygdtRUyL7nM4E/iBCgerl5dz+ohpuvheX1t8KOvdepooGkoKmphUZbIiPkM2Ie4VjtIJ0bD3wwjIVE+0oN3Xo5AwJaYqnuyfRTmfGWgTrq1AcjZkNkZHs6t5cyWo2WCyJLcBDkuNDve+Nw8dGXwOcwj5LmH6bFyPiUdF5mKAs0mgQnr8UDZBhAGxPcB0vLtE2D+2AEOBARNHnDZ90nYQbJlV2/GzEOiuOvnJElj1KNrVYy73p0THp9A0p0TM/Lt6xAo8cxWUSscKo09zG4GJ6DEmTp/bJk8/FTL6dPwrjWLJ2ObBCj3m9GRKThpclPKgxhXwz60oyNXcxGwDLQrPjCfO0N0eLFkNBLOSt67/+GiPwNs3kpARVlyZIly3PlSnhMi3gpxCsYIPeglQygVdWZuRWBUUzO5tQKYZ2cAgl6xoTN8WlfIRKfbU9gDZtMyXvw6z60gMS5OvrH7eopDW2j0fVhM9FapdyF1yhxVrbo4CNF9xhyXuQ7RtBw5Sidl3PuP0Xe57MH2N7Gml+tWk7QNpDRgt1+NzT2ZD9c7OZrQD5amQSXKmjZip7i9Mq+9bnwcB4uw42MapTjRs0yQeUPVgOZQ4vX0J7MDFlaPk1kWJnDxIgN9K6xZFG8QiWeTwuWG/JEpm1FmBpHl+7XIoJtihp918vN3bEJWFW71FKlHIvlkEwMIJHeyePwx+P3WzlFTN4EFVHq3XDR8YRZFnCvdswGntRmMz3f25i90ZaMbpfpD8IPkOuz5Up8hFwpUm156UDWehjLmNSMvk9M6bpYeDk5QokfFtfDD60kIQrS7vRB+DWvF6Gx+TxMyyZ1mGvd+Rw+OrN06pG+CfzbmF/5kVRf99R22WnSMyxBGpZ1rBPfoo0105ryw8WkYAixYKgFa9d3SBSP37RMt+hcB6L3zMfCg2iTqWGZgpVOlwWmone/t0Jb4WL7LBipfcLt9ZJx8WM4AcH/2p8I5zpTkugQY1JPkF735DTcPyE97o8fW03G34koSTx4DOmY6Wxl4LDHP7iF1+Dx/UY5jWKKkvRcLQbAWTSU4NkZp5MMZg7765mT6W5MPiYSv2PxvnAfGoyN7TpVSo0Va5RVevpRWD+628oS7wNLLN14O/VTiNVb7X3asUlOUzN820WnS35k1lT4CF5mkrX5sVxIcgBrlixZLlWuBBIqCpHZjpMp01Uuwtf8FF/UEyQiOzoKqmIxb+R0ERBOA71C7TcG3KjB5hVAIyVQwPJhaOvutwORWlUBEV1/A8ezjHTR16yUlNjkXw23AyIUasJP1QzbHk0KKWByL87YFLQKHTNJho8J77nEGAF10RTMRP6jndDO8cednB51Sf8qLXUNbU3zKzQwg0sffhA27OyExkfqZEdtGkMUOmhnpoaocU5Fp0pcexGAj8zuwTUCHadpWE35nC5rUjofI1cNQa6uEcZ1IjU+xw2dEtMp2upN8HonF3aDiEQEdIqA26N7SK53FLazxFGBG5tdLxUV1kB7mraFtAPGn0UAdFDgqKq1KklBPGLIC8Z23mlhhwLTh3aRvnvOIJ9YMjs5rJduBveLd3i5FsFMW47gzvIYZLimYzkOv6kZK12+oGQklCVLlkuVK4GEXOFkMiukHqO8MFz3P3o/TDYfPA5f1sVp+BR76aQRZjij9gvnrHAuyTI6OJ7QWYvqCLzTk3fDDlRblvF+WF675mQE+2b0qk9NqhYhafK1ftrQ3n5q+6IspAWZQs0zIunec2gMY0F3ALKAaMOk66xROmcXJYIe+EbOjkMjM6TXIMfAUsyaElfL+Ia2nn6AINNXgWrqNKxATfeV6/FieA69feG+gFBnQG7Q8hOw9JMJgmvhQkG+4QzOm137/7f3pbGWbFd5366qM9+x7+15fnb7ecKYxEI4iRSEMxiEQPmRyChIlkBCkSxBokixLf9A+WEpUSKUSAmJrEDIAAYEJFhICTgmEX/A4ISI2H5T+409vL493L7DmU/Vzo/9fatO1e12dz/83jlPby+pVX3OqVO1a9e5tb79rbW+VVgxskhtI0ZrREYZVuZUKaGzRiKbsJftXqZ81BsGGFGrAMf9sN17LbxxyJQEtRlXEEC3fjYCWpRpaZwiKq+lPowoijfcY0NIcYdtFUxXEzn7O0RCByXiFs/XlVxuUX4GzBHtJsynlBRX3c+EBVlOQ1Czu5Pj2rfC7+nmM+Fix/dEsBOx9cIxWpefLEQfkVC0aNEWasuBhFwQ8WpysTykpsfdO+GJq3Vuhzqrw6JAe6oQjZBC9XnqiTRmFnIIm/4kuJmDIUP2r4bPC0prpKt8ul9qYJ2lt+JSzINWT11GbWoOoN5cUJG82Qgo6DnHA4ZWV+klM/FKOkc1HC3OyNVqQ+S5O/Se28cTjMmpTRk5mXHAbUbDZvJg9N4tRgibK2WawvxlGY3AMTXbQKvHsP2Ux+J3hOwaSoFY4xySC2v2woWcejog0x4LiQ/I++0z5DwZ5OiymHedKLVJ/iuru9Ci6tWFTqbkR8SpJGpgYLIoc9IvNVmTGXkQhcMPbhBZS4ztLstt+lXEIa5umgMbJ/UbDe+peFfywUNynsN9cYYcC6+zz4hb/z63avCpdlVIkFBbRGgxtangykAywkq3EDo0wMTrzasI6HUivpf/zxC3uWrIKZWs8pFOK/zdbq+HSUwLy7x9LItIKFq0aAu1pUBCcMGJzbg23rsbnqStmco4gmuQlEdjNEHOR7nkMcTiT0QEKNeilv8hGdd7d8OjvrnHz5lZuE5PMr1bYEwvriiFiX8punJEzrW6rnYW8QqfS7YDvizWlfaIav469IItlW3w87KlMfOgavyMxqS1/trxFAmrXcUd9Bl1dLyeMUsohDTOng8n2zzBeVdBa01KVpu06bCyprnQAMOm2bZLDdfRDZ8fPx/GtLoZPti+HHZk2hB2Xw5w5S7bS/uZR4/NCfZPh+vYuhxgxtqWIqJEYyorYPumMXN3hDSsyFcUlxJBM4eVY4ruKWcn7HtIiZLDazU0si8JEuYJ7ZKDJJclud52XmAyEloMJ87Jvc343T7PMeQxc/GbzAPqE4WNeR9HfD1lIfK0KCxfywTdlJxLhC+53XaHUik1zmumCCnHdJMI6IU/CbzsztURJkRAbYZLlRi5sU2OkTleuTKEH9MiEooWLdpCbTmQEMITOWEh3MZ2eG90gZmhd8LrGSVQe92G8Rm5+A4+ypMJM3IJkUzIShmuRFsjogLP8E4xICkw37xPZQ98VGfMwSn7EvK7ugh5F6V7SKqB6/Ix8ykwBZoq8GSmsRotCgGp3EHRJWVjl8JqPFetSNYxFylbTbCi/JT18N4KebQVFp32NgPykQdrsE2PRPaTuRKWebPW1KlDkyjKtS0ZRzuFSx1o4sOmzezhtFdV8r/5AnnAV5XhrjF4TJgdf0j+ZbAf5uzcB8Ixjp8U0iAapjcfMndmzNdTyacQrSgS14SDY2KWazC3iscaUixudE+oijkxzBM6YK7OXcrXemWGKyeolWFKNKJMYvEu4qo03pw8zIjHHB+Ig+H3OLUTZmlnRMl+WiIZ/Q4Gu2F770aYzzErBtZPhy+dPisOkmiMKH2Honjf+ioj06+MOXe5la50KIdz5hyjrkJXvJejJ6OEIhKKFi3aYm05kJAHUDjLizjDiMnxc4xk3QyP1juvME9oVmZRD5lxLMH0fa6rD4VilPrJ/JSMaKvJpJ0ua5k65AI2yAO1VlOk/L81n6sMeE7w3iJZ1cvS/qqxmdHbFDOPjJnQHW6bREA9ciXiFFytPY9Jl2gk4qE4RuU5jQYFporGKIdF4v4EIZ1NIh9lZZPXKeVCdBYiIld9nSQOmXJZTPye10wvPySikJyt8p2UK6M23vdvMPeEHrxI5vxjKhnU8J07zzISxGM2P8xsd17XbED0wt/CmMhhzPclKGe5PEUZ5euwRdGQ0bnDO+R8DhSZIk9DbuiQ+40nutvM/1IB6zTFlOffe53om5xQllXnTD+nKbmfnAhJqH9mxySyc+V96KyG7wrF3mYdoETzdLGDW2HCM4Ly9W3ybaokeCZM6uvk5EaH4ThZkqLDqOq5y+FHuXqceXRq164/h8mTYZuIhKJFi7ZQWw4kBACFg6cLztrKlyBvQomJVdZ3zYa5SSaM6YnuvRS+c/1blBm9z0zoUVVHoUNmv9sNKKtH7sWRezmkl+mMPTJ6IEsDovdQLo+vi2RJikFRGnJBymbVGj7tuLIdD69R/ETG9TWL/EtvyXOIpzHpUglM8WVGrmntVIaEylUHRION2jmtGZ+rXmdRBY9lLVwtjdjDldEvbSWhIq8uJKQM8JqYltBLj9xSnxxMU9nnmbOM9TERnZDD/q3wnxvM4F07TUTEyOmYvJ84IeX8KD9Hom7pikemjG5OyoCZ0Ye7ymYm8iG6OiAyOpwoB4sox2RcwvW0Wok1pBxcVwp32GydCuPVfZAciFpDKYpmUiyc40yKC0zzb7ccNi6HH9eI/NHgOn//5OTSXlXG+M7L5aoCKBUU+teJItUMNC9VKU6cDwhofYsyIb1qVLgwuYaIhKJFi/Y2sqVAQh5Bv6UuVqUcmJTvd5jfUqykVsndWgmXcLgXFqbtu+H1+jiQKlZMjGrVeYN5K9LzGdyjJ+O5VlcTtFmHlbS19g77JiX0CRsiHzVUnCllRgEjIahGGclSs0UvoTNdOqN1aVO8Ba+51iZG3EpiAkTVvCjXddgkp7ayxQHQszbaNfhibV/4SlySXldLtuZkUf0RAS4xI5IT9WrySBSiUjNrSknO6+SlcL/2rlNniB660XYWnRwNNaFhozygPvmMZpd5T8cY6VKmu0Kj5C4KkmXjSdh/81ILK6ck/hW+M2D28uEtqjUw90hc5IBRsgnDUorGqkK+0w4X1uqWjRoOiKJS3qstIjdxPiOiFInO5Zw7tUOXKF6P+Vvrp5hzdTxBthb+/9KfjngM8DvME1LkivWXg0Oixp2wva+o5CTMw3QaXktkb/tEhotPhWtqroX3FAG0RpVSaYicULRo0d5OthRIKEGo6K63TbY8XVsLE734oqylEnFh4vLkVMj9eHIr8jaqj+qRe7GMa/IGh1wkHx5PsMrckabqcbIqcjBIYFKr5HVqLYHVhmUsBT14k21tylNxHH1yWS0iNd+oXp+3dipchytCorniNnGwtNiUldWmB2RHqHI9D9MnP9JE0PbzZd6S9rHhVvkxfcV4MW4bvKc9VpifuhI+ePmbwWPngxRdahQJQQhlCSLNKOCfUM5Q12ftuAm/Mt0P5sYoIrp1ITNx/9FhFQbqN3fAHK9DyjGMrR5PJBdrAKm80CbX2Oolxn+1iYR7vB61l7rPSNutFxnRUtspqhe0pT+k/C3+drcuhQ+6myn6zKEbEcENxXuJbyKKbKkRBOu9DlmdkLP28pBcUEq4uXU6wJ33fLiL1eP8m8oUTVVOHXPahPTz2h/yIywioWjRoi3UlgIJwQFIkyP5KUfa/4oj8s68uNry9MjfNJkAMZT3U56N+BCJ6FgFNdffqvNSBGVUmF6wtHPreTJmtda7lilt7XroIdTUceLLY+gzjmc6mHK8Iq84/loLZEOND6l0B+Z1gqueq94E0PSH+dqUImvtbFzteH5un/KcVZRoQInIQpm9M15AxrNKReD8d7c4hmC3X5yUrbKlI875blC/Zu14OIbyVlTrVowVrQmbNjPxW70wqd2TRBJriUWcrLU3vbqylEfSpyLqsmgYJ0UIo0U01mROTaPlTPmxR06zsxH26ZP7ufVSIHAOmBEuDm7SogImM5Sb0DF9Zc6SVoLBPdWyselnocxv8LqqUVjPkOMBayhHVEEYDMIfwNZ2gF1XvjsksJ16KoMjAhMQ1e9c+lmZmjFEJBQtWrS3ky0HEoKDc0npqevFWK7+0pXtd5hQssEcojHzOZSlrAiKIj37rEOSF2okQlD0Umyd0+klxuU4oRFDCNVxlbpB1eiesphTXleqyv9piRBm6lygDhgMr1i9lnSgLWtZ6KQambPGitK9npbXrs4PrZ7OoTwhkSYcr4W2qtGyeqeJUofYl5E0chASMZgNK1+17idCnJqbBrO1td1eD5MtBLu27bBPDR91ktDvZPNiuLBT58N2nUqQpvNMNNPZpJ51VvXYLfI2WSMx9KSonTqT3KOec79f46OkqaSGnY0wbt3j9oqITGdzIyRUcPy3Xwon2b3OyC5v+tTQojKsFVMM1lStlqJSDugfEkGzNjLhD35ILkj1XfptTskhHdymbjshX7sRXl98T4hWXnxvQERrx1PM6l1KhKCFznV9o+p4H2VL8hBCRRCsXJbxomztUdjOthrhDVUh6PpFhhwlQD5Uyj8fSpQjmJKAXGmHSZ51q+RzcyVBS6Fe9ffW+Gr/UTKdsLtC8kpunCgc7yVmVZgkR6LiRBHOjfIhAgAp/zgli2o9zvWgmEr+gcQqQ7ODXQ831ZK1CpvTXvU6NN6ipoNq/b60nNMyR5c79thX76sdLldYNNrVcoTlJ831KoHeYrlHd71ctgBlsuXmKSaTriQYXJRj4b3kQboq/uVW5OpYCYRaFvB3pOTT+jIimf8r4L79vgpUWQA6KX97QLmM7PAJ3yIh3WX5RGe9LGlQoqZy+EYs/bh7TT2YeC/1oOYNSWodT7W8X9kQuy/BMo+cpSF64JmaRiIHydFzDMPdcG49ULI0fGHrbPh7eNd3hWXYBslo10rQsN5lehhVKQIFW+oFz4+yuByLFi3aQm1JkJAPj+pal7qSCK4uD8KTVssTebPwOO5tM+x6sYoo3B16mYkOSjkLeh9P+L/K4432PA4JWddbWtYI6dAj2HJG0LsaRte5DiSERfGq0X6O2VQIjl6eMFoebXIYPFKjrUS16lLKCHXO2cHt8Po22/W4AljZFEIQ2Uq0p0aDIikLhfvD++oHr2VNw8LpVcJxeFDg2rNhgm88H86rXdZIph47HbYnLoWtCi2VLKovCI0oCVXLt85agvYax1MoYROV69HvRsLrQ86z2t4I6YFlBqghQedgaRa57q2+orQMbk1ChnsoGbDNAuRNkt2dDud+5Gw5rB/03VsM8x9WqQL12ZmonZNQLu9H1hPyqxLE04m3ItLxUFsdW8t30hZqxJiz0SVTVk6cCMc+/sHwu9u+GG66EnXhYb3ntWR1JqcLvl99/bgWkVC0aNEWakuBhDyCFz7S/7vGS1tynZ/bp0aeNth7+9QHAhnRWwsubu8VFnPuhUs+YOErLLEwvLQyjgOPgiL4akUkL44qPVAWgqZa08sT15o47irl3yNXUhj7QTfpoaZ7TJakXGjnGL2eakCMryECuhde37jKlsCvS0qisLW7iFj1jZdciFypZHPHJAwOd+iBGe5dpQiaoziakNL1V3O89gLLGojExL1JjkIFrC1KgK6fJgdEJEQAeAT1miXOZD/kib2Rw+LYiB6JAqbc5iTmxxyD0JVIaEtV8K6UB2YIvss5W6O0bEKUcihBeRVEk/A9RrGwdZZ/SAp46sti5JHkTSiST0CKgdCVIFEtwt3muLfPZpXrkJTsbFJYca7E+kSsz9QWvRN+VyeeCpCtKUlc8lXn3sPiVLZekoyNFTPDlzK/tYb2Pq3u+7Ck14dZRELRokVbqD0SCTnnfhHADwPY8d5/kO8dA/BrAC4BeBnA3/He7/KzzwL4SYRV9k9773/3kecAn7T1DLd68uK8i6hDIXOOXLfSQ3Uv8DtKWpRg+aDqgicmhBXeThseOUPD98klJOcV8TEXGjZaE9eSFCf0Ro7bTGjNAwNGMfZYOHuCyKBL1HF/J3ypJwRxQmv5cBCJbF1/hQiIXJCiU9Mix2A9fNbk+LOGkARRF3mEHSKf++RS7l8NUOIkCyTFVUwZ3hmwTdHN56aYku/qKJV/LmwMwCqIpyzeLIqs8rlC2ibDK7d4BCEBLqlKqGgfzYm2amut34TKHtSEsB7F8Yk3nm9C9CSw2+2S21FnY3FzPPb6yfD5BmU5lDTr5mRuC17bXd5rRWiFljRlU0Uj+dUtRsHOvCsc+wTTUPSTl0ytK6uYkEoGmPN6jBzppQ8FIuzcB8MfgkUjxfNwBSE0X5boeJsrFVPb36FutbhFV+77JPY4u/8SgI/X3vsMgK94768A+Apfwzn3fgCfAPABfufnnQRaokWLFu0B9kgk5L3/A+fcpdrbPwrg+/n//wDgfwH4NN//Ve/9GMBLzrmrAL4XwB9+23NAnkfwprYerYuHze1qjjKtekd9LinTNqM1DYrVp8xHmYyqnngmcfQ2ULAt70v0/ONx8EzHiIjErST1Vrom/xrebyi6J06im2A2lkiW2giTd2ICXZOeaXgvwJWNE8GNF0QlkuPsU5g8Zz5UPpUnTK0dTYdISMWiM+Yzqb3Li18NGrm7RIltesNDoscJr38kSV1yQqPXc6Sq35QAXbfq13QP1ahP+VAt8TwClRZR0RwKSXgrexBStu5Cimjxm9acktE8NUfscGxdSlAoEdJyrgAUEl9To0QiI7U9Sujee5IHIc/TZaugrtokKfpkbaFLMfzDnSnHWY2stYTgeEEN3q+zLOY9/+Fw75VDNVJekUpGms7GefI9TU5buDEXng6vNy8w2soonvLhhC6PyNTUZF5c6kouTZIdqFpR+xN8XHujnNBJ7/1NAOCWvUpxFsBrc/td43tHzDn3U865rznnvnZwuPcGhxEtWrS3u32no2P1hyNwJNzBN73/AoAvAMDlC0/72ocPPEIlEGZIqAaJalyRnvCNllLdiQqayulhMSFbTQ+5Xsf9BAXRyvB1ClcxotXfCx5qm9nZq/SGLfJQoiyUnVqo2Z5yYpplw0GhlTvK75FQmtbsDeW6KDLFTF5GWA7uBO6omVRLE/Kpt1IQJQDlhdBfePvw9fD53k3yNkNXOcboDrkLevfZVJnLPN4gN8SifgKzmnSH/iPx+WJc5T1c7Sdjou8CQulcZne9eJfoN+O9bPWq++kgEsOTPIrE5Y6otKGUGJH4v7Lg19qqiOYc8VxqLLDKnCylFQ0tl8lbtrUjCuzwry7jidtqMslzrFHW5Mz7A3qRiFhODk/ytPptpKnDqTOMoB0LMM+58N0Os/4bKtsohCbT+SmySRbiU1mNsvwBN3cjfPW7elslO9MHPQYebm8UCd1yzp0GAG53+P41AOfn9jsH4MYbPEe0aNHeAfZGkdCXAHwSwD/h9rfn3v8V59zPATgD4AqAP37UwRyqjHpZKMrXQhYmQzAPuap8TJlJXV2gKgdDkYoGa5f6jFjI+3fpLcfDHKnIBoZs+nfFt5CHYeviwWl6oQuMkKghICMpalktSdfByKPtlIciWQcek9tMbYQp+6AmdTMTp+dYyA/kamkkh52UUSLlkEh/6+7NcI4b3wgfjJmTZPVrRIsT8mMgnyYE0rQ2RQlyIoWmBLc4ZVMeS/VPWSZhNY7Fol/k/8T71QKkcHNfUg2WIjY8p1BLlyJlDc0Vz63xqqBV5ufOqfvfUMsnbh2F09SOW80pJaMhTkgRyJmE5U1UzFtUdGON+VYSpON9kVhZk6jqGAtyu2zJNJwov4xTwutqdxThAporEvmv8jWW1S80bihR+XEPzk1yqie0ZD1XtjUXyqtJvViEcfRkSOhxQvRfRCCht51z1wD8LMLD59edcz8J4FUAf5sX9A3n3K8D+CbCT+ZT3quMM1q0aNGO2uNEx37sIR997CH7fx7A5590IA5H+QE7pr1dIqQy67Wau5DXKHp3pB6NXoTr7yYF5Z0qrlmfMyu8VZHrMaqM13RXYmfhO8rQlTtpv5+Vx5nW30QryoidlRXVqrZ26zz9VjjGFvNPJLavcSu3ZIM1SrsnxBFpDU8E0nSgOigKclm5hKsYWVM1t3gwNT/sZuKOGMEjItrqCb2F/TZPpCU6JMqQxzFpW7Uw6gqN8HPOxVT7W4V2eN2gp3YoBexkzu5xldAQn6M6tKxGWpRUY41kdCWqUlW/7tVkRUlgYSMZWEUB20Rfym8ShSJkn6ZuTjmAc8RxTVXTxx00R+IW+4xK7jEbXbWHPXJEq8zmbrSSsolCUp0Tu9S8imLKSeHfT619tzduTmHmEqWqHk1tt3XNU8nV3tRdfTyLGdPRokVbqC1F7RgQcmrKh3SV55GnmK8Tc1UnV4q42xLWaddwDDXZYy1WkzzNhBKgBzerp17pOGRr4cWd+2wMx+xgQ2ziJg6ZgcwWMmsng0teXQm7teih++IVsjKapCaBFrEhV5WtKH9DiIDXQY+8xmjMusTHiVqyNvNx8gRNnn+V9VopkcyIGdKiCzaJ/saMfqVsEiBUM2Cd3Sp5kONXQpLMmaLAjWc5NxRYl5icopAdIrkW6+6EUiR+ptol8X26PmvbnTn7ASSZeD2aVYhbqC2Mv94q24Tx6yihzIGxvDLxSxSAy2qclerwhPwk+m5J/sqoZuTKN2Bl8opOqjmD0JLxFRI1I5+0z/q72y9QjJ5IfCbUzJyy1eOpDcBbDg/PpQy8mlCd/S0pojXS3wNfS0FBfFV7rqKeXSO16lD2uGr3br06wZNYRELRokVbqC0JEnIP5INq2u7mjUJ1uJ789ZwFeUVf2WZ8oq9IKY6IyFQc6ZoP6X06ANrMj1lhhvTrLwU3cUDdmvv71Yhcvh+FbEcdAAAgAElEQVS2N5/l976XzeI65IYY8ZoNSxQiNcN0Jk9EREFP1K55ecuSJW+wwcrqxHEdrnyhZoJNCr93T4R9dtj6d8RomPJpmox+eXJG8pqKyqgOqbkerqe7RkTYTK0WbO815lnRo7aImlaJhISIhoNwjluvcw53qVRAAusYubA1efuuK9tQ19QlNXdHGhDwx5DWfj/WebHQvS/RgVouF7zmnFGx0V5u+wBAwuuwIBPRjWk0kTNSBvZsAmREpENGzmZMIlJba0VCFV3qUxnhLvPT7t9gHSEjYDmlHkacu3zWRFYVWSjhiWrz7M+hjDADZZ5a/141Aic1UF3X+laC7jFpPVUjapr/wb3w/sHtJ8uZjkgoWrRoC7UlQUI0c/rycFW0k8ztdyRiUm9Pg6rX1FbC8uKZ1Cq5YMuWhB59cN9b25nVYwFGbRwPLuuVZ4Pk4rWrAX3kRDFD9nS+TY2dkxQLXyW/01ml1s/Qw48VKeFJyBeJI5nyc9NmqaVziOM6wZySY2x7YxXlDVciHCEefkciPo4fNNkIL8uqmdErPOapy6zifoq5PlR77LVTpFfCIRWtGzLLXJ2V2lQYHDJ35MU/oxLjCwFV5oes9CcSus5zXv6LYe7OX2xaxrMxOjVCsDZF9jvJaqjY+BCiA/Fws8Ij55xMxbtMqluQ5iiIDApyQfX2Nxkz88E5SjMAV6gDROH6nZepoEhF++5WVSdIiotTNlrssHmlCfVzvzajYxnmM81rfxems80NkZA0mKRbdUgUo3MrAqn23aOsVI8UwtQqQi2g+qwPbMyqY3iURSQULVq0hdqSIKGQk+NKGMN3taVXQfn+I3Mya49Xl9aiTHSLKT1191h4PdoN+00OCqtYl3f0jDxd+lBY5Ps8LKBfeiZUoXtmBati/Pbzwdv3vofN9pjfUYyB8V3+/4g+Lz2VcjBqHS9gERZ+b4Ucl/ST89ILGXeiLg3kiLbOke94ndEyopcJsV9vK3zxKapTrnF/ZzpKYZOlCdhHEO1uOMfaTJE1elZ63DtEmPdfCUjo4Dpzl4j8Gim9KDmJF0ZhTjtJgrNPh5M0rIEfh1FDzuKIcs2NgjlOqFPRNSJSX7aNLkyjmfd8wO8q21q8jRUG8vNa9FL1gSVHlFpOmnK8JoygNXf422OUKymq6Hf9OFG6pd2EsUlbauO0zjWnevgQs0aY/H0cUN/pkJyX/calyKi/Dx52OingmR1vPCqPrZWA47Gb7SfDNhEJRYsWbaG2JEgID6m1D6Ynrq9kRz8CC9W8pEy1TM4qy9XRIHw+Yr5N2kus/mfKaMaQCGeb3Mjl7wrucXZAXej7YX/p905Y6b53hzrE9GDttcKyrBNyJSYoqJ5lyp/R2lzcVqOKRuo1c4IHLin7hskLdlfDQU6Sx7m3Ek62xjnqUEVwbSt44OamUnl5HOneCHnAm+eX1o3pDBNFqn2y2hPPDqrzL8Q3ZrptV4jvRnj/7remOMGaPOU56RqVc+TLNG0Ac326vOq+qnOmc6sfGADrkGKRKzU75OfiYYQ0O/XQre6bUHuq3KbEUNQGE8Z0Hf274gjD52P+3pq9qqykIloZEcY6O5go58013ZH8H18LUOlXojyuIbWhJgdSN+BvlmNQh5gZo8abswTb51UJwGvjSXL1cVOeVuvJsM3yPITcUcKxzCmrk2slFLc/wZpIVpm8pbKMOpEdTKHZkeA3HweJcxgQfo761eQ+EYbHL4Zf17l3hdevfj3g5g5J1gZ//IWkF9TldTXBeKCbXCU4lRynokeRfs4a9XF/XY8eDLyTls6Ql8sxLT2l8rV+KhxjbavaEkjTLILUSjGUGqFEOFOF8OWyJK1ehyXg8XVD49fbdR/CJ0JRhD9U62qb50aSWhdafcUSUsXKlg9HAMiVEcqgQ1H7fc2/kvid0gG8nqbcTJm+kEhWQyR+ra2QjmMP5ySxEzWZwLixzULnVZWIaAq0rOGg9Fq/CT7oG9SD1VIR3s39jVSX8Wa6LzxmwYeRUgnUxHJ3J6+MISPjPW4VyHPKhNSoAxH8awwqrOzH5Vi0aNHeRrY0SMj7uYSwGtI1eU+9Mf/otCStqtdADaIXxlYSdivySijcv0vvaXC2sOXXvdthrXacHmzCglDPisljFDcb8BjDvoYWXreELOgdm22gy5IQLbtyJa5x2K7q2MwzlW2guaPIZyEnScvOH6RZnb+UTQMd50JCVsVUqmHVc2VOJSRERkoDKLzNf70wUks2hep7FO7qMgmxyblSoz+hrhZD21q6nricWqJpvf+2EJ4hohpprKWE5Dgk4yLBdot0wKHBAEVzg0StpsICB5wDhqm1jLTGiiqaJYmspoJ5Wo5Hv02bf7Vo1r0UkZ7q/SraFTwuVwxCxWWRb20hUCnSrbzPuVKDz93bXC5zGdqifG2HgY9GyxktUVZV8Xp47Wtsd3QhfTJZ+YiEokWLtlBbCiTkvaQzyN9Ye+IqJFKSmUtg69Gja1/uY8WNPIRS/nlICXgpXX3IFHi1jZ7OAE9v15RUhIoZyfGIzNygiLgjh3TzxRnHG7bttbCWTpqlJ+5ZI0Wen62G8mHlksttzTPXHZ2Qn7L1G43yu850JfgdU4YXaiJSsP6+Io/5MtHu4jAoRZsDM/E02ifVqUQKh2Otk3/aO8cEyf2ALtv7YTJVInPmQkCXl98b+LZjFxpHUg0UsjaRLc2FsfvihKqhY4NrQhpzIfyEc7FCOdQ2Sy0sZM0AggTiZkSwJsaWK9AR3pfCaZIUJZ+n6S1zC8KmJv8xH1wASpSve61tiSAKVLFRiQr12vg8plTMJLvB0pD+3bDNONaxUBrnZZaUAZoj7Zi4bfACmysRCUWLFu1tZEuBhJAElGBPXZN5EGeBytYXmFuMB1M0Qu5NTm9mYQF6k1nNo42qURCJg00GcyFHiZdJslMCYxLRosDV5pWAiFbIHakUoMWiR631k6Rsg9wRglHph64ns6mZvyyLltW1b5Pajh6Az8v/A2VkzSQxlAQnz5ZWj1Fv51tY/JfzMQWGiiqqFELiX0ry43ad4eQLT4cLW9kKpEMxC9tNlsZ0maCXtnObB7t2hdGZXGkCYjUuRVIeEpSYkuCxkLzQIqNMrblkv6bCzzzmxOuc4XVOtF7UoKpkdRWlmgplNkIyYTh/OLYQtgFUosWM41aioFDtkWYA9h+he1+579Wda3CF47RkRW7VqLPgb6PFsSiVZDKaon/Ash3eB2urXRNMsxXMY1pEQtGiRVuoLQUScj489ZUCr0fqBFWvk8wX43nJTJQyoOGrEnEKu86s7IHeRvxOLeLW32NynbYjjxaPrfV1k6UJx8hbtFlYaRE1HrNBuZAmoUU9UuHgLLdCzeh0kJSyHyqREMIoIxLF/KEMrdSF37wrEZshHiU8KqdIUEGe11W9ZT0fT5ehdjH9fo4RpW5dVuXoxixFWNkiTyDpEbWzbgXk8/y1MIjnXwuDbLFB4Ek2ALh4NsMqI2rKaVErI7VSUiO/piEi5vDwhzNRY0h5dXIbbUbiNk8CTUbHhAbTWglFrt/iPBrHnFCZyhwkEmY5VR5NJRty3lUiIWQmaQxJxVojT0EEy7GqIgzjipwzvszytnR+i7S5ylzs3mZeFn8DQrDlT5VzzSTSIinKfCzjhNz8y4eK5j/KIhKKFi3aQm0pkBAwl9WLcl2tJ2RaW/sX8EciCeILLCIyF/kAYF3pSm9O78koSLNB/kbqUBmQ0JN2NsMxT76bkqqUrRBHNGBuhfKKFIFoEM2oDbFyL1zDWfZ0Rl7CMVVfpxchJdTysEiJaVdJGlTi6R5liQdDNqLRCs7shAipqchWrTDxCDckfkTC/31vDfkyyeUyE3dCD6oM4owlICpofeaZAEu+/vuh1mVyWOVFdtgeZ/h9XXzoo935KbEIpiKFJgnLcYrHOLytYtRqkow4jDHbK03XEzS6QhW6dCFmcW7VSJzuRyauUYEjaxMVNmNfjivrCXESZXCuFFGTyJqaHdYLCLQiqJ+rgDPeq92dWy0AVvKhcwyYF6RomJo7Ck3WI3ATfrHZLLC6oR8j/+amNeRjEdLICUWLFu1tZEuDhELEK/y/sEJJmmXwlvk6QkfyxFZoKMRg0YrqVhIY8u4S7jr/oQBf+mfEDeV2zBOXgifuketRjsje68EV7d4Ib4xY9KeIUMY1foeiZttnwzlWtkvvbfk1ijQYIuRrV916ur+s1vKoni5VWZiL2zJPSv5CNWG18KPqpixPSMfhOcaUKR0PvCEXNXZULona1wxZdyeh+122rblHgfzJbkgVH3JSPU+a5WHO955L0H865FmtrpfSG2H8RMyuGr08YBuiwT0hQUERXqaihPOoUgBCzl15ZWoLLYF+6HNUjukNlaByoGJaIjfNv9Bhfljl7BJJqqrhoihSoXzJjRDN3GfLpsOBhyesPX8l5FetrFWRjSLNisypRlFIulYqZ9sm7+/W6WbZnspWIbWtfpNPCG0iEooWLdpCbWmQUOFKBr/M9qx6umSOHLLqeH5fH9VSWWw/C2vUHrsZBafWL4QP1s4QHRSurNpX3glrwvZuVJvSSfh+elDNXylua7hhvwGR0ualDCcpT6FWvhYRyY38eeC23ia5SKpeveKdavOpCIqOkVmhWvUkJp1bmytxdQd7RH4HhfFcY+buzKgOkBBCjA+Ch15jHtBoHF6P+2Gb8SSSPp0SaySN8Hmz59FSE8lG9dpVsyQvrvtwyMiPtyr08LmEylT31e4U9rkiVSYLLIRKXi2TFKyQjyKLQtxGxQhN8nPn7bcJi+aFbasXtuPauO6rkr2UKwBQRrYGbNm0fyvsN+jP0CCSvtkM8/bu90lVjeNSex7e6y55ygNiHqvkF2okkl0hl3fmUoY2JUlcLWJr3CGqbz+uRSQULVq0hdrSICE4Z7kWkuk8svQUMsnncxL0FK6RPa7qBY9UUMvrK8IiXiQrk2PkHQ73wr57lCS9/xrX4syRUXueIVFAk6JOkoeVJ1QO0u2rI6yvBc6j3VIukVwvK/Qt/bTmJ+oi7zURd6FG58v5LEV+GLGyZoJJ5WNvCU+uOiccwlh1dYwqFWOPgh5YQm6ux8ghkYJaUQvxrChTnJ62w+56DeNLwhhOnAku/Px7m6Y4IK4tbVbH1Vf2721Jxorr4uUrl0fohdcpdJPONT80XsMCQeEDRTrBez0RupxVI28CCfZ24iybWlyUxt9eq3I9L1MC99YrYaAtouKU96mpe8s8o9t3phx/qTXUZdQrnzLznI0s9VNIiSqtjfj1QKSpXY8kaLtEaVunwrmPnc+QKvJWq/53teVHHoXuo0WL9naypUBCHsz9qXETJRioZmbCFUf5ioeo73knT0CuQu8rk1S8ga9611le4ICaN6qKP7hJJJFLVY9IR66OCGFCfkTyowmRkSRDs4azHAtr1WLtj8XjhJeKnKS1LOaywZ+iIFVvVIRJ5Xs8huqilINEqOBnRWU/i7xpMuTxZppD8VYFRocch/gaIR9ee1NIj1/Z2KDmzNMBAeV7HOOQHNlGeHnlw+GAp9/dQkov75o6LTk6zs3uLSoRkJPTXFnxvKRX+Yauv0O526ztykr2mkKnuDZFARNyYBYpFOU4qx67qfkAkKnGUPOaVu/dhL/Na/+P5BDrFxPO3a7y0XiyFvWI1CQgTYHVQvk8bA+kvKZaw0hpW6nJ5PBSGNxtklzKXztxPhznFKWMu8cTyyUy7apa80n7aR4N1X5beyQScs79onNuxzn39bn3/plz7lnn3J855/6Lc25j7rPPOueuOueec879zScaTbRo0d5x9jhI6JcA/CsA/3HuvS8D+Kz3fuac+6cAPgvg08659wP4BIAPADgD4H84597jvSnYPNScxRXK3IUZOZVcWZ2NEu3UkYu1xuExCqe6L0kScmPLWiKIslSZ+/P7cDikEPnOa8FL7Nykjg6TfApDEtQiopi76tPESWQkPDr0XBvbGRpcm+fzuSph4JWBiNdJ5Natxkf5HuH1dE58PhzPl9xPTQvYVBvFJ0lh0SKI1TW/tdIh8hiy/XV/32OWVz2t2iB3iBh6RBuKCAllXXhPiN7MmHN0n3zI8TNhx60rwUO31lMj1TSf4ur2rrON0FU2rGROUqeXVPbT76i7Kv4qvN+m4HzyoL8CA+WK0PK1suGJKJShX2ocha1pVifl8Z3pNwntEgkRYUqVUchadWlTtqYuxlU1AP3OnAdIraFL3WrVrqU1DfIGebXuZjjH2feF9zfJ/SibfONU+MLq2fB+o5PMVclX/9asblHbJwyPPRIJee//AMC92nu/572ESfFHAM7x/z8K4Fe992Pv/UsArgL43icbUrRo0d5J9p3ghH4CwK/x/2cRHkqya3zviDnnfgrATwHA9uZJq30KH9o+AEqnZC134cvuAgoq1Z6+yvo1CGbrVjf/tZJjqfFOvvDG7YyYgTu6R8SjMBEf/S4t67UAQJ2DdUw1g+tsBHfVOt4ywGN6zUI+NW0W6/5gyolCcPXroOfT9ablNUvmryxyrqI/i6jV+CXpESmqNOV8DPbDPOzfKwy5CQm1le1LzqvDFtLKCNd1rjH7/NIHwpzcpz7xOrOie5v8XiOxgYrbGZA6eelqGNgh87bEuUwmjJJNhBTCWJT9m1DvuqUmfbkrs/Kt24Z4M567Fm0Voijpj1q0lX9ZaTrH59Vq86wJCr+jlkxCnBPqjqvwa7WjaJnGFg7Q6gGb1LASIjtg5LZLlcOWWjjzPnQ4B21GK7cYLcvUwUPRwESqFGXzLLtktSxXxrd4KEUhH9P+XNEx59znOJRf1lsP2O2BLJX3/gve+4947z+yurL+5xlGtGjR3sb2hpGQc+6TAH4YwMd8SYdfA3B+brdzAG48xsGQOGeV7fLyln6gRoDS0smSMvtXkZtaRvRMjdnoPa2FMPR52KpBXlk7o0xsb95Dlck9kgnTvOr18hpiM0ikCyAqULO7554bwXMRf7r2XTWfUy1Wk1mqXY6hSc4layj6oXoeRflKdyv9F+O9asPKfRXFKMJTphiH12NGAbVVo8hiUnJS+gmoeSBYK9ZWRbjcnUQKGOHZYD6QkJHel+phkZaN/UbMj3mRraQPySPp3jZqyeaHw2pEdMyxrWyH1w1lAMPP6fFY/JQbcULVEJC0jFRbltrvJ3xPuWFp5myfsotMNbdLKgtr1F66Q27RS8sc4rTYNnoj7L/KCFen69Bg5FN/D+M98U+MGHL8DXJ1hniUUS1+iUipMOnIOVyh343y4varfFhLba7HeCJ7Qw8h59zHAXwawF/13g/mPvoSgF9xzv0cAjF9BcAfP84xfYIy8VB3y7paVsXNwh+/7igngpWp9ltxVeJW7JmILCMU9UCxP2J9H9ggWXfyvUzAazEp7lY4ylTSHRPdOP4A1ULHxMI4VD5YDl+c4gWt6Ai5tc/Brq+Mr6kfKOVRN8i+qYWOQuG2pCqj5+VDUQ8Kyeby3DNCbRXTqrjRHlLqQMsxenbpdDbnZbKi/ugU0u5yWVWSsrYI4RsaP8l9Lo2Sco0YNomzP4gJk0GnLBhu86mj0hU7tFoTScOXP4KV7TCm81cCKZ4q3J7MyZjYQ7XqHOryGUMWn+4zxWBAgl0tizaO8fqaKNfItY6detypwPn0u0jG8wExPGApyzi83yHpvMI0h95Gef12j/Q3onvMeyopj3Za/RuzgI5SVjQoxXMsL2bOgUsc7kDH5lYPIeMjHs8e+RByzn0RwPcD2HbOXQPwswjRsBaAL5NH+CPv/d/z3n/DOffrAL6J8Pf+qceJjEWLFu2da498CHnvf+wBb//Ct9n/8wA+/8Qj8fPkkYjSKhk4d5LyiW9tXOS+uU9SfeIL4sgjFBZv5NcJkcYkccdDhymFydskSZ8ifh8fDzsPKaWgY41ccD8Hw/D5eKCGchwCvVMjTzG4Hl7c6yj8zPNyyWG90ekVCybiWUj8cvhey0oPam1WfEk0lw0geQyVHqiFkkjYiSYnHEuwWmFboZ3OChMLWzOgIfI9fKaCz9SKHbmVEFatQaGzxMjK29VGmByWiZmRMO8QPRmBq6ADr6vHEPzmybDfZSZIdrcCsjC5l8TPLbeqSEwBDi1V93kfXvtmGMzgJpeoTFR1XEb2Keh/8V0ZXK8eGOAplNjJJfaGesxzuVVMlSwqhCGExy2PNBkCM6YnZHYfuBJQgirvi7a2ghDCq9Mb9jfIwTpv6EryuH0u+SSjqzKYkZ+DT49hsWwjWrRoC7WlKNuABzCbC2Ga2Dg/t0aHc09nX/Wo8rBlqL7mW60pYnhZkphhv0FQGcWLJDtvXJ1itM8yDK591ZanySLNjHzNynrwrOsnAtdwnpKxXaopjNno7w6LC/d2chzcDZ8dMuzfIZKxVkVk1FusAFWrogFLFAbMUV9lEaQSOcWpOJuouRQHzS8T2xSytmREJs3NiABnXNtbASavp0N5h9aKMwSR0ds7ciJdNgEQJ1QYdOVgjJzVPa0iEUtVmCMlWpzvJqVwJY+a5UKDLF0gGb59OdyXM0wT6G4qY69yqjBZ1kq6Okxv9yNs9zT/r0guhPwNmdGC6QGjAYX7u12cJ9dTtmUKG5OoIdpVs4C0LRTPkhfJ6bJB5gF/q8oFyQ+9tSTKyHMluseSDWnpmsOxp8brzc0BMNfmqcrdFXnJg+0xZeUuk0Ula1wobSF7MgYmIqFo0aIt1JYDCcEHl6Mwe1p9Clv2n6Jl3hlaUsmBn/dq5cbMmfBSNYo0ZFHpi8+Hp/rzfxio/4PbM0zp/vKaIFRG997gGvigFbzefi16sXY2bE+eD27p8ungEQd3pti9xgRIoqyCbYblFZSUqCaJSo6bjelp9xQOZVSmIxcm7gVH3Lk8cWqhLHpFVkLOREEoSiZ+QEl+RBAbjMqMjycYEQEoSudYGqE5SFvVIthSRrXKk1j4QhIZeSlaJ3S3wlbRl95HKd57QqoKYbMVE8eQrZYlB/PXUSYmOhuaiXlJ2pbIVImbxawqXKcxqU1VKgjL8bfJqx28PMXsXBhXRnkM+x0X1d+ilXzUIlgmqi8xfspuTAflln0FoPxL44TEzSnpUnySrlfnrCGhenRwOgLuEwXeu0ZRO4n5UUplyJSI/YlaeD6eRSQULVq0hdqSICGE5Lpa40KhG0t0k6SrS82rSbAqeSghX4NGWuMzhaTPhKs7L4ZFdX8nfDAdFfCFkuFqJBX5mumEhZKMaI0H4XVZ4MnICT3W9rvJTaxk6G2Rv6B3UwRtxHGp/bA4leaKkFHKc4UxjEfhnD3CGImCwbtSpMzZW+GlAIGKHGucmziiBhGToh6ORbdNIoz18ymSHYWuwmad+TEqpLQok41FHtZgWnWMQp1zbxvI5T3vboU5aK9Ls6N6j03MzZoAMBfJdqicGXAOnsirbAJQJq0CZaqPCo/FB6pVkaQ7xpzDPpHteCfHkLK+LTVYtKJrXXN1qyRS8Zb289P9alb/TsaDArNxuAEJC097ayzXEFen3ClUc8PqS4ailqNXCgl6jJinNbhnmZlhH0X5+NUstoGOFi3a28mWAgkVvsBwPDAhe0mxFoQ31gKI3jZJknI9b46/5lktJBQ2xgdo6c9jaW3caoV1bLsdkk5mE48Z3UHBR33ZallcCvOAJOM6I0opKPLOZ3zOCFfB6d5+KoVricsSB6KQifgvIiSrlQ3fzU2YnF5pEI7dOGTujsIhyVwulDk3evcagaao2EDtkjk3kvo07oLfkixH+xiseHfMqFBjNXw4Vo7LQEiI18ExWZGs+CYn+Q3m26i8Iz86XssbElqU5AWjl9ZUcFaNOkmUX+NXdnDqnNXxJBaSdZVzCBkkFMfP03Bj7uxRDIwHVVb8WJnwRY67RIsN5leljFSZpEctd00RqxkjXBq/NRsk35MzbJauFWhygjvKgSLqmqjIWi2ZLI/Oz13lHBqzq+fcKcds5tGmFk3RCNvJvq/sk6stkrPErceyiISiRYu2UHNPKsX4pgzCudsA+gDuLHosD7FtLOfYlnVcwPKOLY7rye2Nju2i9/74o3ZaiocQADjnvua9/8iix/EgW9axLeu4gOUdWxzXk9ubPba4HIsWLdpCLT6EokWLtlBbpofQFxY9gG9jyzq2ZR0XsLxji+N6cntTx7Y0nFC0aNHembZMSChatGjvQIsPoWjRoi3UluIh5Jz7ODu2XnXOfWaB4zjvnPufzrlnnHPfcM79DN8/5pz7snPuBW43FzS+1Dn3p86531mycW04536DXXmfcc59dBnG5pz7B7yPX3fOfdE5117UuB7SyfihY3mrOhkvQ4flhT+EnHMpgH8N4AcBvB/Aj7GT6yJsBuAfeu/fB+D7AHyKY/kMgK94768A+ApfL8J+BsAzc6+XZVz/EsB/996/F8B3I4xxoWNzzp0F8NMAPuK9/yBCGegnFjiuXwLw8dp7DxxLrZPxxwH8PP9O3qpxfRnAB733HwLwPIKm/Js3Lu/9Qv8B+CiA3517/VmEFtPLMLbfBvDXATwH4DTfOw3guQWM5RzCD/UHAPwO31uGca0BeAkMcsy9v9CxITTdfA3AMYQayd8B8DcWOS4AlwB8/VFzVP8bAPC7AD76Vo2r9tnfAvDLb+a4Fo6EUP5YZA/t2vpWmnPuEoDvAfBVACe99zcBgNsTCxjSvwDwj1DKg2FJxvUUgNsA/j2Xiv/OOddb9Ni899cB/HMArwK4CWDPe/97ix5XzR42lmX6m/gJAP+N/39TxrUMD6EHiY8sNG/AObcC4DcB/H3v/f4ix8Lx/DCAHe/9/170WB5gGYC/AODfeO+/B6EGcGG8noz8yo8CuIzQA6/nnPvxxY7qsW0p/ib+PB2Wn8SW4SH0xrq2vknmnGsgPIB+2Xv/W3z7lnPuND8/DWDnLR7WXwbwI865lwH8KoAfcM795yUYFxDu3zXv/Vf5+jcQHhTo57IAAAFgSURBVEqLHttfA/CS9/62934K4LcA/KUlGNe8PWwsC/+bmOuw/Hc9115v1riW4SH0JwCuOOcuO+eaCMTXlxYxEBdEin4BwDPe+5+b++hLAD7J/38SgSt6y8x7/1nv/Tnv/SWE+fl97/2PL3pcHNvrAF5zzj3Ntz6G0Pxy0WN7FcD3Oee6vK8fQyDMFz2ueXvYWL4E4BPOuZZz7jKeoJPxd8LmOiz/iD/aYfk7P663ipR7BDH2Qwgs/LcAfG6B4/grCPDyzwD8X/77IQBbCKTwC9weW+AYvx8lMb0U4wLwYQBf47z9VwCbyzA2AP8YwLMAvg7gPyF0DV7IuAB8EYGbmiIgip/8dmMB8Dn+PTwH4Aff4nFdReB+9Dfwb9/MccWyjWjRoi3UlmE5Fi1atHewxYdQtGjRFmrxIRQtWrSFWnwIRYsWbaEWH0LRokVbqMWHULRo0RZq8SEULVq0hdr/BwG9bfHjJP0DAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Perform a sanity check on some random training samples\n",
    "ix = random.randint(0, len(preds_train_t))\n",
    "imshow(X_train[ix])\n",
    "plt.show()\n",
    "imshow(np.squeeze(Y_train[ix]))\n",
    "plt.show()\n",
    "imshow(np.squeeze(preds_train_t[ix]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "_cell_guid": "4f66b75c-c694-41a1-8c91-34bb6595837b",
    "_uuid": "d4ccbb559375bc2777ffb692a20adc313159f2cc"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Perform a sanity check on some random validation samples\n",
    "ix = random.randint(0, len(preds_val_t))\n",
    "imshow(X_train[int(X_train.shape[0]*0.9):][ix])\n",
    "plt.show()\n",
    "imshow(np.squeeze(Y_train[int(Y_train.shape[0]*0.9):][ix]))\n",
    "plt.show()\n",
    "imshow(np.squeeze(preds_val_t[ix]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "a6690535-b2e4-49ac-98d9-7191bfabfb6f",
    "_uuid": "6a34c98de7c6ae473f676a34fe7e099b46764eca"
   },
   "source": [
    "# Encode and submit our results\n",
    "\n",
    "- [Link](https://www.kaggle.com/rakhlin/fast-run-length-encoding-python) has an excellent implementation of run-length encoding."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "_cell_guid": "59a0af60-a7d7-41ef-a6fe-9e3c72defa07",
    "_uuid": "4f99c1bf852e82b60bd4f982ca0df293f712cdf0",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Run-length encoding stolen from https://www.kaggle.com/rakhlin/fast-run-length-encoding-python\n",
    "def rle_encoding(x):\n",
    "    dots = np.where(x.T.flatten() == 1)[0]\n",
    "    run_lengths = []\n",
    "    prev = -2\n",
    "    for b in dots:\n",
    "        if (b>prev+1): run_lengths.extend((b + 1, 0))\n",
    "        run_lengths[-1] += 1\n",
    "        prev = b\n",
    "    return run_lengths\n",
    "\n",
    "def prob_to_rles(x, cutoff=0.5):\n",
    "    lab_img = label(x > cutoff)\n",
    "    for i in range(1, lab_img.max() + 1):\n",
    "        yield rle_encoding(lab_img == i)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "31133f8c-3f40-4dff-8e1d-898d56672332",
    "_uuid": "2e07f6afc4787b068ba714428145dcb3951d718f"
   },
   "source": [
    "- Iterate over the test IDs and generate run-length encodings for each seperate mask identified by skimage ..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "_cell_guid": "22fe24a1-7659-4cc9-9d23-211f38e5b99f",
    "_uuid": "089587843ed6a3955fdcb9b23a6ec3bf5d703688",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "new_test_ids = []\n",
    "rles = []\n",
    "for n, id_ in enumerate(test_ids):\n",
    "    rle = list(prob_to_rles(preds_test_upsampled[n]))\n",
    "    rles.extend(rle)\n",
    "    new_test_ids.extend([id_] * len(rle))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "20b6b627-0fd6-425d-888f-da7f39efb124",
    "_uuid": "849184a40a2c9c21506d8b8eb10ad9155fa229e8"
   },
   "source": [
    "- Create submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "_cell_guid": "1ba0ee3a-cca0-4349-83f6-09a1ac6fcb44",
    "_uuid": "ba589f56f5be1e6886bc88f5bf9e7d0a408e4048",
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Create submission DataFrame\n",
    "sub = pd.DataFrame()\n",
    "sub['ImageId'] = new_test_ids\n",
    "sub['EncodedPixels'] = pd.Series(rles).apply(lambda x: ' '.join(str(y) for y in x))\n",
    "sub.to_csv('sub-dsbowl2018-1.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "_cell_guid": "222475b9-3171-461a-90f0-a820a6bd2634",
    "_uuid": "fb5e6f8cca872f1bd7036f6d9ac2ed2cab615536",
    "collapsed": true
   },
   "source": [
    "LB score history:\n",
    "- Version 7: 0.277 LB"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "_cell_guid": "3f5e5a47-6133-4870-976a-a8e4fa7bf46c",
    "_uuid": "2a83eab66bf55194f300953bea5534b6a043130f",
    "collapsed": true
   },
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}