Datasets
Models
Downloads: 1
Diff of
/notebooks/04-Efficient B0 Brain+Subdural Windowed Image.ipynb
[000000]
..
[d735e2]
Switch to unified view
| a | b/notebooks/04-Efficient B0 Brain+Subdural Windowed Image.ipynb | ||
|---|---|---|---|
| 1 | { |
||
| 2 | "cells": [ |
||
| 3 | { |
||
| 4 | "cell_type": "markdown", |
||
| 5 | "metadata": {}, |
||
| 6 | "source": [ |
||
| 7 | "# RSNA Intracranial Hemorrhage Detection " |
||
| 8 | ] |
||
| 9 | }, |
||
| 10 | { |
||
| 11 | "cell_type": "markdown", |
||
| 12 | "metadata": {}, |
||
| 13 | "source": [ |
||
| 14 | "<b>Competition Overview</b><br/><br/>\n", |
||
| 15 | "Intracranial hemorrhage, bleeding that occurs inside the cranium, is a serious health problem requiring rapid and often intensive medical treatment. For example, intracranial hemorrhages account for approximately 10% of strokes in the U.S., where stroke is the fifth-leading cause of death. Identifying the location and type of any hemorrhage present is a critical step in treating the patient.\n", |
||
| 16 | "\n", |
||
| 17 | "Diagnosis requires an urgent procedure. When a patient shows acute neurological symptoms such as severe headache or loss of consciousness, highly trained specialists review medical images of the patient’s cranium to look for the presence, location and type of hemorrhage. The process is complicated and often time consuming." |
||
| 18 | ] |
||
| 19 | }, |
||
| 20 | { |
||
| 21 | "cell_type": "markdown", |
||
| 22 | "metadata": {}, |
||
| 23 | "source": [ |
||
| 24 | "<b>What am i predicting?</b><br/><br/>\n", |
||
| 25 | "In this competition our goal is to predict intracranial hemorrhage and its subtypes. Given an image the we need to predict probablity of each subtype. This indicates its a multilabel classification problem." |
||
| 26 | ] |
||
| 27 | }, |
||
| 28 | { |
||
| 29 | "cell_type": "markdown", |
||
| 30 | "metadata": {}, |
||
| 31 | "source": [ |
||
| 32 | "<b>Competition Evaluation Metric</b><br/><br/>\n", |
||
| 33 | "Evaluation metric is weighted multi-label logarithmic loss. So for given image we need to predict probality for each subtype. There is also an any label, which indicates that a hemorrhage of ANY kind exists in the image. The any label is weighted more highly than specific hemorrhage sub-types.\n", |
||
| 34 | "\n", |
||
| 35 | "<b>Note:</b>The weights for each subtype for calculating weighted multi-label logarithmic loss is **not** given as part of the competition. We will be using binary cross entropy loss as weights are not available" |
||
| 36 | ] |
||
| 37 | }, |
||
| 38 | { |
||
| 39 | "cell_type": "markdown", |
||
| 40 | "metadata": {}, |
||
| 41 | "source": [ |
||
| 42 | "<b>Dataset Description</b>\n", |
||
| 43 | "\n", |
||
| 44 | "The dataset is divided into two parts\n", |
||
| 45 | "\n", |
||
| 46 | "1. Train\n", |
||
| 47 | "2. Test\n", |
||
| 48 | "\n", |
||
| 49 | "**1. Train**\n", |
||
| 50 | "Number of rows: 40,45,548 records.\n", |
||
| 51 | "Number of columns: 2\n", |
||
| 52 | "\n", |
||
| 53 | "Columns:\n", |
||
| 54 | "\n", |
||
| 55 | "**Id**: An image Id. Each Id corresponds to a unique image, and will contain an underscore.\n", |
||
| 56 | "\n", |
||
| 57 | "Example: ID_28fbab7eb_epidural. So the Id consists of two parts one is image file id ID_28fbab7eb and the other is sub type name\n", |
||
| 58 | "\n", |
||
| 59 | "**Label**: The target label whether that sub-type of hemorrhage (or any hemorrhage in the case of any) exists in the indicated image. 1 --> Exists and 0 --> Doesn't exist.\n", |
||
| 60 | "\n", |
||
| 61 | "**2. Test**\n", |
||
| 62 | "Number of rows: 4,71,270 records.\n", |
||
| 63 | "\n", |
||
| 64 | "Columns:\n", |
||
| 65 | "\n", |
||
| 66 | "**Id**: An image Id. Each Id corresponds to a unique image, and will contain an underscore.\n", |
||
| 67 | "\n", |
||
| 68 | "Example: ID_28fbab7eb_epidural. So the Id consists of two parts one is image file id ID_28fbab7eb and the other is sub type name" |
||
| 69 | ] |
||
| 70 | }, |
||
| 71 | { |
||
| 72 | "cell_type": "code", |
||
| 73 | "execution_count": 1, |
||
| 74 | "metadata": {}, |
||
| 75 | "outputs": [ |
||
| 76 | { |
||
| 77 | "name": "stderr", |
||
| 78 | "output_type": "stream", |
||
| 79 | "text": [ |
||
| 80 | "Using TensorFlow backend.\n" |
||
| 81 | ] |
||
| 82 | } |
||
| 83 | ], |
||
| 84 | "source": [ |
||
| 85 | "import numpy as np\n", |
||
| 86 | "import pandas as pd\n", |
||
| 87 | "import pydicom\n", |
||
| 88 | "import os\n", |
||
| 89 | "import glob\n", |
||
| 90 | "import random\n", |
||
| 91 | "import cv2\n", |
||
| 92 | "import tensorflow as tf\n", |
||
| 93 | "from math import ceil, floor\n", |
||
| 94 | "from tqdm import tqdm\n", |
||
| 95 | "from imgaug import augmenters as iaa\n", |
||
| 96 | "import matplotlib.pyplot as plt\n", |
||
| 97 | "from math import ceil, floor\n", |
||
| 98 | "import keras\n", |
||
| 99 | "import keras.backend as K\n", |
||
| 100 | "from keras.callbacks import Callback, ModelCheckpoint\n", |
||
| 101 | "from keras.layers import Dense, Flatten, Dropout\n", |
||
| 102 | "from keras.models import Model, load_model\n", |
||
| 103 | "from keras.utils import Sequence\n", |
||
| 104 | "from keras.losses import binary_crossentropy\n", |
||
| 105 | "from keras.optimizers import Adam" |
||
| 106 | ] |
||
| 107 | }, |
||
| 108 | { |
||
| 109 | "cell_type": "code", |
||
| 110 | "execution_count": 2, |
||
| 111 | "metadata": {}, |
||
| 112 | "outputs": [], |
||
| 113 | "source": [ |
||
| 114 | "# Random Seed\n", |
||
| 115 | "SEED = 42\n", |
||
| 116 | "np.random.seed(SEED)\n", |
||
| 117 | "\n", |
||
| 118 | "# some constants\n", |
||
| 119 | "TEST_SIZE = 0.06\n", |
||
| 120 | "HEIGHT = 256\n", |
||
| 121 | "WIDTH = 256\n", |
||
| 122 | "TRAIN_BATCH_SIZE = 32\n", |
||
| 123 | "VALID_BATCH_SIZE = 64\n", |
||
| 124 | "\n", |
||
| 125 | "# Train and Test folders\n", |
||
| 126 | "input_folder = '../input/rsna-intracranial-hemorrhage-detection/'\n", |
||
| 127 | "path_train_img = input_folder + 'stage_1_train_images/'\n", |
||
| 128 | "path_test_img = input_folder + 'stage_1_test_images/'" |
||
| 129 | ] |
||
| 130 | }, |
||
| 131 | { |
||
| 132 | "cell_type": "code", |
||
| 133 | "execution_count": 3, |
||
| 134 | "metadata": {}, |
||
| 135 | "outputs": [ |
||
| 136 | { |
||
| 137 | "data": { |
||
| 138 | "text/html": [ |
||
| 139 | "<div>\n", |
||
| 140 | "<style scoped>\n", |
||
| 141 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 142 | " vertical-align: middle;\n", |
||
| 143 | " }\n", |
||
| 144 | "\n", |
||
| 145 | " .dataframe tbody tr th {\n", |
||
| 146 | " vertical-align: top;\n", |
||
| 147 | " }\n", |
||
| 148 | "\n", |
||
| 149 | " .dataframe thead th {\n", |
||
| 150 | " text-align: right;\n", |
||
| 151 | " }\n", |
||
| 152 | "</style>\n", |
||
| 153 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 154 | " <thead>\n", |
||
| 155 | " <tr style=\"text-align: right;\">\n", |
||
| 156 | " <th></th>\n", |
||
| 157 | " <th>ID</th>\n", |
||
| 158 | " <th>Label</th>\n", |
||
| 159 | " </tr>\n", |
||
| 160 | " </thead>\n", |
||
| 161 | " <tbody>\n", |
||
| 162 | " <tr>\n", |
||
| 163 | " <th>0</th>\n", |
||
| 164 | " <td>ID_63eb1e259_epidural</td>\n", |
||
| 165 | " <td>0</td>\n", |
||
| 166 | " </tr>\n", |
||
| 167 | " <tr>\n", |
||
| 168 | " <th>1</th>\n", |
||
| 169 | " <td>ID_63eb1e259_intraparenchymal</td>\n", |
||
| 170 | " <td>0</td>\n", |
||
| 171 | " </tr>\n", |
||
| 172 | " <tr>\n", |
||
| 173 | " <th>2</th>\n", |
||
| 174 | " <td>ID_63eb1e259_intraventricular</td>\n", |
||
| 175 | " <td>0</td>\n", |
||
| 176 | " </tr>\n", |
||
| 177 | " <tr>\n", |
||
| 178 | " <th>3</th>\n", |
||
| 179 | " <td>ID_63eb1e259_subarachnoid</td>\n", |
||
| 180 | " <td>0</td>\n", |
||
| 181 | " </tr>\n", |
||
| 182 | " <tr>\n", |
||
| 183 | " <th>4</th>\n", |
||
| 184 | " <td>ID_63eb1e259_subdural</td>\n", |
||
| 185 | " <td>0</td>\n", |
||
| 186 | " </tr>\n", |
||
| 187 | " </tbody>\n", |
||
| 188 | "</table>\n", |
||
| 189 | "</div>" |
||
| 190 | ], |
||
| 191 | "text/plain": [ |
||
| 192 | " ID Label\n", |
||
| 193 | "0 ID_63eb1e259_epidural 0\n", |
||
| 194 | "1 ID_63eb1e259_intraparenchymal 0\n", |
||
| 195 | "2 ID_63eb1e259_intraventricular 0\n", |
||
| 196 | "3 ID_63eb1e259_subarachnoid 0\n", |
||
| 197 | "4 ID_63eb1e259_subdural 0" |
||
| 198 | ] |
||
| 199 | }, |
||
| 200 | "execution_count": 3, |
||
| 201 | "metadata": {}, |
||
| 202 | "output_type": "execute_result" |
||
| 203 | } |
||
| 204 | ], |
||
| 205 | "source": [ |
||
| 206 | "train_df = pd.read_csv(input_folder + 'stage_1_train.csv')\n", |
||
| 207 | "train_df.head()" |
||
| 208 | ] |
||
| 209 | }, |
||
| 210 | { |
||
| 211 | "cell_type": "code", |
||
| 212 | "execution_count": 4, |
||
| 213 | "metadata": {}, |
||
| 214 | "outputs": [ |
||
| 215 | { |
||
| 216 | "data": { |
||
| 217 | "text/html": [ |
||
| 218 | "<div>\n", |
||
| 219 | "<style scoped>\n", |
||
| 220 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 221 | " vertical-align: middle;\n", |
||
| 222 | " }\n", |
||
| 223 | "\n", |
||
| 224 | " .dataframe tbody tr th {\n", |
||
| 225 | " vertical-align: top;\n", |
||
| 226 | " }\n", |
||
| 227 | "\n", |
||
| 228 | " .dataframe thead th {\n", |
||
| 229 | " text-align: right;\n", |
||
| 230 | " }\n", |
||
| 231 | "</style>\n", |
||
| 232 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 233 | " <thead>\n", |
||
| 234 | " <tr style=\"text-align: right;\">\n", |
||
| 235 | " <th></th>\n", |
||
| 236 | " <th>ID</th>\n", |
||
| 237 | " <th>Label</th>\n", |
||
| 238 | " <th>sub_type</th>\n", |
||
| 239 | " <th>file_name</th>\n", |
||
| 240 | " </tr>\n", |
||
| 241 | " </thead>\n", |
||
| 242 | " <tbody>\n", |
||
| 243 | " <tr>\n", |
||
| 244 | " <th>0</th>\n", |
||
| 245 | " <td>ID_63eb1e259_epidural</td>\n", |
||
| 246 | " <td>0</td>\n", |
||
| 247 | " <td>epidural</td>\n", |
||
| 248 | " <td>ID_63eb1e259.dcm</td>\n", |
||
| 249 | " </tr>\n", |
||
| 250 | " <tr>\n", |
||
| 251 | " <th>1</th>\n", |
||
| 252 | " <td>ID_63eb1e259_intraparenchymal</td>\n", |
||
| 253 | " <td>0</td>\n", |
||
| 254 | " <td>intraparenchymal</td>\n", |
||
| 255 | " <td>ID_63eb1e259.dcm</td>\n", |
||
| 256 | " </tr>\n", |
||
| 257 | " <tr>\n", |
||
| 258 | " <th>2</th>\n", |
||
| 259 | " <td>ID_63eb1e259_intraventricular</td>\n", |
||
| 260 | " <td>0</td>\n", |
||
| 261 | " <td>intraventricular</td>\n", |
||
| 262 | " <td>ID_63eb1e259.dcm</td>\n", |
||
| 263 | " </tr>\n", |
||
| 264 | " <tr>\n", |
||
| 265 | " <th>3</th>\n", |
||
| 266 | " <td>ID_63eb1e259_subarachnoid</td>\n", |
||
| 267 | " <td>0</td>\n", |
||
| 268 | " <td>subarachnoid</td>\n", |
||
| 269 | " <td>ID_63eb1e259.dcm</td>\n", |
||
| 270 | " </tr>\n", |
||
| 271 | " <tr>\n", |
||
| 272 | " <th>4</th>\n", |
||
| 273 | " <td>ID_63eb1e259_subdural</td>\n", |
||
| 274 | " <td>0</td>\n", |
||
| 275 | " <td>subdural</td>\n", |
||
| 276 | " <td>ID_63eb1e259.dcm</td>\n", |
||
| 277 | " </tr>\n", |
||
| 278 | " </tbody>\n", |
||
| 279 | "</table>\n", |
||
| 280 | "</div>" |
||
| 281 | ], |
||
| 282 | "text/plain": [ |
||
| 283 | " ID Label sub_type file_name\n", |
||
| 284 | "0 ID_63eb1e259_epidural 0 epidural ID_63eb1e259.dcm\n", |
||
| 285 | "1 ID_63eb1e259_intraparenchymal 0 intraparenchymal ID_63eb1e259.dcm\n", |
||
| 286 | "2 ID_63eb1e259_intraventricular 0 intraventricular ID_63eb1e259.dcm\n", |
||
| 287 | "3 ID_63eb1e259_subarachnoid 0 subarachnoid ID_63eb1e259.dcm\n", |
||
| 288 | "4 ID_63eb1e259_subdural 0 subdural ID_63eb1e259.dcm" |
||
| 289 | ] |
||
| 290 | }, |
||
| 291 | "execution_count": 4, |
||
| 292 | "metadata": {}, |
||
| 293 | "output_type": "execute_result" |
||
| 294 | } |
||
| 295 | ], |
||
| 296 | "source": [ |
||
| 297 | "# extract subtype\n", |
||
| 298 | "train_df['sub_type'] = train_df['ID'].apply(lambda x: x.split('_')[-1])\n", |
||
| 299 | "# extract filename\n", |
||
| 300 | "train_df['file_name'] = train_df['ID'].apply(lambda x: '_'.join(x.split('_')[:2]) + '.dcm')\n", |
||
| 301 | "train_df.head()" |
||
| 302 | ] |
||
| 303 | }, |
||
| 304 | { |
||
| 305 | "cell_type": "code", |
||
| 306 | "execution_count": 5, |
||
| 307 | "metadata": {}, |
||
| 308 | "outputs": [ |
||
| 309 | { |
||
| 310 | "data": { |
||
| 311 | "text/plain": [ |
||
| 312 | "(4045572, 4)" |
||
| 313 | ] |
||
| 314 | }, |
||
| 315 | "execution_count": 5, |
||
| 316 | "metadata": {}, |
||
| 317 | "output_type": "execute_result" |
||
| 318 | } |
||
| 319 | ], |
||
| 320 | "source": [ |
||
| 321 | "train_df.shape" |
||
| 322 | ] |
||
| 323 | }, |
||
| 324 | { |
||
| 325 | "cell_type": "code", |
||
| 326 | "execution_count": 6, |
||
| 327 | "metadata": {}, |
||
| 328 | "outputs": [ |
||
| 329 | { |
||
| 330 | "data": { |
||
| 331 | "text/plain": [ |
||
| 332 | "(4045548, 4)" |
||
| 333 | ] |
||
| 334 | }, |
||
| 335 | "execution_count": 6, |
||
| 336 | "metadata": {}, |
||
| 337 | "output_type": "execute_result" |
||
| 338 | } |
||
| 339 | ], |
||
| 340 | "source": [ |
||
| 341 | "# remove duplicates\n", |
||
| 342 | "train_df.drop_duplicates(['Label', 'sub_type', 'file_name'], inplace=True)\n", |
||
| 343 | "train_df.shape" |
||
| 344 | ] |
||
| 345 | }, |
||
| 346 | { |
||
| 347 | "cell_type": "code", |
||
| 348 | "execution_count": 7, |
||
| 349 | "metadata": {}, |
||
| 350 | "outputs": [ |
||
| 351 | { |
||
| 352 | "name": "stdout", |
||
| 353 | "output_type": "stream", |
||
| 354 | "text": [ |
||
| 355 | "Number of train images availabe: 674258\n" |
||
| 356 | ] |
||
| 357 | } |
||
| 358 | ], |
||
| 359 | "source": [ |
||
| 360 | "print(\"Number of train images availabe:\", len(os.listdir(path_train_img)))" |
||
| 361 | ] |
||
| 362 | }, |
||
| 363 | { |
||
| 364 | "cell_type": "code", |
||
| 365 | "execution_count": 8, |
||
| 366 | "metadata": {}, |
||
| 367 | "outputs": [ |
||
| 368 | { |
||
| 369 | "data": { |
||
| 370 | "text/html": [ |
||
| 371 | "<div>\n", |
||
| 372 | "<style scoped>\n", |
||
| 373 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 374 | " vertical-align: middle;\n", |
||
| 375 | " }\n", |
||
| 376 | "\n", |
||
| 377 | " .dataframe tbody tr th {\n", |
||
| 378 | " vertical-align: top;\n", |
||
| 379 | " }\n", |
||
| 380 | "\n", |
||
| 381 | " .dataframe thead th {\n", |
||
| 382 | " text-align: right;\n", |
||
| 383 | " }\n", |
||
| 384 | "</style>\n", |
||
| 385 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 386 | " <thead>\n", |
||
| 387 | " <tr style=\"text-align: right;\">\n", |
||
| 388 | " <th>sub_type</th>\n", |
||
| 389 | " <th>any</th>\n", |
||
| 390 | " <th>epidural</th>\n", |
||
| 391 | " <th>intraparenchymal</th>\n", |
||
| 392 | " <th>intraventricular</th>\n", |
||
| 393 | " <th>subarachnoid</th>\n", |
||
| 394 | " <th>subdural</th>\n", |
||
| 395 | " </tr>\n", |
||
| 396 | " <tr>\n", |
||
| 397 | " <th>file_name</th>\n", |
||
| 398 | " <th></th>\n", |
||
| 399 | " <th></th>\n", |
||
| 400 | " <th></th>\n", |
||
| 401 | " <th></th>\n", |
||
| 402 | " <th></th>\n", |
||
| 403 | " <th></th>\n", |
||
| 404 | " </tr>\n", |
||
| 405 | " </thead>\n", |
||
| 406 | " <tbody>\n", |
||
| 407 | " <tr>\n", |
||
| 408 | " <th>ID_000039fa0.dcm</th>\n", |
||
| 409 | " <td>0</td>\n", |
||
| 410 | " <td>0</td>\n", |
||
| 411 | " <td>0</td>\n", |
||
| 412 | " <td>0</td>\n", |
||
| 413 | " <td>0</td>\n", |
||
| 414 | " <td>0</td>\n", |
||
| 415 | " </tr>\n", |
||
| 416 | " <tr>\n", |
||
| 417 | " <th>ID_00005679d.dcm</th>\n", |
||
| 418 | " <td>0</td>\n", |
||
| 419 | " <td>0</td>\n", |
||
| 420 | " <td>0</td>\n", |
||
| 421 | " <td>0</td>\n", |
||
| 422 | " <td>0</td>\n", |
||
| 423 | " <td>0</td>\n", |
||
| 424 | " </tr>\n", |
||
| 425 | " <tr>\n", |
||
| 426 | " <th>ID_00008ce3c.dcm</th>\n", |
||
| 427 | " <td>0</td>\n", |
||
| 428 | " <td>0</td>\n", |
||
| 429 | " <td>0</td>\n", |
||
| 430 | " <td>0</td>\n", |
||
| 431 | " <td>0</td>\n", |
||
| 432 | " <td>0</td>\n", |
||
| 433 | " </tr>\n", |
||
| 434 | " <tr>\n", |
||
| 435 | " <th>ID_0000950d7.dcm</th>\n", |
||
| 436 | " <td>0</td>\n", |
||
| 437 | " <td>0</td>\n", |
||
| 438 | " <td>0</td>\n", |
||
| 439 | " <td>0</td>\n", |
||
| 440 | " <td>0</td>\n", |
||
| 441 | " <td>0</td>\n", |
||
| 442 | " </tr>\n", |
||
| 443 | " <tr>\n", |
||
| 444 | " <th>ID_0000aee4b.dcm</th>\n", |
||
| 445 | " <td>0</td>\n", |
||
| 446 | " <td>0</td>\n", |
||
| 447 | " <td>0</td>\n", |
||
| 448 | " <td>0</td>\n", |
||
| 449 | " <td>0</td>\n", |
||
| 450 | " <td>0</td>\n", |
||
| 451 | " </tr>\n", |
||
| 452 | " </tbody>\n", |
||
| 453 | "</table>\n", |
||
| 454 | "</div>" |
||
| 455 | ], |
||
| 456 | "text/plain": [ |
||
| 457 | "sub_type any epidural intraparenchymal intraventricular \\\n", |
||
| 458 | "file_name \n", |
||
| 459 | "ID_000039fa0.dcm 0 0 0 0 \n", |
||
| 460 | "ID_00005679d.dcm 0 0 0 0 \n", |
||
| 461 | "ID_00008ce3c.dcm 0 0 0 0 \n", |
||
| 462 | "ID_0000950d7.dcm 0 0 0 0 \n", |
||
| 463 | "ID_0000aee4b.dcm 0 0 0 0 \n", |
||
| 464 | "\n", |
||
| 465 | "sub_type subarachnoid subdural \n", |
||
| 466 | "file_name \n", |
||
| 467 | "ID_000039fa0.dcm 0 0 \n", |
||
| 468 | "ID_00005679d.dcm 0 0 \n", |
||
| 469 | "ID_00008ce3c.dcm 0 0 \n", |
||
| 470 | "ID_0000950d7.dcm 0 0 \n", |
||
| 471 | "ID_0000aee4b.dcm 0 0 " |
||
| 472 | ] |
||
| 473 | }, |
||
| 474 | "execution_count": 8, |
||
| 475 | "metadata": {}, |
||
| 476 | "output_type": "execute_result" |
||
| 477 | } |
||
| 478 | ], |
||
| 479 | "source": [ |
||
| 480 | "train_final_df = pd.pivot_table(train_df.drop(columns='ID'), index=\"file_name\", \\\n", |
||
| 481 | " columns=\"sub_type\", values=\"Label\")\n", |
||
| 482 | "train_final_df.head()" |
||
| 483 | ] |
||
| 484 | }, |
||
| 485 | { |
||
| 486 | "cell_type": "code", |
||
| 487 | "execution_count": 9, |
||
| 488 | "metadata": {}, |
||
| 489 | "outputs": [ |
||
| 490 | { |
||
| 491 | "data": { |
||
| 492 | "text/plain": [ |
||
| 493 | "(674258, 6)" |
||
| 494 | ] |
||
| 495 | }, |
||
| 496 | "execution_count": 9, |
||
| 497 | "metadata": {}, |
||
| 498 | "output_type": "execute_result" |
||
| 499 | } |
||
| 500 | ], |
||
| 501 | "source": [ |
||
| 502 | "train_final_df.shape" |
||
| 503 | ] |
||
| 504 | }, |
||
| 505 | { |
||
| 506 | "cell_type": "code", |
||
| 507 | "execution_count": 10, |
||
| 508 | "metadata": {}, |
||
| 509 | "outputs": [], |
||
| 510 | "source": [ |
||
| 511 | "# Invalid image ID_6431af929.dcm\n", |
||
| 512 | "train_final_df.drop('ID_6431af929.dcm', inplace=True)" |
||
| 513 | ] |
||
| 514 | }, |
||
| 515 | { |
||
| 516 | "cell_type": "code", |
||
| 517 | "execution_count": 11, |
||
| 518 | "metadata": {}, |
||
| 519 | "outputs": [ |
||
| 520 | { |
||
| 521 | "name": "stdout", |
||
| 522 | "output_type": "stream", |
||
| 523 | "text": [ |
||
| 524 | "Collecting efficientnet\n", |
||
| 525 | " Downloading https://files.pythonhosted.org/packages/97/82/f3ae07316f0461417dc54affab6e86ab188a5a22f33176d35271628b96e0/efficientnet-1.0.0-py3-none-any.whl\n", |
||
| 526 | "Requirement already satisfied: keras-applications<=1.0.8,>=1.0.7 in /opt/conda/lib/python3.6/site-packages (from efficientnet) (1.0.8)\n", |
||
| 527 | "Requirement already satisfied: scikit-image in /opt/conda/lib/python3.6/site-packages (from efficientnet) (0.16.1)\n", |
||
| 528 | "Requirement already satisfied: h5py in /opt/conda/lib/python3.6/site-packages (from keras-applications<=1.0.8,>=1.0.7->efficientnet) (2.9.0)\n", |
||
| 529 | "Requirement already satisfied: numpy>=1.9.1 in /opt/conda/lib/python3.6/site-packages (from keras-applications<=1.0.8,>=1.0.7->efficientnet) (1.16.4)\n", |
||
| 530 | "Requirement already satisfied: pillow>=4.3.0 in /opt/conda/lib/python3.6/site-packages (from scikit-image->efficientnet) (5.4.1)\n", |
||
| 531 | "Requirement already satisfied: imageio>=2.3.0 in /opt/conda/lib/python3.6/site-packages (from scikit-image->efficientnet) (2.6.0)\n", |
||
| 532 | "Requirement already satisfied: PyWavelets>=0.4.0 in /opt/conda/lib/python3.6/site-packages (from scikit-image->efficientnet) (1.0.3)\n", |
||
| 533 | "Requirement already satisfied: scipy>=0.19.0 in /opt/conda/lib/python3.6/site-packages (from scikit-image->efficientnet) (1.2.1)\n", |
||
| 534 | "Requirement already satisfied: networkx>=2.0 in /opt/conda/lib/python3.6/site-packages (from scikit-image->efficientnet) (2.4)\n", |
||
| 535 | "Requirement already satisfied: matplotlib!=3.0.0,>=2.0.0 in /opt/conda/lib/python3.6/site-packages (from scikit-image->efficientnet) (3.0.3)\n", |
||
| 536 | "Requirement already satisfied: six in /opt/conda/lib/python3.6/site-packages (from h5py->keras-applications<=1.0.8,>=1.0.7->efficientnet) (1.12.0)\n", |
||
| 537 | "Requirement already satisfied: decorator>=4.3.0 in /opt/conda/lib/python3.6/site-packages (from networkx>=2.0->scikit-image->efficientnet) (4.4.0)\n", |
||
| 538 | "Requirement already satisfied: cycler>=0.10 in /opt/conda/lib/python3.6/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet) (0.10.0)\n", |
||
| 539 | "Requirement already satisfied: kiwisolver>=1.0.1 in /opt/conda/lib/python3.6/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet) (1.1.0)\n", |
||
| 540 | "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /opt/conda/lib/python3.6/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet) (2.4.2)\n", |
||
| 541 | "Requirement already satisfied: python-dateutil>=2.1 in /opt/conda/lib/python3.6/site-packages (from matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet) (2.8.0)\n", |
||
| 542 | "Requirement already satisfied: setuptools in /opt/conda/lib/python3.6/site-packages (from kiwisolver>=1.0.1->matplotlib!=3.0.0,>=2.0.0->scikit-image->efficientnet) (41.4.0)\n", |
||
| 543 | "Installing collected packages: efficientnet\n", |
||
| 544 | "Successfully installed efficientnet-1.0.0\n", |
||
| 545 | "Collecting iterative-stratification\n", |
||
| 546 | " Downloading https://files.pythonhosted.org/packages/9d/79/9ba64c8c07b07b8b45d80725b2ebd7b7884701c1da34f70d4749f7b45f9a/iterative_stratification-0.1.6-py3-none-any.whl\n", |
||
| 547 | "Requirement already satisfied: scipy in /opt/conda/lib/python3.6/site-packages (from iterative-stratification) (1.2.1)\n", |
||
| 548 | "Requirement already satisfied: numpy in /opt/conda/lib/python3.6/site-packages (from iterative-stratification) (1.16.4)\n", |
||
| 549 | "Requirement already satisfied: scikit-learn in /opt/conda/lib/python3.6/site-packages (from iterative-stratification) (0.21.3)\n", |
||
| 550 | "Requirement already satisfied: joblib>=0.11 in /opt/conda/lib/python3.6/site-packages (from scikit-learn->iterative-stratification) (0.13.2)\n", |
||
| 551 | "Installing collected packages: iterative-stratification\n", |
||
| 552 | "Successfully installed iterative-stratification-0.1.6\n" |
||
| 553 | ] |
||
| 554 | } |
||
| 555 | ], |
||
| 556 | "source": [ |
||
| 557 | "# Install Efficient Net as it is not part of Keras\n", |
||
| 558 | "!pip install efficientnet\n", |
||
| 559 | "!pip install iterative-stratification" |
||
| 560 | ] |
||
| 561 | }, |
||
| 562 | { |
||
| 563 | "cell_type": "code", |
||
| 564 | "execution_count": 13, |
||
| 565 | "metadata": {}, |
||
| 566 | "outputs": [], |
||
| 567 | "source": [ |
||
| 568 | "import efficientnet.keras as efn \n", |
||
| 569 | "from iterstrat.ml_stratifiers import MultilabelStratifiedShuffleSplit" |
||
| 570 | ] |
||
| 571 | }, |
||
| 572 | { |
||
| 573 | "cell_type": "code", |
||
| 574 | "execution_count": 14, |
||
| 575 | "metadata": {}, |
||
| 576 | "outputs": [], |
||
| 577 | "source": [ |
||
| 578 | "from IPython.display import HTML\n", |
||
| 579 | "\n", |
||
| 580 | "def create_download_link(title = \"Download CSV file\", filename = \"data.csv\"): \n", |
||
| 581 | " \"\"\"\n", |
||
| 582 | " Helper function to generate download link to files in kaggle kernel \n", |
||
| 583 | " \"\"\"\n", |
||
| 584 | " html = '<a href={filename}>{title}</a>'\n", |
||
| 585 | " html = html.format(title=title,filename=filename)\n", |
||
| 586 | " return HTML(html)" |
||
| 587 | ] |
||
| 588 | }, |
||
| 589 | { |
||
| 590 | "cell_type": "code", |
||
| 591 | "execution_count": 15, |
||
| 592 | "metadata": {}, |
||
| 593 | "outputs": [], |
||
| 594 | "source": [ |
||
| 595 | "def get_corrected_bsb_window(dcm, window_center, window_width):\n", |
||
| 596 | " #------ Correct Dicom Image ------------#\n", |
||
| 597 | " if (dcm.BitsStored == 12) and (dcm.PixelRepresentation == 0) and (int(dcm.RescaleIntercept) > -100):\n", |
||
| 598 | " x = dcm.pixel_array + 1000\n", |
||
| 599 | " px_mode = 4096\n", |
||
| 600 | " x[x>=px_mode] = x[x>=px_mode] - px_mode\n", |
||
| 601 | " dcm.PixelData = x.tobytes()\n", |
||
| 602 | " dcm.RescaleIntercept = -1000\n", |
||
| 603 | " \n", |
||
| 604 | " #------ Windowing ----------------------#\n", |
||
| 605 | " img = dcm.pixel_array * dcm.RescaleSlope + dcm.RescaleIntercept\n", |
||
| 606 | " img_min = window_center - window_width // 2\n", |
||
| 607 | " img_max = window_center + window_width // 2\n", |
||
| 608 | " img = np.clip(img, img_min, img_max)\n", |
||
| 609 | " return img\n", |
||
| 610 | "\n", |
||
| 611 | "def get_rgb_image(img):\n", |
||
| 612 | " brain_img = get_corrected_bsb_window(img, 40, 80)\n", |
||
| 613 | " subdural_img = get_corrected_bsb_window(img, 80, 200)\n", |
||
| 614 | " soft_img = get_corrected_bsb_window(img, 40, 380)\n", |
||
| 615 | " \n", |
||
| 616 | " brain_img = (brain_img - 0) / 80\n", |
||
| 617 | " subdural_img = (subdural_img - (-20)) / 200\n", |
||
| 618 | " soft_img = (soft_img - (-150)) / 380\n", |
||
| 619 | " bsb_img = np.array([brain_img, subdural_img, soft_img]).transpose(1,2,0)\n", |
||
| 620 | " \n", |
||
| 621 | " return bsb_img\n", |
||
| 622 | "\n", |
||
| 623 | "def _read(path, desired_size=(WIDTH, HEIGHT)):\n", |
||
| 624 | "\n", |
||
| 625 | " dcm = pydicom.dcmread(path)\n", |
||
| 626 | " \n", |
||
| 627 | " try:\n", |
||
| 628 | " img = get_rgb_image(dcm)\n", |
||
| 629 | " except:\n", |
||
| 630 | " img = np.zeros(desired_size)\n", |
||
| 631 | " \n", |
||
| 632 | " \n", |
||
| 633 | " img = cv2.resize(img, desired_size[:2], interpolation=cv2.INTER_LINEAR)\n", |
||
| 634 | " \n", |
||
| 635 | " return img" |
||
| 636 | ] |
||
| 637 | }, |
||
| 638 | { |
||
| 639 | "cell_type": "code", |
||
| 640 | "execution_count": 16, |
||
| 641 | "metadata": {}, |
||
| 642 | "outputs": [ |
||
| 643 | { |
||
| 644 | "data": { |
||
| 645 | "text/plain": [ |
||
| 646 | "(128, 128, 3)" |
||
| 647 | ] |
||
| 648 | }, |
||
| 649 | "execution_count": 16, |
||
| 650 | "metadata": {}, |
||
| 651 | "output_type": "execute_result" |
||
| 652 | } |
||
| 653 | ], |
||
| 654 | "source": [ |
||
| 655 | "_read(path_train_img + 'ID_ffff922b9.dcm', (128, 128)).shape" |
||
| 656 | ] |
||
| 657 | }, |
||
| 658 | { |
||
| 659 | "cell_type": "code", |
||
| 660 | "execution_count": 17, |
||
| 661 | "metadata": {}, |
||
| 662 | "outputs": [ |
||
| 663 | { |
||
| 664 | "data": { |
||
| 665 | "text/plain": [ |
||
| 666 | "<matplotlib.image.AxesImage at 0x7f6514ccc898>" |
||
| 667 | ] |
||
| 668 | }, |
||
| 669 | "execution_count": 17, |
||
| 670 | "metadata": {}, |
||
| 671 | "output_type": "execute_result" |
||
| 672 | }, |
||
| 673 | { |
||
| 674 | "data": { |
||
| 675 | "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQUAAAD8CAYAAAB+fLH0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzsnXeYFtX1xz/z1u0N2F1YelFEmqKoKIKiokbF3hW7qZoYk2hiQWMsSYyJmvgLUaNGY429K1LEggIKgkjvy7K9t7fM749z7ry7sNQFpNzv8/DMu/PembkzvHPvud9zzvc4rutiYWFhYeD7vjtgYWGxe8EOChYWFq1gBwULC4tWsIOChYVFK9hBwcLCohXsoGBhYdEKdlCwsLBohZ02KDiOc6LjOAsdx1niOM5NO+s6FhYWOxbOzghechzHDywCjgfWAF8CF7iu++0Ov5iFhcUORWAnnXc4sMR13WUAjuM8B4wD2hwUHMexYZV7EdLSOwGw/37dmfv1NwBEYs3fZ5csBKWu63baUqOdNSgUAKtb/L0GOKxlA8dxrgGu2UnXt/heID+nYYeeDcCUSf+gILsHAIWVq763Xll4WLk1jXbWoOC0sa+VNeC67kRgIlhLYW/BqIGDABg9sicAjuMwasRYAAo/tYPCnoKdNSisAbq1+LsrULiTrmWxm2DqvK9abQGmfvre99Udi+3EzvI+fAn0cxynl+M4IeB84PWddC0LC4sdiJ1iKbiuG3Uc56fAe4AfeNx13fk741oWFhY7FjvFJbnNnbCcgoXFrsAs13UP2VIjG9FoYWHRCnZQsLCwaAU7KFhYWLSCHRQsLCxawQ4KFhYWrWAHBQsLi1awg4KFhUUr2EHBwsKiFeygYGFh0Qp2ULCwsGiFnZUlabEvwtdPtvF1uqML/zxvPAApyfJT+8lLn/KTg7IAuOfjp7SdjXLfnWAtBQsLi1awloJFO5FDOPVY+ZiWCUCfDikAfPvtQ1z7/NMAfHT1lQBU19Xw01Eny+dwVwD+/uH9QOMu7LPF5mCzJC22Gj06HMSK/94LwG//+gkA97zzIQnlvf1l00H0dZ4eux8X//dmAAZ0vxSAUwb04JvyBgDe/M3RAPjPOm0X9N4CmyVpYWGxPbCWgsUW8flt/wTgsPOHQEGG7EzuINuP5/K7+z8G4O6379EjRKJzYPcLmLdujeyKzNXv6nDfexkA97sSAHzXjwfiO/MWLATWUrCwsNh2WEvBog34AVj+x8cA6BmJye7Du8Gxo7VNsEV7qenw6t9eAOD9l14C4JHpryGavXDeERcA8Pxn9wPikqz/4FkAUo4fu4X+hHXbtI33YbEBrKVgYWGx7bCWwj6NTKBWP4cAuPig0/jPbRfLLsMfHKpBSa4DjdXyObmXHtfSYtg+TJkyh2OOGbqZFjkAfPfoIwD0v+q8dlwtCYCnL5nAxf9VgfHYp+043x6FrbIUbJzCvgLfYYTT+gDQVK0vNiH8WfLi/+4oqeR0crdMmudIRGIoO1maVVXKNrM/JOfv8K6NHj0EMzlNuPKHANzx+D+973t0lriG/ft11D1JbG9cQ1rqfgCM6pLOvb+VaMsePX4HwI9+MZHKmmXa0rwaptLhvrN0scsHCwuLVrDLh70evXWbBtTr5yTd+snrJiU+F4wfBkB2biqkK7Hn0+p/HSVCkVF9IbVfu3ozZcoXjB49fIvtRo8eDUBebTqfr5WlzcoHLgHAueA+4PN29WPVnG/pNvgA+aNJllBO6kkQM+VJTOVD88z2iohLSzRaWFhsO6ylsNeih24NBxAlQQqm69YBv+QrPHTGYAAu378TQbUQllbL7JjbMRWAcEqQtP21kvmIIbLN7rLxpQuluPGUSbMZfckZANR/LsFLqUck+IOtheOYWXsUALePO5YuuXIv1/7rt9t0rtYwz0Y5ExqB/vp5kW73qqCqrbIUtntQcBynG/AU8mTjwETXdf/mOE4O8DzQE1gBnOu6bsUWzmUHhR2GProdrFvz6KtIpCibwcFBlhXQp5+Y0jcP68KcMjGZ//1dKQC162ukeSQGbrkeO1u3ZUz+v4kAjL726tZdaS6DuJrdi6S+cN+TH2bJmie36Y4Sg4LkVLjuqhbf+fXT9ry86kEJ6rOKvLYd59ijsNOXD1Hgl67rHgAcDvzEcZwBwE3AJNd1+wGT9G8LC4s9BNvtknRddx2wTj/XOI6zACgAxgGjtdmTwBTgN+3qpcVW4Zojf8m/F4llECkp0r2lLVokCEaBC36JATi5p0QZhn0+ShujAJzQTZYWFx4jZOXIbplkhOTYN5eOAeC6KcsZ+6vJADR1KZDT7qd5EQXdIE335Utm5OwHztqme5pw2bXe58nvPpL4oq5KP2yrhZCq22RguXyMLN/Gc+zd2CFxCo7j9AQOAmYAeTpg4LruOsdxcjdxzDXANTvi+hYWFjsO7SYaHcdJA6YCf3Bd92XHcSpd181q8X2F67rZWziH5RS2C+Kqq7ztIQDunLGaB2fIejtauVDbaN4CARKzqpkLgsiMCRNOEddkfkqQ9Q0RAK4ZkAdAcpK0P/fthYzrKf+VN05aAkBDeSn3nX4oAFXNcq27zhkEgBN3ueixLwFIUfLyvosPIufa89t533DTcaLTcN8kE+S0WdrKQrBziUYAx3GCwJvAe67r/kX3LQRGq5XQGZjiuu7+WziPHRS2GVmU3nQ3AHfMWgvAQx98AazX780AENFtmITvPaRbHyQLMXn6UDHzp66rIa6/id8MFc/ClEKJgHx/1lqIm0FG/fdOCpceLed4Y6Ww+PtlSJzDH4/uxaiHX9b2Jq5gIK5rwopNolP7sfTjGfQ9+vAddr69FDuXaHSEEn4MWGAGBMXrwHj9PB7Y6yldC4u9Ce1xSR4FfAx8Q2Ja+i3CK7wAdAdWAee4rufH2tS5rKWw1RDzfe7PbueGz0QG7cOZM/S7ahL/FWa8N0uFEAmSTV2SgUzOHC7xDJXNQi7GXLh+oCwb1tZJSvQNH68AILK+FjKVrAzI+Tukh7miv8Qu3D9D+hOvMKZ8CWDyFSbrNsZtp94IwH8Xy89i8YK/4pGfZWWy7dBhi0+iFZbNgk+kn86l5+hO+7PaADs3Icp13ekk7NENMWZ7z2thYfH9wmZJ7jGQIKMvfnwbABe+v4R5C7/Q7xp0GyIRmJS2wXdJeP/djqZEpyczt1y4gVUavTioYyr/WSxuzDeXyEwe8cvYH+ySTm6K8BH5KXKd2UW1zCiW3IGgK3xDE9/oNVOQcBbzGeBQ7nzjfgDy8tQ9WTyf6tniQs08SQRXRo2S6MUpU6Z4T8AEMSVnnsg7158p7e64Qr7s3Y/aaZLhmJon++rWP4bFtsPmPlhYWLSCtRT2ECz5hYiiXvzRUgDmLZxNInOv5dhuGP0S3Zp1dQBjPQRyJfchWtnI4BxxSa5QSyESc6lqkhm/i/IHPdLFOshLDpKiXMLbK8TT4FbU84V6JJoqTYZhjm4rgbX62QROzUBCWuCGQyXMOPbeYjIvfUC/l3yEqVOnAuA4R9G3U1GrczRUNTPmbyIxf8dc0X743V9PIe1QCYN+aqxoJnxZcisA977zeyy2HjYhag/A/531e5Yq6fend9/XvfUkKB3jJkxm4/RoTWAKZkJUTHmng0QqupE4qWnywteV1AEwblgBecmyNMjVUm/ZIdlOXVfDx0WSB1GxWpOIgn5orNNrGZLTxEisJiFOYvqYzle/lwFg6C0SrzB6/x8zdZGUkFv22iQAep12lPTV6QII+Xj7TRIY27AI/vjyn/R8AwDo2204i285AYA/viPX/6JY+hVrquXVWf/AwqZOW1hYbAfs8mE3RkG2mNl9spP54SuGVDRSak0kUqCbdRslkddg2qk3OJIJjuQwFGTJkuGQTqletaYVGo3YOyPMyho5n8mBKG+S7fS1VSQH5fypubIUqVtXSsJi0RoP3pIlHzDyZhK/dtUxZzJUcyquP06Cr6YuKiaYJNWiep1mApBMgFUWxlJ49rEPAFhY/Dl/dO7T7+cBsGR1HWm/FUulUQ2WWJ0sif78gwG8OguLrYS1FCwsLFrBWgq7MT69+ioABj8xC+JGQLSli1F1DrygpGY2zm8waScdIV0si0M6Sfs4sFS1EgJ+mR+CjsP+SjC+vkqCkOZr+DJxl3q1FGJVS/S8Poy78YIREp5yzSAhC4/55/Mc3Pc4AG44TEjF11dU4Fz2xAb3As3f/p9+SqEl7rnjV9x8u4i5LioxGg5BCOnSuHmm7iulrswESGnAVFgEUx5fWIL79P8AcC6+UtsYYRWLDWGJxt0Qx+wn6sXfRiXacP2yBYheDSSWDDE21g10AI1B8JYRmqSa1YkT95OXpVKXCourGqnSJUJ+qpjrl/TryBolNf/zhSgo0WjIwigJwtBIw2eAIwPQ4P5S+OX+I6XfLi6PqVDL2ytlgAn5fQzUqtQhXXVce2AeZz14qp6up55XlaRxcZzWBu2UVz+k9PWvALjh7a8BWFVUAXynLUx7GZwuOfoo/jpSztusA1bn809nH4QlGi0sLLYddvmwG+LcISMAuG2W8fFXkog/MDO1S8LtKNbBsIHDGaYE4MSPvmnV/uYjelAfk6WFyXpMDfgpi4pVUFgj1sCLy8tZUakWiAlG9DItC8lJ7wtAeY0hN98GV6ySXunXAfCtnuv65z/x7imvQLIwj+6ewc91th5xiRCpjDiKxE8xRms4iZoQEyRWw83M5qzHJH/irFrNCp3+NW+9LKTjTzSWY+VSKXz7n2kfk6niMA+ddSAWm4e1FCwsLFrBcgq7EYwLcm2FzHSk/EC29V/ROjIRZM0thOHwIXLcid0yKVQ+4NGpsr4eOVjW99PuPZGF00R27Af/k8jDovpm6tap61KJRn9mErF6sQwyO8j5D1VicnlhOXefK3UfzvtrWyrKJg/O8A0dSc0Q9+NFh/QEYH1DhFfGScCRc85AadZ7MAkXZDvwiBSsnfGNREAe+ZQIvMTqpoEjFoL77g0A3P3aDH73j9vbf809CztfZGVHwQ4KgtuOl4i9Oz98R3a4mfpNMQmvgkkpTuaKY8SnX63E4ceFNawvEu3CUXliLk9Z/lTiAjMkdHjUj0Xi4uNVVTg62MQ11uDE3tmUKPmYE5YBqFiVmL5+4mycQ47eZP9TMyRluS4qS52j+3dibZUsRUzM49AOKYztKvf18xO0sEz/TrBKvQFnqHR8Wt9NXqcl3NmfAVBWXM0t94j34eUFEvpc3ST31lQ9HxNl2TtvHABLZ/8ap2Bz9Sv3Slii0cLCYtthicbdBj7u/OBBAPJ6inDV+hVf6XchEi46kb8cPWwAGSZmIC4zouvAr08V0/y+53+3wfkjUC0E4P+pe66oPsJkJR2/1loPcReSdSnxwZcrAEjPFPLSOWRcmz0/+WhJVZ705WIA0pIkMWnaN6sgYtym4oZ8tyyVXM2tIKKkYmYSHN5dPoeVUC3XOIiczVsMzsFHABCY+ByzKuRaJTVKjDaaOI4IhpRdtl6iIn9yYXe66nJtTcVXWCRgLQULC4tWsJzCboLeXU5hWeGb8kfoNNk2myhGP6JuBz8/QVSXh3VM9dKYvywRYq9zSojrXrql7QtESqBSCLhPbpXZsmd6mNGvyDWWrJJUayc9Fbd8nh60eoOTrGfU0IsAmPr1M9I+3IuPHhFX4TFXGJVmY9WMhWyxbA5VHmFQxxTuHyH3knWsVrNKDYHyF1MnSo7H/FKxXH49aQm1X4mwjNNL8yLixYw7XoKPXn3/Fb1WGfz9PQDu+UCsjN++ZqIdC/FqPHh9G8CDZws/ct1Lt7bxwPZKWE7BwsJi22E5hd0EywrnAcqGNy/WvSZUOY9LR0mp+OG54h5sjLpefYaj8iX0+QcPJaopbYRgJyLPiRbDIQMk9PmAP09h+WqpExHMluCiSEMjMHODg2W2d123RV1HwTVDR7awEAw0p6FDNslqARzdRfqYnxIiLUk4hebPxRJ5fM46TuwlwiwnvLYAgAM1FPqSoV1w+kkOCPFi7wqvffDqBpesB+UqbjpE7uW2D6VIbLQumuiTZlxCBa+uMAKzRrrOuFL3bdhBYbfBCkDj/9GcAxOb4MvyIvI+WCPE4JiCDIZoHEH/o3pKu1ST9wC8okuRM07xdgV/KsuSy8bcAcDy1fMYPkSIuu8q5KXJzEiidKXReZRBx3XlRRqQ13ujXv9zxlMb7fNQU8tx/SROoi4qLtWZJXW88J0sVQy5uaCygc/WywvZHBXy8Sc6cF15Qj9K6yX24qXPP9roEq/9TWITxl13Aqs0ceuH78pgcNux4vL8w1draVpjwjPNtsZTmHr8fIm5uOK59lSw3ntglw8WFhatYC2F3QY5JIg9k1eguQ3hVBZqPsI5vcXMzg77WVsnM/nBLTL+VjwpKcJT50gAz/gzWlyiWmbG9+ZL5mJen4P5Yp6Qj/36SQblIZ1SeatCzldd/TYAt54ilsWC4m0sxNq8kK9KRQ4uqGXjZhTXsbpWXKNL1UV6z/BuTNHIyqGau3HegZrd2T2T0UpSvtTGJU6/YQIA7uVDmbpE7uudL4QofSes1Qv9AcjSQreVxk1Zx6z5IgBznkZWXkE3NiZX9z1YS8HCwqIV2m0pOI7jR5ipta7rnuI4Ti/gOWTqmw1c4rpu8+bOYQFwBGCES1oWgIXLDutBnwyxGubp2r+0McpND1/c+hRNpQz4uYQw1xc/woa47pw/A3DvGJFG++GkJTx4huQEmOKwt77+AYZ469PrUgD+9rHmYnA4+DXMOvbWVtxTEdWNYs10UMJxZH46a5Uj8Kk7PC3oo0Il3x5SLQYModkYpa+pSuXNYS3Kz8eFP2h89VvunCFycL4cycKMlxvLpiPUyzHJeXK/DeunQ1yskzs/E7J19YRb6TbBFkLfEcuH64EFJNQ97gMecF33Ocdx/g+4Etj4F2rRCr7UFOJ15r9D1Yd8StJF4t5Le9X+Yo5/Xd8MXQ3xp+TZupXUV0zUfUmtv3vuTX6s5eAWa+Rfw91jueoZESl5bLL4+HF6UNCnMwArSiVjISVFz1WdBzGjWGTyBmLAOv1sUqxF8SgzfyApqgQ9V7Ug562v8ZYSl2sK9b1fFVKh9+fXwaBMcybS0sPMLzeeA+P58LNhinXy5Q9w6ciRACxZIcuIQQdKHsU3C1ZCXO+52eST9EKqGsLDWqn75sO60SlDyNiS6tfZV9Gu5YPjOF2BHwCP6t8OcCyJ5d+TwD4pcWNhsaeivZbCX4Ffk9AI6wBUuq5r/D5rgIJ2XmMvh+ol1hWRiEsQUze1o8xqnVOC9EqXnIAHlRic+NHdiVNUiglNz4PYaJx/VIjHz2cX0jNHLJBTtcT8yPs/ZvpcmSUL+ki04Npl61m7RJYxpx4+GID3l6k/PyMzIRLtUx3EeAwwFotYCocOFlfgl/PWQYGQfTO1rkSsqBSQ8y3sJt/NnreMC0aKWf/QfBFNqVcX5ivdMpkw1SyrTOBrdxLydIr4VzTFpFYEMXF1/kTJyiuuGEbaHR8C0FxhYh2SvGdVVz4XgBcX7MfKW6UcXcqvrKWwzXAc5xSg2HXdluLZbRWcbTOE2XGcaxzHmek4zoaRMhYWFt8j2mMpHAmc5jjOyciwm4FYDlmO4wTUWuiKBJ5vBNd1JwITYV/PfRih29V4XIIjuQEn9RNSLxJ3vSKuT75xk3fkl7+RqkeH3nGJ7mnGiJW4j4uF8MDbYkX8+JAC5muAUOdrJV8gLz+dM0fIrP6qztC4ZZw0XMLjzfq+SV2IBP2kdxceoEbFXCgrA0czG/2yNfzA8EFd6GwK0So/sTo5jXBA2r39lfw0evTpRpYGZ43IE6PzHa1A9eTnqzmopzyHaeX76X2mkiiLZ5QaHF75ZoX2Q6wvE+kZ7NOB60b0BODP75o6FNkkVLAlOOu6977lqlOED+nSQQrcFpZNZV/DdlsKruve7LpuV9d1ewLnAx+5rnsRMBk4W5uNB15rdy8tLCx2GXZG8NJvgOccx7kL+Aqw9cDbhAklNsE0HTHirP1VKt0oH62oaWZZmc6I6cqeN5Vy6H0/AiCti7jRfE1Rpl0sGgGvLpPKUBOOFz2CCR8s4Y43Rcz1tCMkO3F4p1QemCMcRbxKi7gmFxBVV+Hn68Sy6KNr/6ywn1lzTRUoo5MQxp8jLsyT1bKZWybeglsO6sI0rT3ZoBzB6sXzCBWIS/SK/uJJubJ/J4Yd1k3vT57B6WpFBH0OH601RIZxRZZj8jESloJL06w/APDjn/8HgA/WiAqVf/Iy/nTuIAAe/EQsheaaQhI1McQiaqhczDcz5f7m3ynPNPsn+56lsEMGBdd1pwBT9PMyYPiOOO9eDb+QeMQMDZNEh25C2Bmfvdm+M78It/bR1seHO/KH8VI6LTlJk4565fDpOnkJbx8lxVeG/eNzAGYvLeKZiw4FoEbFTW74aCn1paV6QjWlw0G+VFLQ5FuUqzxbVXMUr2BsWExzJzWZgwvEGz25UK59mCZtnXpEd049/mAAZjw5BYD3v3ybe0dIIdj5GnOxoKKRYVp3AnU/pmkUIxlhxiyUPv7HM/dXAAfoZ1W8Dh4MfWSw+8cbEwB49pJ7AaiOxLy4h+fPEjflGU+Uk3DbGldqDePfF1Jz9jGa1u2paJvaF3s/bESjhYVFK9jch+8LYTGhqdeMyJQ8Tu8r5rexEF5aKgTYHSfuT2LGSuCWZ2YAMP4oOdcl/Tqyf5bMfhl3i4hpLCZWwYrrR3PPVxJk9M8vdQlQ30zCDarEXWUjlZVCeFaaykxpOqPWNECWuiIrxTR3fT7mank5xy/nyvFpKvI150G5nPewCVdL+wlXw9siTFs6bQUAa+qaeeV1SZnur/0/YIgEUNEcI+Az1pRKttEMYfF0jzzwFwBM06VDS1xwuRCmrz41mzVKpJ4+Wiyoq1dW8K/JovaMT5YxxCPMWyv39dl7QtCeeMB5ALy7YDPZoHsZrKVgYWHRCtZS+L4QN+kgMiv3KMgipDNify0Vf3SBrKuvufL4jQ6veH8G44YLiejT9fKrKyp4+EMJxCnoJTPpHcOFwDvqhTmsWa1BTh5JmEZiPa0uSVIT+wK6hlcXY7hzJk1FRphE8woavqYpUScWgBdVx8Rx7meB1mLo/0MRYln/xJtMniRZjIaDWFffTMgn89NvD1ILwbg84y7dlG/I7S1BV8XL1kKTWCDTJt+jVzUyay2wXMjWSNz1MjILioQ8vWtED/41TWtPxlRcxcknEpHrTlNu5uzBYoW9u2Dj0++tsIPC94VGfbn8woAXpIU8IZKMkLD/52qadH63zI0Of+7RDxmq6kSPapr02tVzIVUGihot6HLV66rz2FBJwjBsWa26Sj8bNn8toJ6AqAamahRJJBoG15Sy+3qrbvO5TyX9uvyZSQC8vjrA2D5yX01axi7Z7+MdrfvQN1OWSQdrjke0qpEyXU7F4yacJQ3QJK14y8HAfC+D5Ce6TDpjYB53T5dlWkBXIof3zIZkjQsxpefcGATlurVKxsbi+14IjV0+WFhYtIK1FL4X9MToAe7XWyLo8pMD9FQfvSkFf990MdG7df+QU/8s7b5+WHLNphfV8NyXMvvFG9WtmNIX6mTmr64zrkaxMoYP6c9s9dtHy0yMRAiczMRnkPJxKolmSsmldBaXY31DlLYIz83hjv/8Rz+JxmR+n+FeLPyzS2Sd4bjQpOnUp/cUy6lIoy9jTVHMZF1fa5ZcdRiCdN5HUvlp4JldYMVs+bqnXOvRhbLEOLJfR4+8/ULdrc8uLef8oRIP8tx0k/kZoUCtr9SA1tTYDdTOdzWspWBhYdEK1lL4PhA6GGIyEw3NEzLv24pGTzNh0mKZQUf0lfXtaQ99SurTsob/2cFCtv3347ngF/fdyIMlWu/j2UuBer2I4Qrkv/iLuavI7y16CkVVsqZ3wkEO7Cmfv6sUtjBaH4EKIehGDJRgKqOFQHUjwQ6yL7LVOQGaJJshWZBFJXWU5InL8pfqdqyPxvnb+0KCmsAqdIZe3xBhhZa2D2mQVj0uhgP52z0vA/CvM49h0h+mADDmX2IpvLNKnsEXy8o9fuafCyRLMjPkZ1CO8BGvZ8i56qtrqFIr7bQemt1ZaiIm9x3YQeH7gBOCmPzYFqjgiQNM0h/xJfqyHNRRTFkf8PlKISbvfUcGh/OPGsygDvKj/t1rRuU4jpdUxYG61ZcsEKTISJpranEovzPzdEmRqfEBVa6LkyMvkBmkjDnZrUcWeSqjviB+IQB1FdNJyKcbr4a80JkdTqCqWo4eoS/lp8sqyFThleXqEeiWFsKQn9PVO3CAemAK6yN00MEgU69dSSrmp/voTIk1+Bcw5urWxW97Z8hS55Xl5VyiGpQmJTsSdxmoqeSnDJaB64XpM0lPlyWF8VYcvV9H9jXY5YOFhUUrWEthl0LIQpqqwJFHP0SJrReXlTOokywluqcJ6VemOQczl5USrZakoAmnSC5BdjjA9a8Zt6ApYtIDp5MQda4Sa2lKXtZWNEC9uiKThDgc1imV1cnSj9W6RPD7fRyieQdpQZkzDtF+pQX9RJT1M5GTwztd7S09nl0qy46ze0kfgj7HK2tvZugzB+by6Xrpx2glMMubopif4jO6dDpSlxhzy+sJahzGmkpjkeSQKAMnJOF/fv0Il/zxR7TEp+/9EoAjxt7P+RoteoW6Ot9cVcmXmo5e0WSI1Q74lGQt1kI7KUE/+xqspWBhYdEK1lLYhXA6iaCKW1rMgQf0BBIyZccVZPCd1nYo0Ai+H7/0MQDp2Xk8fJ4kntZFZMb913clEJWgpREHigxZSTRGJ113r2+QGdoQd7VAfi8hKTPCMvv1SAsxS6/fNVvW8NcOyDUcH6WqxHyCWg5fl9V75391ufATqQGfpzCdqgVvf6jVnQ55fi6/HyHis0Ykpmd6mFK1HkxOQyTmEsgXEnSpRiGOXaip3LhkdpVZPlZioi4rSeRsCDF46Z/e5ZLbpfitVykrV/iBQTkpTNXoScMzFNY3s0ZdnH6TWxFzWFskFtlhWn+icP2+V0rOWgoWFhatYC2FXQKRMLvsQJkNR+T1YXKhzEirdLaaW17PLweL1+HHr0puQCed6U7rQZjGAAAgAElEQVQ5INeTOTcaBPPmz6ZqoejXZDgy8/7oR896gT491UIwlkhtZhIBv8yIi1bIbLyoqIaj1SvwC5Vbn1FcS5Gupw/X2dKIvQzPTeWYPjJr91Te409zivhU61senC/t52iNyON65lCtHox8tTBSAz6PRzH7ihoinKMCLR+uEAugtExm9r+c2J8svf7lzxplvxwS+gY6r4WD5Bb8GIDiyqdpib4ZYZ7TjNOJR0uW5AV9OnCLhkEvWmTEXGOMOECsqTnlcg+9O6ayr8EOCrsAF44cD8A1BwjJ9VFhNTF9eRdpfQM3GudhTe99/nypqTBXX679spL4laYZF6+RKMZHb7+GjP2EuLz9jDsBGQCy9QXqqG48Y+73y0zi3UUmylHNZcfx+vGoFn3tnRH2zjFIXXZHjBsgjQ7vhhl1TvhAKmMvrmpkXYMMbGdq5egFugw6u3c2TXqBSiU+U4N+L8ejo/atpDHqLXMCSm66mpiUHvQn8g+CmpIdaQJHXa+aO0JTFSVNpqJ0a+QmB/lME5xe0MFhQHZyQlE4bgjMEEkaxWmS05LClmi0sLDYx2EthR0Go0D3hbcnJUvShftp2TMTENMcc1msFkLxapm5Th5UwPM/kJl/tpJbJ2gQ0zFvLKC6Ts3lNC3RPsEoOMOdr2rMf0Ym43SJYtyaplzbmrpmTtpfAnHMcuDGwfl0U4KxVt1+DQ0R+o1V1eQjRbqMsIqQRFdCmUZK6pJhbLdMmnUmNxaGcVt2SQmpuxE6K3naOSXoLXE6tmg/o1iWOb/Re75xmRR6verVuTx+hvTDnyIkYaxqPbhilfhzhASNFce4YOQQ2sI/vy0GdTtOUsJx/6xkDtalwaqO8lyaSkvoqs/tWJWYW9QY3fB0ez2spWBhYdEK1lJoFzRjMHA6eT1k5lq/VLMNiXBoHyGtlmvsvlnnf15cy+xvRJfgZ8eLiMeDI3vymgb/jNOQ3HuUCDulR5Y3c93/UMuisrI2D+dJ+8MKMjyrpFhdklGdlvtnJbFM+3H3yJ4AFBxcAKpfkKX6C0TjYPQb1EKoeEeqKzF9CS9qZuNpms0YibteSPLrGop9dGcRdV1T1+yt201dh7DP59WCMNmgXVKC9FNXoQmj9rQRmip5RYnRkAZaNdTlQlQslYPyhYydWVrPC5rX8F9a4/LhXb0M1OfmCW9zTf9O3DhYyNUPV4v101Ra7XEgrtIuKVlJ7Guwg8J2oa9u5QeJ38fZSrJNT5aIw8Nz08jQF8G8qI8qkfjFohJvMDhf2fyXlpThmOi5cVLC7eZ88ULU1zWTpAMF/fZv0Q8ZFB4/TvpTH43ztZKTPdNlcDLEWXlTlEPUXC7Qa0bnryegkX5xVUz2HdAJJouAyUGnS9rznGXyXfyuExl3iAwAnUz+c0kdL+hg9sg3cn/Ha1zDqtp6PlR59mO7yKDWJyPJZGR7npf0oI9hGjX5r++MJ8CQfzFml8o9NdRo6nR2KqhXJV8HmF59O5C6iejDayb+ku/G3grA/p2FrHx/TRXD9HkcpNf+cKWPKUvkXtOOFTXnAfug98EuHywsLFrBWgrbjGwS9XTF5M7vkslLWnzFmM4ndM3EBMr94lMp4urEZWb/8urhNKpb7vezZRlxz8VHMPRnZ7W+1EBxBaawCcwTgvFIjQ/4v2+LPQETYy4vUvfg12X1/O4gWc68OU3yBo7qlkmT9nudloZ76cnZ/OlriQdo1ui+zE5yT9e/t4i/GfekEpqxdxcRNmIs6r57fmlCG9HMxvupteTieuXonl4ss3Kv9LDX38HqBp3mV8soVkuhpi87GjHpNsUgRVKbP9Wsyu5pIXqkb1oApr/GQUxV1+RHhdWeRXOLpqN/OGct64ukT3X6IP0NEfY1WEvBwsKiFdplKTiOkwU8CgxEmKErgIXA84jm2ArgXNd1KzZxij0QYYxysC9birOWVDUSKxcX2ovLPpPtJ50geAQAT58rrrIUnelu+3Ityaog+s4bUreA3AIcx5QxU7LLNTJhLdGc+PiOSAwb3YOze+ewXmc2E9dvSsCd36cDCzQa0pRyu+erQj5Vko0ymd1vO3MYTR/8VPYNGKQX2vTcUTB6DKbk7flviK7DzX9837tfM3sn6b13DAe8bMrF2h+/A+epmOs72h9/nszi8dpkXCVIiUmuB6FcTNm9HCVvl9c0c4cpPdcGrnlYMiifO+S3AMwvjVCk5OrFakUEslOJFkuf3lIr5jzN3diX0N7lw9+Ad13XPdtxnBBi6f4WmOS67r2O49wE3ITUl9xLEAWUnKsw6kazgZUbtFvJWYdKIRHDvBvVny6pQR6dJMQXjgwwEyb8HZMGvFksl0hCeh0Iv5JIycENkjxUf/9bfK7KzobEM+x/0OdwuqospWt/phbW8LMjpMDK+X+5XBomGVXnbUfPU48F4Fnd8tUsbv6d1BcuqpdBMOBzOFVVjczy6oVl5VRqHIGRczdJWM3xOFNWynOJlejzbq4FxDOypFB+wvmdUjnxgoM20zutUalek8mLS73yeN+p0M0tw7tyxwx5Yq+qIM3pQ7tsyyPYK7DdywfHcTKAo9ECsq7rNrsytY0DntRmTwKnt7eTFhYWuw7tsRR6I7XG/u04zhBgFnA9kOe67joA13XXOY6T2/5u7k4IQUBj8KNGfmxDK0Hwvy+k2EhJo7q3NHpwdNdMz0IwuOOOX210/FlHnSvnmf5CYmevAzdqR7L421OuOoysG98E8KIMR6hYSWF9hKyLZDlzwYEDZbt2CRSo8EuRipakNEJGfuvz14iF8/kDr/Pn98RS6aEuz6L6CHddKeftNX5c6+MOGsY9b2uU4VsfAHDRfVMZpWTsg1oR+7KVlV5BG1OE5ch8aZMT9nsJWvNLTAxDNUajMU+f6cEdUwmNPnTjZ7MBrrtQ+tMtLcRtM4XkNcupueUNHpn5hUZYXv7q/C2ec29De4jGAHAw8Ijrugchuts3bf6QBBzHucZxnJmO48xsRx8sLCx2MNpjKawB1riuO0P/fgkZFNY7jtNZrYTOQHFbB7uuOxGYCOA4zp4jrh8cADqbEP1ss03T82XGzVKh0uc0S/GRp65po3XDRnt+9oOTALjrzOu45Zk7AJjyqjzubyfN5MeP3tL6gPyBXHT2CgBef01IyIfny+PvlhbijE6mxoP+txsrASC/V+JzufTzuesfBfACkBZVNXouz3U6exfWNXPqn6T6061TRZH5vMcTVk/9828BkHLeaQA8k9eBB+8VIrJSc0FSAj76a87D089L2TtDPHZKDzNvoqREz/zfdAAOPfsKjARdpZKFT58/GPwZbBpi1X3+oZSaP6N/J0+wxkjLDeuY4kVZfqEVq15evRU8z16G7bYUXNctAlY7jmNC7MYA3wKvA+N133jgtXb10MLCYpfCcdtRAcdxnKGISzIELAMuRwaaF5C64auAc1zXLd/CefYgS+FCiBixjymbadiXUMExAPh0Vk3VIJ/S4kcZ2OM4AM4+VKyJO/73X0Q8BGCzj8uD65oMvkR4b/HfJfL/1pdbr4W/Ka+nb4bMiE9NvYfN4ewjxVn0rQqlFursmR0OeHkLS3SWj8RdkvW+jNDrVf3FM/DrV25j+Z+eACBLvQvZt1zJ+vulrHueiqRGG2M0qgnym49XAHCLBlqlh3yknXKAfHn4KNk21OCkjNXeiifoxStHcfajV+o+de3WiVVAMJm69+cAcPCN7wAw84IhlFWL9bBa7+/A7GRPlGbsK1qDs9xYCq9s9pntIZjluu4hW2rULpek67pfA21dZEx7zrt7Qis/RwqBL7fYune/sSxbKqa701HM2sqpN+i3LvNXCbE3f9VH3jHrvhZNxvwhGjWoNRvefeRluuYJ8TboLNUhpBy+0SrSgw7wznHyPyV1u3+21nHQGIbc5CBBv8OmIeNyoNOVxDQp6Ng+Oa1arKlt9tKjTcxFfTROUH2L6+pk8PubJh39Guh1sbgnP7ztRQCOI07euSPlhO/Lc5w3Zx1DTxKD8wrVcjSp5WG/w2GmXsXh2pHkdFx3CgD7d5OlxQ1Tl3PSva8CkKrpz2iMxC0vfOMlcC3SOhfvLyvnrKGyZPlO+zunrN7LwfArURtrVbR2z5m72gMb0WhhYdEKNvdha2GiDQMORBo33S5dKictW7wMAuI2u+sIibQL9x/gNXNdce3dctkfAJg+638JC6FBudmg/Pec+KNzWhzXtuSYwcy3fwjA8WcISZipgUohv8OX6mZ76od/ASBvSGemq4vxLs2M3K9LhneMyVEweRqHdkr1IiZ76Gw8q6SOsFoK2pyR6k6kpgg052GERg02/PMFkq8V8RmuFHJzKMBUIStNrYuQHnfYDadCQVtRhXL9havlPr959AUue2IWAOV6joFaFu7Bjxdx4ngxM0IaHDVtXTVnaeq0yWZ9aXk5n6jATZKmaddVak0I+gKL2+jH3gdrKVhYWLRCu4jGHdaJ3ZpoHAzAQ2fLbP2zl54AlrbRzuQJGA9sfwwJ9thNQipecU+igpGj0+rkd6cBMHrsSL74+UMAfLlW1r1mfXv4NcNhqAQcMU2qQs146Wv6HCkzaMeTdbFd3sDHE98D4LNvpR8LNUsyjsuLmnNgJMcWFtZ4scYdVSiltKaRTkpIGsETI6qanxykUEvGG0m3vOQgKzXDcqFKnR2uFaKOGZDL3c/c3PoxVRRC9gahw+Xl8Oy78vknF7Kj8MyPHgDgtvcXM04tFcOxvLK8gpIfHQZAqSo33zWrkAPVujDh1ue8IpoWDZXFwLs7rG/fE7aKaLSDQguMGiXs9tSpiYrK/zz3LgA+WS0eh6c++8fWnSxwLiQJ0eXOvk329evrff3QX54D4Gc3iCn9j3P/4OkZmhJrLdOgTaGV15R0c4Ez9OVbqwTftxUN1Cmjn6aCI2uVWU/y+5inMvHfXCI5AtOWlzN+8jIASnVZ4PM5jFGVJ/NiNCrxGHNdz+uQrP05uEOKJztvkq9MGvQnRbU8fKrEQhx3n/EMtEhv1kK3X//8Xwy9XdPGO3bd+Fm2E5OvvZ9TX5DB9LWzRSn7uMc+44wjpIL208fK/8t7qyuJ6z2c1F2ebUwHzQP+M5u1yw3BPGuH93EXYasGBbt8sLCwaAVrKQAbPgMx7WVGIajKxpEX2BYkd7mabPXpr1389022W/qy5ARccMdbng6jIcqMAvJbqyrJTZJzZauQyeF5aazXaL4ytTCcFp9NWrARN3lndZWXe3GvphgfkJPMFR+JpfBNhczamUG/l0VpCq8elivLmIWVjR75GFL3ZtfUkKdBafr2iQqfBHyOd6yRhTsqP93TqjQFWu4+ri/db71qk8+ovXCcAcAC/atAt2tbtDDZlakQEpGXw3qJ9XDjYZJF2jk5yK9nSHr8p1+/o+0X7qwu7yxYS8HCwmLbYV2SLTBlypQWf6nbMdJm6sZmILNxTkqQO4YVbKEt9DlOLJLPF6/lz5+LbFutxuS/qFJpQ3JSqFWuwLgLI3HXq59g9pU0Rgn7ZJzfXy2EkM7QI/LSvGzDv2uwTqfkgDeTm0pVSQEfPdJk3Z8ZFKvDZDA6TsLt2F/zBUoaoh4haeo9GDLv0E6p3uch+TIDf1Ne7+WCjNJS9Jndt6DhUKHFZrPzN99uQ9SbyNAFLXaubaPhV4mPqmEzY6HkZ5yjxkAo5RiyO6rVmKqTbd0eZylsFaylYGFh0QrWUgCOHXs2AJPf/1+Lvd9tsN069OwsHoyc5CBjVD59s8iQPIHlFfVeBakFmnNwis6gc8rq6ZIqs7Fh9ueU1XsBRGpYUB+Je2pCwzrKzLywSv6ubo4xUnUMTABSl9SgJ7pakJqo62iuf5QGIX2nf4d9PjooH2AKxyYHfF7dxaXKLdTqd/2zkrhzuEqkKb+yrKiWJXo+Y7lkFtdBjeYYpG9gNXz6MX3HiQdoyYInZF/Hzpt4mBsgRcK0Rw06hanfvLl1x2wCzfWTWb9qcrvOsafADgpsOBi0D727SFRiXTTO9OXiPuy5Fcf97pNV3ss3WmskzFYV40jc9VSOvZwAn+OZ68Zl+NaqSn44QDRt3taU34Da+/tlJXnH9tE4hJpI3FuiGBfjgOxkj6Q0RWHfXy2p0+khH6PV5HfUyIzE4x4xaVIrhuiA1CMtzDK9Zj8dwAJOYlDyCsfisv7uZwHIaxHLAcCIkXTtIgPKqaMl5fwvv72SfheKoNf8ryRm5MCuYZzc1hqNkz+WZcOUuW/Acmnn9L4MgIy0ANW187RlKRYJ2OWDhYVFK1iX5A5Ghx7XAnBghxSPSLvzzQmbOUJM7qGDfk2WRgmaROg5Gml3VH66Z6KbjESXRHCTSf0N+hwvkc+IoJjj8pKDngVwhEq0raxtordaDcYNmhRwvGhFE5RUqiXoypuiXpDTmb1VBCU56AX8GAvgoA6pXh/NEmeFLi1Cfh+rNQLSKDGf1iObT1QS7fKrj5R2x43ezDODVZPElO9xnKmVsWnB8Ccm/Jvxt18mf1Rqu2ff5WeaXv7YZ0IYNtS9tNlr7gWwLkkLC4tth+UUdjDK1om7b1pRihcuvDmUviTFW0fkpfGt5il8ozO0mdHDPocSncm76czbUtugQcOis1NCzFQewsz2BRoAVdYYpY8WkzVh1BlBP6WNMrubkvWNMdcjH9MCYrPk58iavqwpyn6ZqkugXEF9NO7lUhix07Eqz/5lSZ1nRRjOoro5RsBYPSYQyudQqeTkz+6RYK5zX5Gw5DE3nApBdQ93T4jWdu9l8ic29xMWF+ZlE67nsgkiYe9ZxheOIe9d0aM442AR1v3vx5s51T4EOyjsaDRP1w9HeAVZNoeP3xQTti4a90x4Y3IbIZMVNc0ckNW6mnRywOcVkz2tpzD2r6+oJE2PCSvrZ6pOpwR8DO0gBOBkTVxK8jteHIOJjeibmURUy9u5uhap00HEh+PlWfh08RL2+1ir95ni5WcIyZkZ9nv1J8zgMDAnxRvgzOBT1RzjTa3tYGpBvKXVrb+96Vku0IK4Hf+gcQqxMPQWURa3WjQdnYy2PBIa30AqXvG9Wq0dkZnLEh3YTIQlgQMhuu+pN28Iu3ywsLBoBWsp7HCoeyslkzE9sjffFHhPy6Qtrmr0ZqxU1To0BF+XlKDndjRxAkurGxnVReIITDm44oYIB6k70GRJGtGSfplJzNKS7ul6/qDP8dqZWbN/dhJzyoyIjOMda/pYr5GVXVLEmhmQncwKJQ5NnILpfzTuUtiU0HcEWFjZ4BGeJvNzVW2TR4gadNIoydpI3FvifHLVYwCM+/cvOGfIFQD8+6U7AXjlz//mjBsvb/MZQ2dwpB9OuhSudd0ITzwmmZuv/P5lACb1HMX8JdZSsJaChYVFK1hLYSchLSNMdSS2xXYmNyA7HCBfZ8fGmKzhjUrymrpmz0JoUoshOxwgop9NluLAnGTP2ihSK8Os8yNxl3rlDUx0ZElDlIBP9o3Q6MWShig5mu1otBNM0FN2yE99TLMjlfdIDfi86EYTYGX6c1DHFAyvZ/iScErQC3Yy5GPchc7ap3S1XIwVkRnye5aFuc5d4+7mxTmPy4ldCaw6/ZeX4f7yMgDmzhIX45BDTF2LJZ6r9pCC4YmH30cyIY2l1TUnGWsn2EFhp6F2/Xre+k5Y+Ks30868XHWRBCtvmHhjUDuOQ8hn1JPl5RqQmcx0LbHWXROYgj7H8yyYuICTumV6fxsT3qApHqeuXq41MFte6CXVjR5BaiTbu6oHo6QxQqp6JMxg1hSLe4PGQB0UDLnZFHO9QckkUlU1x7w0ajMI1kfjXpKUCVgx5GaPtBBmaO2k6eOPflfCoGtFZ3LcX06WL1OTvSc2eJiSkLVCNDppiUSqmWs3VuI2Zewqm2O8F9LzNb+9Ubt9BXb5YGFh0QrWUthZcBcxu3jLZcx7aYLTrJI6TyDFiJUYeTZc13PVHZYrsQvzKxo87URT4yHs8zFL4xQGqqBKjc7okbjrpTivVksgPznozcyGQKyNxGjSmb63ti9V07+mOe7N8ob8S/b7vNRtI/3mJW+lhfhc6zgYQjPkd8jS+zPLpLjreIrKZl+W/v11WT3Ha9zDosp671x3q+DJuArpB6mm+GwLpOZttGvM/qM22pei18oK+cnsLPkTVSuT9NvNKHfvpbCWgoWFRSu0y1JwHOcXwFXIUvAbpGxcZ+A5pAbabOAS13W3HMWz16GBwsqNi8ZuiMpmmYXTQwkZtBLNIWhWIjEnKeDlPJjMxcqmKL0yxMowM3nnlKC37jYzeoOZecMB1ilhZ2bj0qYoh6vIyooa+S/KCgW882lGthd4lOR3vPRuc1xdNO5d0+RgmLlmQWUjmcoVmIzOnHDAcz8at+aKmiaPhzD3bKyOnKQAxQ2t+w0wWy2Q2reEGky7dhAbQyyL2yc8wh0TJPty0sLVG7X6VgVt31xV6aWmN3Q9BYBFa/b6fIiNsN2WguM4BcB1wCGu6w5E8njOB+4DHnBdtx+SpXLlps9iYWGxu6G9nEIASHYcJ4LEka4DjgWMeP+TwATgkXZeZw+EQwxni63O7CnZhi8uK/fyCcysaSTT0/x+unvZhjKDdkwKosS+F3Jc2RQjI5QIHYaExVCQEmSJFlQ1FZ/8PoevNKDJYFVtM0M6CB9hcirMzJ4c8FHheUZk37LqJk+araihdSZnfTTuWRnGyxKNuyQH5GdnuJNvK+OeK3WG5k8crtyJ44igDOBdu6wxSlStmcUqeX8QMRL5pWLNEBULYMLYwTR9JjU473n3aTaEEZgpb4rxqyHiqbj61ekbtdtXsN2Dguu6ax3H+TNSWboBeB8RxK90E+WQ15CQz90KBIHI9nZpN0M6oaQtP948LT7iLnM95WVjOhsXZWbIT43GGBjyMS856ImaGPddJO7SR5cUKQF5gcwLWtYY9WIASjSGYWiHFEqUMDSDSO+MMHo6aiPSLtdEFzbFvLgDs+wpaYzSRwexDD2/cUNWNEe9WAujp1IXidMjTfaZVOtOSQEvj6OLKjSt06IzEdf1+m0GSRc8RZfrtOzd+yPeI3mw1tXQSE+Cmg9x+AjueXcEm8J3moh2bu9sL4nNbSzZZPu9He1ZPmQD44BeQBck6+SkNpq2qZXgOM41juPMdBxn5vb2wcLCYsejPcuH44DlruuWADiO8zIwAshyHCeg1kJXoLCtg13XnQhM1GNdgN+ffQu3vvKENIgtb0fXdgd0pLlqy+6soAb8dEsNe7UXTKSfIQZzk4MsrTZSamIJlDREvRnZBBn5HSeR0ahWhMl0rI3GvBk3SWfZpljcu0Z+SsKllxSQ74saxHow6ddRn+ORm+VaE6JjUoKYTNF+mHyLrqkhr6Sc6TcklhJmeVJUH/ECmYxuoymr5wfqlWA0SyFHOu9dH+Cg37zjyczdMEiWAGMmai6Ek8Pm8NPXJU27b6cchqpAjMxx+yba45JcBRzuOE6KI/+DY4BvgcnA2dpmPPBa+7poYWGxK9EeTmGG4zgvIW7HKCKePxF4C3jOcZy7dN9jW3vOW1+6fXu7sxsiBM31W2yV1lXWsHUtZnJDBJpZubI56rn2DOmXFHC8GdrMljlhv8cNGJi1W0444AU7DVcSb3Vtszf7mlyD7mlh77x99frG6shPCXqchuEqcpMDXph1s87oTbqNuS7Z2u+odiQj6Pc4k346s1c0RT3OxPTDuF59TiKwaY0e53MA3dfsib8mnocJE2eN8gLd2rYUKj+QkOdY4yQAFq6u4KynJFCKwNGyjbZU8+6u21rdlrM3ol3eB9d1bwc2fJOXAcPbaL6PwY9HmsZ1cPClbNxMX65OyUFP6MRE85kEoOZYHFeJdWNe90lPojGqiUrK4jdEExGH5uU1QiaRuOvFAJgXOjXoIzmS+N5sayOti9SaUnEVTYnCL+bFzg4FE94MbWfyI2Jx1xN7MSZpRVPUi+IsVpIz2uKaJnfDLKFKGyPeZxM5GXOhk5Kypn1DNO4NKF+XiQfjrNUqstJtf9rCP/5uyr+ZFPcKQEVYom/oPrOsygdk2XPcQKlAnpudxX8/fka/X9PmNfZE2IhGCwuLVrC5DzsNLiYmcNbjMusMu+q8jZupRFrHpICXk2Bmd58jxxekhjwJM+O2/LKkzpu1TVm3yuaYt686IvsyDAnpc7ylhyECs8OBVqnbIFaEiRv0Mhz176DP8dykxmqvjsS8GAoTu9BQJ2folhbyXJiGLEwL+rzitGq4UBOJee5Gc968FhmUJm18rrot+2SEvSxQYyXlJge8vhmLpWiaFM/NH7FxvgPA49/p8iLQS7bRZW20Mi7yRCTkh/OW6Kc+kDZaPtZ+rvuWsKfDWgoWFhatYC2FnYY4plrpS/+T+PxhbVVb79EDgIDzlVcUdnVLQg0h08wsb8i2gtQQ2brGNmv/uBv0RFjMbGkIvoUVDYxVbYXp62r1O9ebaU37wvpmuqsL0q/XNLkHeclBT7ats7owkwM+LxgqqBaAyR8oa4p663yzrYvGvapVBp1TEpZQY1Tuz7g8yxqjVCk/Ylypsbjr8QsmmjMSdz2rxFg46RoY1ibmL/QCt4hub6HYKoJ6Kzm5knuxvnjPtxTsoLDT8CXQG4DnFkjl6nvaatZVUnX3y0xinsYpGJPeRAM6Tpy6iElKSkilV+spDHGYEw6wqs6EQct/rSHwgj5nI/GU4oaINxiY5UNNc5xczRo2HgDTpksL74Mh/2oiCYIv3CL+AaB7WoiqJkOWSv+zQn5vqWDaFTVEPOl1400wqtS5yUHvJTdCM+khv+eR8LfwjJh+mvapV5228fNWTLz3FSINZmmwrZXFBScPPZ8D84Wk/NO7D27XOXZH2OWDhYVFK1hLYaehEhMVV1K9uXgFMcPDQZ/n00/2G5ehzIbfFTeQ5EUvyr7sUMCb8Qs1KjESdz1S0LjqTBLRgdnJrKlrnVeSEw54yw1jWYR8jlcubpCa3+AfRmoAABl5SURBVCaGoDnuetaLQbiFBFxOWJYdWaFEaTsTHRn2J9SlDSHougmLZUNS00QfxF3XW0o0az2KsN/nJY0Zi6Ux7noFa0doSbs2XcCKX3ywhEFa1u+L4u3L7L9kYAEXvvKN/lW1XefYHWEtBQsLi1awlsJOQwqggh1RM/+ZXIikjVrn7NeRjz6ScundNEKwSDMFu6cnXHAGwYDfIyQNCRnwOV6ewvr61laBz3GI6brdEH2RuOuVoWsZxRjbwD1oZujihohX08EIrVZHYtTpsSb1e39NpV5W3eRxBUZAJu4mSMdva8QiibkiJSefxSowQVpxElaMScN2nLjntjWWUHbI75GwB//ugg0fbwLVUtavvqyGksyN/x+2BVe9Phc3ZubV03VrIkrfaOOIPQPWUrCwsGgFaynsNByMsQjiNSLFzhwNjhkyYOPm3TK9gB1TI8Gsqyuaot7a36zXS5sSWZKGgW+KuTSoBLzRKti/hUaDya0wOQodkoKeB8DwB02xhCaD4RJMv/yO43kfTC5GYV3zRrkJpj/pQb8nRutvEb5coftMlmRTLO7pMxi3o7GtAo7jaSu0zI8w1oZ5SB2TAhzRU8OVMzZdmWvmi1PlQzTCmqL28QB11c8CRhw2Tbd7frizHRR2GlLwlgkZQni998wUAMa2NSh0zvBeRkMcmmIyDdE4ucnyX2UGgqrmmPdimBe1LpLQMDRmvnHxNcZcMlWVqSBVrpMW9Hs1G0zF6J7pIa/obKfk1j+PtXUR70VeqcuZjJDfe4GLNliy9MnweQOKcZvWRGI06ABXpfe3pq7Zi4o0yxiTLJXk93luWJPmHXddL0Xc5FH0Sg9z7RO/Zku48eHP9FMlkZo2pT62Ees32O75sMsHCwuLVrCWwk5DKp6bqkEe84ffSNbe2Laa9+5KRkjEPkx25Lq6hOCICcwxy4iKppg3oxtC0Ock8iCMAW6Clzq2kIaraBFQZFx7plZDzE0ELRliz8slqE8sQczMv7Yu4p3bnMOQkEGf47kMzT1lhvyeWnWSzwQlOR75adyyIc8t62OlBl2ZazZE456uorEYLrjw8LaeagtIu6lfG/GeLHBnbOGYfRPWUrCwsGgFaynsLGRlQqUSjBFRpHt2kYbEEiGRp68I57JcJdcMYWfyAbqkBL1ZcnWLUGVTByEnbLIIm7zgJsPD1TRLm4E5yZ7IitFOqGpRANcEILlugqMw5zLaCR1a1KI05KJLYgbvnJxQRQYJMjKkpiEfY3EIBE1ehvSjIDVElp7b5FkYiyfoc7zwacMtBBzHC+wqVBIy54zj2Bye/MNT5k51m8kmlAL3edhBYWeh8jPwa4GS2DoAik1xmBlfw2GHbnBAkG79OgLgWy6y5cYsD/gcT+vQEyHx+1ilA0RDSoLZNyXZTO5AfVSOa4jGNxJxqYvGvRfziE7Cnr+6osJbItS3UIkGWFrTRB8l9kzcRNDntFq+QCKKsaY55iVJmQrZaUHHG9hMjERhfbM3eJiYC2PC1kZipAVaezf6ZITx69Lj+UcuZmtw2S3P6qeQbj/equP2Rdjlg4WFRStYS2GnYSE5nUXco3xNGQCRcsnG++sf3+Xn/9vQUoCsy44HoK8WT90/S/YXN0Q9098sB6piMQZqbMGsEnEn9s1M8rISzcxsCLwVtU1eXYbaaGLZYJYepu5Cz/SwF+NgYMjKZL/PE0tJbuEONecw2Z3GbZni9xF15frmnMkBnxd/YSjRZL/PO8aQmuZcGSG/V8Y+T9O1+2YmkWvUoQe2VS6uNSbe+A9A/g9IlVR16vYeF+KOhrUULCwsWsExsenfaye07sPeB3XRdR4PQOU6iXZL7diZ2pLHN3lU9LEXAPjTG6IkHPY5LNM1vIkGXF3X7JWlL/QiCRMZiwY5yjGUNcXQpb4nvJIS8HkRhKYcXeeUIBUaXZi/QSn6zFAiQtG4HaNx1+MSyjfQX0gP+j3Xoul35xSzpk/oL0TjbsLVqtfuqfkfA7KTE65ItVLG9ulAr7t/uMnn56FUrAOn0zHg7yf7Yi9v+bi9F7Nc1z1kS43s8mGnQqtHVy0CIJAjoit1pWuo+GAaANnHH73RUYFLxwDQd6r41N9cVel9Z/QKo67LvHIx+Y2vfljHVMo0QtGkVZe18CQYE90UfkkPJlSTShsN65/4vsTTcjRFZHykxFoLqvTJCHtEpBnZjaBKYyxOVbOWlFNyM+x3PKKzUds1Efd0I81AZJKygj7H68+oAbnSjxsu3eiZtQUn/3z9lATx+GbbWiRglw8WFhatYC2FXYD6erEK+nQbDMDS8lqGjH8CgFWFG1sKBDsAMDBXRVoao16U4SytEp2fEvRmfuOyK6xv9iIHTaSicRemBf2eq9GkVTeF/Z54i5mNHRyPkDTnqFZyMy3F7+UymBiGxljc02Q0hVpNfsaaumYvtsEQpZ1TghQ3RL3PoEVp9BgjwGJiEoI+h6P27yT7ttJCGHf4DfJBXcEED4CILVS2tbCWgoWFRSts0VJwHOdx4BSg2HXdgbovB3ge6AmsAM51XbdCa0r+DTgZCR27zHXd2Tun67saJgIxstlWm8PShRow4/Ri9TqtXlSs9QRyu23U/oBbzgBg0tX/9lyAwRYKy0uqhXwcmS+l19fWNXszvifw2iL3wUQJGsGTjKCP+RUyu5soxK6pQY+sNASjybhcVdvMAdlJrfrRHHO97EzDFRhfY0444PU7PZTof43hC0yYIwkOIcvjHqR9h6QAqddvWoC1JT7789MAvD7DBCZ11JNX0p7/t30NW2MpPAGcuMG+m4BJruv2Aybp3yCl6Pvpv2uAR3ZMNy0sLHYVtmgpuK47zXGcnhvsHgeM1s9PAlOA3+j+p1zxc37uOE6W4zidXdddt6M6vKtx+y0PADDh9z8HoHfHoSwvm7OdZ9Pjsg+C8hUA3HLtvwG465XbNm6eVQDAT28/mQcnvA3g6Q40RONeqLHRVYi5rscppKToTKtr+kjc9bIeTXjxsuomL0fCrO+j8URYs3EnGqugKRb3PAsm67A+GmdVqQRPGS9IlxTDOyTmHMNnxFyXTupBMedKCjhe/UkTim0Coo7s1wFCmxZNMaiY8gkjfvW0/qVCrE5v7eoLWzy+vbh5/G8AuPuJO+XSTninX3NnYaviFHRQeLPF8qHSdd2sFt9XuK6b7TjOm8C9rutO1/2TgN+4rjtzC+ffbeMUhnU5GICZa2fJjvJ1OB26tPOsAyBwoHyMSnq1677FZsfoYtFv/PX4JwGJVFxb31qFuKo5llAsMrqNumTomJSoOm1e8ribEC4xZn6/zLBHGCYGEfmBr6+PeEVYTHHYuJuIeDRJTQdrKbzF1Y1ezIIpd7eoqtEjKbulJnIfTM6DqWB9VL7EYBz88I8wWpdtwV0sasq+AXe0KOqSqlvzE31vk8fvKCTeI1lyOU5w042/P3wvcQpOG/vafOEdx7kGWWJYWFjsRtjeQWG9WRY4jtOZRImdNUBLxqwrm8hPdV13IjARdm9LYVbhBjxpTmeCSRI/H2lcuZ1n/ZZg6hA5R5XMiGMG/pxJ8x7e9CG5fQD444vXA3D3RQ97S4MlmnKdEfJ7uRFmmWHcirXRuLcvoBmGy2savTwII/dWG4l7Zr+Z/ExgUWMsTq5f2hk3aHM8TkWT1mBQctBEX+aEA56gS8tCtoZfNBGKPdPCXom68YfIkinzpgv1xjdhJayV8my+QXfJ39GVgJaJ8/eU7S6IXrz9ut+3+nv0gWN2+jV3NrbXJfk6MF4/jwdea7H/UkdwOFC1J/MJFhb7IrbIKTiO8yxCKnZE1ClvB14FXgC6A6uAc1zXLVeX5MOIt6IeuHxLfIJeY7e1FAxGHXkyAFOmvwWVUsLcyc5t93mdDpcD4JYVMvMNIauGnXLMlg+sKuXX50v9wqXqmnRIBA4dagKfNFBobV0zmRqubArIrq2PML9cNB5ykuS7muaE1Jk5l5ntG2LxFrJwMsv3zUhilhKNnvCK/m/mJm9siGaFAh6BacjH9KCPK047AIDARWdu+d5L1xPufSMAzTWmoKsP/F3lY8xUbVqw5XO1E4n3x9Sk2K1Df7aKU7AJUdsI140C8mMe2UPSn6ev2uK4txmIyZuacx515YvlGrH35avNlD0DoFmqR3/4i0f17xjHHCuMu39YXwC+e0bUi38/eRkj8mSpYgrGLKlu8qpIp+iSYXl1k/dyb4gkv88TSDFxBwOykr14CSOCYmTlG2JxzxOR7MnROx4JaiIhz/vhCBgxcvP3CrBCiESn/23QVKE7Vbgm3B2adaXqTtnyuXYAJr80idFnHQvA6J6yHJy6cu4uufZ2YqsGhd16WLOwsNj1sJbCNmJo5/34qtC4vtRsd9pXfkxwKKagSChJCp82Nby6A86riNTh/u8DAFyVhfvr+0tYoTJvhpBc3xDBZF8babceSlCWNcW8HIzK5kSkpPkJLVWCcWC2WD9hv+NZDYZIDDoOI9TdeMDvz5YD0zfv4v1u4ovS/tondU8ZnrsxLMVk/eEYsernt+ZJ7ADItV23AirFYnGyc3bRtdsFaylYWFhsO6ylsB1wV6lDpVs+AK/dO5HTb752B5z5VN2KfkJKisyodXVv74Bzt4GmWs48ViLwemjQUF0k5rkPm9R6MDN/3E2IoJh8h6rmmJcbYURWhqsIbNDneMVmTaDS0IM6k3zNWdqBzQX4CD8x+Ud/59gn1S3cYLzbcQiJdRFIVqGWqme27d7bAbdJVbpDaS2sxKZNtt+NYC0FCwuLbYe1FNqBls9OEkcBKtpuvE04RberZBMswG02oSA7J3z2tcv+CMCc8novWMlYDF1UQi2O6/EHxjU5u6zOk1LrqtbAEA1zPrlPDgNO3l8OOGaEXmkLHpX6agDOH3s3AM9PXwiU6pfCtfhy8kgLqdZD0RPbdqPtwG8vvRmAPzwpfXvxgb9z7g0/3WXX3wGwLsmdjZNOugKAt99+DBDXnuMk74AzqwEXUJ99tAyQWICFH/4FgP3GHLkDrtMGKotY96935VpLROOwUMvX1UZiHmFoalIsqWpi3DHiBu17mUbzpedvx4U1KjPzKgBqqjWlnIAnOuPvkAlArGQhxHZ13YZsXLe81R7HaSuqf7eGXT5YWFhsO6ylsANQs2g1af0kmm7KExJvf8zlZ23ukG1D8tmEQmKBNFfJDNo7vydL59wj3+duz8y8GyFWj5N8gXyOGJFaIT59mV3o3VlEZJZ8Z8jEHbFE2za4bgwzhzpOmu6t2+X9aCespWBhYbHtsJbCDsKGz3F0v1FMXTJtB15BchnCnc6V60XiNGsORmaWiJBMmngpw845YQdec+fik6fF1XrUJXdiQsdBCNucbpItWb7uO4hO/R56J1g+VwLVeg7aj9HHSubm1MnPbu6Q3RmWaNy1EMVh1y329jiOagSakmU7FD05cIDENRhvQdGKCtC8hR7dhJS7/lDJZB988iDGXGTSendEBOb2YeFbn9H/VEn8wi3RvV0hTZZAqSZeosgsFRp3bQcVkx8WFafRP7kIgCkfTuGY47ciUW33hl0+WFhYbDuspbCDMWqkaNxOmfaOt2/nWgxgTO7sDsfRNV9m3BIt6FqkuQ1JAf//t3f2sVlWVwD/nVBaBIIVEKxUPlxwfIlgHYGxhDJh2E5xMqOQJpBhRlhMxtgSFfsHLwluEJkDFgU/pkzH1xQVhCkyaEnQUVazrSvQDhA2GVDGUGQyCYyzP+594X36RUv73velnF9y8/Cc+7y9h9On5z33Pvc5hy9r3FuV8dQ68wsGMKBwCADf/c4IJ7wxt/XUqtjL9NkrAHhl204vPMvFZCiZLroqGNaHD6tdmH7q1DutN/4V8PBDLhfnmrUuNyc1/jX5m1r+mnwaYJGCYRjNxyKFJPHcrPn8YHFxRJZ/u9twtL3ywwAa+EKunVwEkJWRw6De7n2BL3yi1XMXlCO+jP3Z0z6KOK9kdnGLmnf1cGsP4/NdcdZ2t+Vw3629Adj9qXvTct+mD1i4zSUz+fLzM37s+K7LMyBuB+P1Xd1iaOcuWReTq1Qd8AuxF5OipJapk3/Mr1f/3J+5SEsks+EPXH3YQmOqWfvYAgAeWvh4RF763nbGFuSnQKPGiN/8OZDRDwDp4KYiesH9EbfrmEWfbi70//iQ292XcV0m1/kK1Gf9vZTlFztPn6qB/+zzP/eAP0YzUKcD5Tt2AZA3+mvEU9eL1J9o5irHpg+GYTQfixQCMGaEmzaUlu2o0zfxWy7RyDtb1gXV6VpnzKBxlO7eEhWePIl065YahcJgkYJhGM3HIoUUULPnY3oM7FdLquTf7jYXba8sCa9UG2fMwPEAlO7Z4CWXNnDFZsQAmPfivMBaBcciBcMwmo9FCinDbWjSeN2CztfXuaJ4qkvx9tPXXgimVVti7gz3SDj2/Pw6fWVlFYwceUdolVKNPZK8uuiOnvBZorvVzQxc+vYHAIx94NteciqQXlcJ7XtzsMRlp+o7elid7tJ33Y7KsYWjgqqVZtj0wTCMK0BVG23Ay7gCspUJsqeBKqACeAvITuibA+wHqoEJl/v5/jNqrW4rWfW+lqx6XxtiyaJXdcmiVxU6+pZ6ncO0G3Xd8vW6bvn6Bm2jqvrUgqX61IKlaaBv2rTypvw9NiVSWIGrDZnIFmCIqg4F/oZzBIjIIGAyMNh/5jlpo1vDDKPN0sRv8r4kRAq1+h4AViZECXMS+jYDoyxSaHkrKpykRYWTVPWsb/VzqKRcD5WUa+HQcVo4dFzK9W5pmznufr1QWaUXKqsajQr02DEdkZunI3LzUq5zGrcmRQp1ywI3n+lAvF5XL2BnQt9hLzNayMrfudyPKyXromzuzO8DEFv2Sy/Jok9+HgCb/lJrtx7AZ8fc9YvXAPCbF3/LgSN/SJLGl6d9h8EAPDl7CrFZU52w5y2X/+C/j1M0xT2ZWbWlFUvrGQAtcwoiUgycB+JpcurLea0NfHYGMKMl4xuG0fo06ZGkiPQFNqrqkATZNGAmcLeqnvGyOQCq+jN/vhmIqWqjX0f2SLL1GNxzOACxmNud9+DM+xq7/DL4VGhHatzxwEl2HnHFWipOulenT/z3HNm+BsRXurgoJq+7e/W6e+9suM0lUiGjuQVYXYq5WOwl5s170svCZ3FuYyTvkaSI3AM8DkyMOwTPBmCyiGSJSD+gP7DrSsYwDCM1XDZSEJHVQD5uC14NMBe3oJjFpfxiO1V1pr++GLfOcB74kaq+W/tn1jOGRQpBcMlP5j46DYCheQ8z6Xv5vq81lpeayGkXecSecZWo/v7711mxY1O48a9dbEej0Vp08sde3NTDZVC6uaNzMCfP/o9DR+Pl1OLZmaPl1Yy0wXY0GobRfCxSMIxrB4sUDMNoPuYUDMOIYE7BMIwI5hQMw4hgTsEwjAjmFAzDiBBwG1ujnAC+8MdU0x3TIxHTI8rVrEefplyUFvsUAESkvCnPUE0P08P0SK4eNn0wDCOCOQXDMCKkk1NIl+IGpkcU0yNKm9cjbdYUDMNID9IpUjAMIw1IC6cgIveISLWI7BeRJwKNeYuIlIjIXhHZLSKzvLyriGwRkX3+eEMgfdqJyJ9EZKM/7yciZV6PtSKSGUCHbBF5Q0SqvF1GpcIeIjLb/04qRWS1iHQIZQ8ReVlEjotIZYKsXhuIY6m/bytE5M4k6/G0/91UiMhbIpKd0DfH61EtIhNaMnbKnYKvC/EsUAAMAqb4+hHJ5jzwE1UdCIwEHvXjPgFsVdX+wFZ/HoJZwN6E84XAL7wenwKPBNBhCfCeqg4A7vD6BLWHiPQCfgjc5XOCtsPVEglljxXUrXPSkA0KcCkH++OSEC9Lsh5h6q00JQ98MhswCticcB6pHRFQj/XAeFxlqxwvywGqA4ydi7vZvglsxGXFPgFk1GejJOnQBTiIX2dKkAe1B64kwCdAV9zmuo3AhJD2oFadk4ZsADwPTKnvumToUauvxfVWGmopjxS4dBPECV4rwmerHg6UAT1V9SiAP/YIoMJi4DHggj/vBnymquf9eQib3IrLp/aKn8a8JCKdCGwPVf0nsAj4B3AUV0n3I8LbI5GGbJDKe3c6EM9/2qp6pINTaHKtiKQMLtIZWIdLMvt5qHETxr8XOK6qHyWK67k02TbJAO4ElqnqcNy281BTp4v4+fr9QD/gZlyCyIJ6Lk2Hx2YpuXdbUm+lKaSDUzgMJJYFygWOhBhYRNrjHMJKVX3Ti2tEJMf35+CK6yaT0cBEETkErMFNIRYD2SISfzclhE0OA4dVtcyfv4FzEqHtMQ44qKr/UtVzwJvA1wlvj0QaskHwe9fXW7kXKFI/V2htPdLBKfwR6O9XlzNxCyYbkj2oiAjwK2Cvqj6T0LUBmOb/PQ231pA0VHWOquaqal/c/32bqhYBJcCDAfU4BnwiIl/1oruBPQS2B27aMFJEOvrfUVyPoPaoRUM22ABM9U8hRgKn4tOMZBCs3koyF42asaBSiFtNPQAUBxrzG7gQqwL4s2+FuPn8VmCfP3YNaId8XCUucHP8XcB+4HUgK8D4w4Byb5O3gRtSYQ9gHlAFVAKv4WqMBLEHsBq3lnEO9w38SEM2wIXtz/r79q+4JybJ1GM/bu0gfr8uT7i+2OtRDRS0ZGzb0WgYRoR0mD4YhpFGmFMwDCOCOQXDMCKYUzAMI4I5BcMwIphTMAwjgjkFwzAimFMwDCPC/wEk8ojMFEt+jgAAAABJRU5ErkJggg==\n", |
||
| 676 | "text/plain": [ |
||
| 677 | "<Figure size 432x288 with 1 Axes>" |
||
| 678 | ] |
||
| 679 | }, |
||
| 680 | "metadata": { |
||
| 681 | "needs_background": "light" |
||
| 682 | }, |
||
| 683 | "output_type": "display_data" |
||
| 684 | } |
||
| 685 | ], |
||
| 686 | "source": [ |
||
| 687 | "plt.imshow(\n", |
||
| 688 | " _read(path_train_img + 'ID_ffff922b9.dcm', (128, 128))\n", |
||
| 689 | ")" |
||
| 690 | ] |
||
| 691 | }, |
||
| 692 | { |
||
| 693 | "cell_type": "code", |
||
| 694 | "execution_count": 18, |
||
| 695 | "metadata": {}, |
||
| 696 | "outputs": [], |
||
| 697 | "source": [ |
||
| 698 | "# Augmentations\n", |
||
| 699 | "# Flip Left Right\n", |
||
| 700 | "# Cropping\n", |
||
| 701 | "sometimes = lambda aug: iaa.Sometimes(0.25, aug)\n", |
||
| 702 | "augmentation = iaa.Sequential([ \n", |
||
| 703 | " iaa.Fliplr(0.25),\n", |
||
| 704 | " sometimes(iaa.Crop(px=(0, 25), keep_size = True, \n", |
||
| 705 | " sample_independently = False)) \n", |
||
| 706 | " ], random_order = True)" |
||
| 707 | ] |
||
| 708 | }, |
||
| 709 | { |
||
| 710 | "cell_type": "code", |
||
| 711 | "execution_count": 19, |
||
| 712 | "metadata": {}, |
||
| 713 | "outputs": [], |
||
| 714 | "source": [ |
||
| 715 | "# Train Data Generator\n", |
||
| 716 | "class TrainDataGenerator(keras.utils.Sequence):\n", |
||
| 717 | "\n", |
||
| 718 | " def __init__(self, dataset, labels, batch_size=16, img_size=(512, 512), img_dir = path_train_img, \\\n", |
||
| 719 | " augment = False, *args, **kwargs):\n", |
||
| 720 | " self.dataset = dataset\n", |
||
| 721 | " self.ids = dataset.index\n", |
||
| 722 | " self.labels = labels\n", |
||
| 723 | " self.batch_size = batch_size\n", |
||
| 724 | " self.img_size = img_size\n", |
||
| 725 | " self.img_dir = img_dir\n", |
||
| 726 | " self.augment = augment\n", |
||
| 727 | " self.on_epoch_end()\n", |
||
| 728 | "\n", |
||
| 729 | " def __len__(self):\n", |
||
| 730 | " return int(ceil(len(self.ids) / self.batch_size))\n", |
||
| 731 | "\n", |
||
| 732 | " def __getitem__(self, index):\n", |
||
| 733 | " indices = self.indices[index*self.batch_size:(index+1)*self.batch_size]\n", |
||
| 734 | " X, Y = self.__data_generation(indices)\n", |
||
| 735 | " return X, Y\n", |
||
| 736 | "\n", |
||
| 737 | " def augmentor(self, image):\n", |
||
| 738 | " augment_img = augmentation \n", |
||
| 739 | " image_aug = augment_img.augment_image(image)\n", |
||
| 740 | " return image_aug\n", |
||
| 741 | "\n", |
||
| 742 | " def on_epoch_end(self):\n", |
||
| 743 | " self.indices = np.arange(len(self.ids))\n", |
||
| 744 | " np.random.shuffle(self.indices)\n", |
||
| 745 | " \n", |
||
| 746 | " def __data_generation(self, indices):\n", |
||
| 747 | " X = np.empty((self.batch_size, *self.img_size, 3))\n", |
||
| 748 | " Y = np.empty((self.batch_size, 6), dtype=np.float32)\n", |
||
| 749 | " \n", |
||
| 750 | " for i, index in enumerate(indices):\n", |
||
| 751 | " ID = self.ids[index]\n", |
||
| 752 | " image = _read(self.img_dir + ID, self.img_size)\n", |
||
| 753 | " if self.augment:\n", |
||
| 754 | " X[i,] = self.augmentor(image)\n", |
||
| 755 | " else:\n", |
||
| 756 | " X[i,] = image \n", |
||
| 757 | " Y[i,] = self.labels.iloc[index].values \n", |
||
| 758 | " return X, Y\n", |
||
| 759 | " \n", |
||
| 760 | "class TestDataGenerator(keras.utils.Sequence):\n", |
||
| 761 | " def __init__(self, ids, labels, batch_size = 5, img_size = (512, 512), img_dir = path_test_img, \\\n", |
||
| 762 | " *args, **kwargs):\n", |
||
| 763 | " self.ids = ids\n", |
||
| 764 | " self.labels = labels\n", |
||
| 765 | " self.batch_size = batch_size\n", |
||
| 766 | " self.img_size = img_size\n", |
||
| 767 | " self.img_dir = img_dir\n", |
||
| 768 | " self.on_epoch_end()\n", |
||
| 769 | "\n", |
||
| 770 | " def __len__(self):\n", |
||
| 771 | " return int(ceil(len(self.ids) / self.batch_size))\n", |
||
| 772 | "\n", |
||
| 773 | " def __getitem__(self, index):\n", |
||
| 774 | " indices = self.indices[index*self.batch_size:(index+1)*self.batch_size]\n", |
||
| 775 | " list_IDs_temp = [self.ids[k] for k in indices]\n", |
||
| 776 | " X = self.__data_generation(list_IDs_temp)\n", |
||
| 777 | " return X\n", |
||
| 778 | "\n", |
||
| 779 | " def on_epoch_end(self):\n", |
||
| 780 | " self.indices = np.arange(len(self.ids))\n", |
||
| 781 | "\n", |
||
| 782 | " def __data_generation(self, list_IDs_temp):\n", |
||
| 783 | " X = np.empty((self.batch_size, *self.img_size, 3))\n", |
||
| 784 | " for i, ID in enumerate(list_IDs_temp):\n", |
||
| 785 | " image = _read(self.img_dir + ID, self.img_size)\n", |
||
| 786 | " X[i,] = image \n", |
||
| 787 | " return X" |
||
| 788 | ] |
||
| 789 | }, |
||
| 790 | { |
||
| 791 | "cell_type": "markdown", |
||
| 792 | "metadata": {}, |
||
| 793 | "source": [ |
||
| 794 | "As we have seen in EDA notebook that we have very few epidural subtypes so we need oversample this sub type" |
||
| 795 | ] |
||
| 796 | }, |
||
| 797 | { |
||
| 798 | "cell_type": "code", |
||
| 799 | "execution_count": 20, |
||
| 800 | "metadata": {}, |
||
| 801 | "outputs": [ |
||
| 802 | { |
||
| 803 | "name": "stdout", |
||
| 804 | "output_type": "stream", |
||
| 805 | "text": [ |
||
| 806 | "Train Shape: (677018, 6)\n" |
||
| 807 | ] |
||
| 808 | } |
||
| 809 | ], |
||
| 810 | "source": [ |
||
| 811 | "# Oversampling\n", |
||
| 812 | "epidural_df = train_final_df[train_final_df.epidural == 1]\n", |
||
| 813 | "train_final_df = pd.concat([train_final_df, epidural_df])\n", |
||
| 814 | "print('Train Shape: {}'.format(train_final_df.shape))" |
||
| 815 | ] |
||
| 816 | }, |
||
| 817 | { |
||
| 818 | "cell_type": "code", |
||
| 819 | "execution_count": 21, |
||
| 820 | "metadata": {}, |
||
| 821 | "outputs": [ |
||
| 822 | { |
||
| 823 | "data": { |
||
| 824 | "text/html": [ |
||
| 825 | "<div>\n", |
||
| 826 | "<style scoped>\n", |
||
| 827 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 828 | " vertical-align: middle;\n", |
||
| 829 | " }\n", |
||
| 830 | "\n", |
||
| 831 | " .dataframe tbody tr th {\n", |
||
| 832 | " vertical-align: top;\n", |
||
| 833 | " }\n", |
||
| 834 | "\n", |
||
| 835 | " .dataframe thead th {\n", |
||
| 836 | " text-align: right;\n", |
||
| 837 | " }\n", |
||
| 838 | "</style>\n", |
||
| 839 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 840 | " <thead>\n", |
||
| 841 | " <tr style=\"text-align: right;\">\n", |
||
| 842 | " <th></th>\n", |
||
| 843 | " <th>ID</th>\n", |
||
| 844 | " <th>Label</th>\n", |
||
| 845 | " </tr>\n", |
||
| 846 | " </thead>\n", |
||
| 847 | " <tbody>\n", |
||
| 848 | " <tr>\n", |
||
| 849 | " <th>0</th>\n", |
||
| 850 | " <td>ID_28fbab7eb_epidural</td>\n", |
||
| 851 | " <td>0.5</td>\n", |
||
| 852 | " </tr>\n", |
||
| 853 | " <tr>\n", |
||
| 854 | " <th>1</th>\n", |
||
| 855 | " <td>ID_28fbab7eb_intraparenchymal</td>\n", |
||
| 856 | " <td>0.5</td>\n", |
||
| 857 | " </tr>\n", |
||
| 858 | " <tr>\n", |
||
| 859 | " <th>2</th>\n", |
||
| 860 | " <td>ID_28fbab7eb_intraventricular</td>\n", |
||
| 861 | " <td>0.5</td>\n", |
||
| 862 | " </tr>\n", |
||
| 863 | " <tr>\n", |
||
| 864 | " <th>3</th>\n", |
||
| 865 | " <td>ID_28fbab7eb_subarachnoid</td>\n", |
||
| 866 | " <td>0.5</td>\n", |
||
| 867 | " </tr>\n", |
||
| 868 | " <tr>\n", |
||
| 869 | " <th>4</th>\n", |
||
| 870 | " <td>ID_28fbab7eb_subdural</td>\n", |
||
| 871 | " <td>0.5</td>\n", |
||
| 872 | " </tr>\n", |
||
| 873 | " </tbody>\n", |
||
| 874 | "</table>\n", |
||
| 875 | "</div>" |
||
| 876 | ], |
||
| 877 | "text/plain": [ |
||
| 878 | " ID Label\n", |
||
| 879 | "0 ID_28fbab7eb_epidural 0.5\n", |
||
| 880 | "1 ID_28fbab7eb_intraparenchymal 0.5\n", |
||
| 881 | "2 ID_28fbab7eb_intraventricular 0.5\n", |
||
| 882 | "3 ID_28fbab7eb_subarachnoid 0.5\n", |
||
| 883 | "4 ID_28fbab7eb_subdural 0.5" |
||
| 884 | ] |
||
| 885 | }, |
||
| 886 | "execution_count": 21, |
||
| 887 | "metadata": {}, |
||
| 888 | "output_type": "execute_result" |
||
| 889 | } |
||
| 890 | ], |
||
| 891 | "source": [ |
||
| 892 | "# load test set\n", |
||
| 893 | "test_df = pd.read_csv(input_folder + 'stage_1_sample_submission.csv')\n", |
||
| 894 | "test_df.head()" |
||
| 895 | ] |
||
| 896 | }, |
||
| 897 | { |
||
| 898 | "cell_type": "code", |
||
| 899 | "execution_count": 22, |
||
| 900 | "metadata": {}, |
||
| 901 | "outputs": [ |
||
| 902 | { |
||
| 903 | "data": { |
||
| 904 | "text/plain": [ |
||
| 905 | "(78545, 6)" |
||
| 906 | ] |
||
| 907 | }, |
||
| 908 | "execution_count": 22, |
||
| 909 | "metadata": {}, |
||
| 910 | "output_type": "execute_result" |
||
| 911 | } |
||
| 912 | ], |
||
| 913 | "source": [ |
||
| 914 | "# extract subtype\n", |
||
| 915 | "test_df['sub_type'] = test_df['ID'].apply(lambda x: x.split('_')[-1])\n", |
||
| 916 | "# extract filename\n", |
||
| 917 | "test_df['file_name'] = test_df['ID'].apply(lambda x: '_'.join(x.split('_')[:2]) + '.dcm')\n", |
||
| 918 | "\n", |
||
| 919 | "test_df = pd.pivot_table(test_df.drop(columns='ID'), index=\"file_name\", \\\n", |
||
| 920 | " columns=\"sub_type\", values=\"Label\")\n", |
||
| 921 | "test_df.head()\n", |
||
| 922 | "\n", |
||
| 923 | "test_df.shape" |
||
| 924 | ] |
||
| 925 | }, |
||
| 926 | { |
||
| 927 | "cell_type": "code", |
||
| 928 | "execution_count": 23, |
||
| 929 | "metadata": {}, |
||
| 930 | "outputs": [ |
||
| 931 | { |
||
| 932 | "data": { |
||
| 933 | "text/html": [ |
||
| 934 | "<div>\n", |
||
| 935 | "<style scoped>\n", |
||
| 936 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 937 | " vertical-align: middle;\n", |
||
| 938 | " }\n", |
||
| 939 | "\n", |
||
| 940 | " .dataframe tbody tr th {\n", |
||
| 941 | " vertical-align: top;\n", |
||
| 942 | " }\n", |
||
| 943 | "\n", |
||
| 944 | " .dataframe thead th {\n", |
||
| 945 | " text-align: right;\n", |
||
| 946 | " }\n", |
||
| 947 | "</style>\n", |
||
| 948 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 949 | " <thead>\n", |
||
| 950 | " <tr style=\"text-align: right;\">\n", |
||
| 951 | " <th>sub_type</th>\n", |
||
| 952 | " <th>any</th>\n", |
||
| 953 | " <th>epidural</th>\n", |
||
| 954 | " <th>intraparenchymal</th>\n", |
||
| 955 | " <th>intraventricular</th>\n", |
||
| 956 | " <th>subarachnoid</th>\n", |
||
| 957 | " <th>subdural</th>\n", |
||
| 958 | " </tr>\n", |
||
| 959 | " <tr>\n", |
||
| 960 | " <th>file_name</th>\n", |
||
| 961 | " <th></th>\n", |
||
| 962 | " <th></th>\n", |
||
| 963 | " <th></th>\n", |
||
| 964 | " <th></th>\n", |
||
| 965 | " <th></th>\n", |
||
| 966 | " <th></th>\n", |
||
| 967 | " </tr>\n", |
||
| 968 | " </thead>\n", |
||
| 969 | " <tbody>\n", |
||
| 970 | " <tr>\n", |
||
| 971 | " <th>ID_000012eaf.dcm</th>\n", |
||
| 972 | " <td>0.5</td>\n", |
||
| 973 | " <td>0.5</td>\n", |
||
| 974 | " <td>0.5</td>\n", |
||
| 975 | " <td>0.5</td>\n", |
||
| 976 | " <td>0.5</td>\n", |
||
| 977 | " <td>0.5</td>\n", |
||
| 978 | " </tr>\n", |
||
| 979 | " <tr>\n", |
||
| 980 | " <th>ID_0000ca2f6.dcm</th>\n", |
||
| 981 | " <td>0.5</td>\n", |
||
| 982 | " <td>0.5</td>\n", |
||
| 983 | " <td>0.5</td>\n", |
||
| 984 | " <td>0.5</td>\n", |
||
| 985 | " <td>0.5</td>\n", |
||
| 986 | " <td>0.5</td>\n", |
||
| 987 | " </tr>\n", |
||
| 988 | " <tr>\n", |
||
| 989 | " <th>ID_000259ccf.dcm</th>\n", |
||
| 990 | " <td>0.5</td>\n", |
||
| 991 | " <td>0.5</td>\n", |
||
| 992 | " <td>0.5</td>\n", |
||
| 993 | " <td>0.5</td>\n", |
||
| 994 | " <td>0.5</td>\n", |
||
| 995 | " <td>0.5</td>\n", |
||
| 996 | " </tr>\n", |
||
| 997 | " <tr>\n", |
||
| 998 | " <th>ID_0002d438a.dcm</th>\n", |
||
| 999 | " <td>0.5</td>\n", |
||
| 1000 | " <td>0.5</td>\n", |
||
| 1001 | " <td>0.5</td>\n", |
||
| 1002 | " <td>0.5</td>\n", |
||
| 1003 | " <td>0.5</td>\n", |
||
| 1004 | " <td>0.5</td>\n", |
||
| 1005 | " </tr>\n", |
||
| 1006 | " <tr>\n", |
||
| 1007 | " <th>ID_00032d440.dcm</th>\n", |
||
| 1008 | " <td>0.5</td>\n", |
||
| 1009 | " <td>0.5</td>\n", |
||
| 1010 | " <td>0.5</td>\n", |
||
| 1011 | " <td>0.5</td>\n", |
||
| 1012 | " <td>0.5</td>\n", |
||
| 1013 | " <td>0.5</td>\n", |
||
| 1014 | " </tr>\n", |
||
| 1015 | " </tbody>\n", |
||
| 1016 | "</table>\n", |
||
| 1017 | "</div>" |
||
| 1018 | ], |
||
| 1019 | "text/plain": [ |
||
| 1020 | "sub_type any epidural intraparenchymal intraventricular \\\n", |
||
| 1021 | "file_name \n", |
||
| 1022 | "ID_000012eaf.dcm 0.5 0.5 0.5 0.5 \n", |
||
| 1023 | "ID_0000ca2f6.dcm 0.5 0.5 0.5 0.5 \n", |
||
| 1024 | "ID_000259ccf.dcm 0.5 0.5 0.5 0.5 \n", |
||
| 1025 | "ID_0002d438a.dcm 0.5 0.5 0.5 0.5 \n", |
||
| 1026 | "ID_00032d440.dcm 0.5 0.5 0.5 0.5 \n", |
||
| 1027 | "\n", |
||
| 1028 | "sub_type subarachnoid subdural \n", |
||
| 1029 | "file_name \n", |
||
| 1030 | "ID_000012eaf.dcm 0.5 0.5 \n", |
||
| 1031 | "ID_0000ca2f6.dcm 0.5 0.5 \n", |
||
| 1032 | "ID_000259ccf.dcm 0.5 0.5 \n", |
||
| 1033 | "ID_0002d438a.dcm 0.5 0.5 \n", |
||
| 1034 | "ID_00032d440.dcm 0.5 0.5 " |
||
| 1035 | ] |
||
| 1036 | }, |
||
| 1037 | "execution_count": 23, |
||
| 1038 | "metadata": {}, |
||
| 1039 | "output_type": "execute_result" |
||
| 1040 | } |
||
| 1041 | ], |
||
| 1042 | "source": [ |
||
| 1043 | "test_df.head()" |
||
| 1044 | ] |
||
| 1045 | }, |
||
| 1046 | { |
||
| 1047 | "cell_type": "code", |
||
| 1048 | "execution_count": 24, |
||
| 1049 | "metadata": {}, |
||
| 1050 | "outputs": [ |
||
| 1051 | { |
||
| 1052 | "name": "stdout", |
||
| 1053 | "output_type": "stream", |
||
| 1054 | "text": [ |
||
| 1055 | "Downloading data from https://github.com/Callidior/keras-applications/releases/download/efficientnet/efficientnet-b0_weights_tf_dim_ordering_tf_kernels_autoaugment_notop.h5\n", |
||
| 1056 | "16809984/16804768 [==============================] - 1s 0us/step\n", |
||
| 1057 | "Model: \"model_1\"\n", |
||
| 1058 | "__________________________________________________________________________________________________\n", |
||
| 1059 | "Layer (type) Output Shape Param # Connected to \n", |
||
| 1060 | "==================================================================================================\n", |
||
| 1061 | "input_1 (InputLayer) (None, 256, 256, 3) 0 \n", |
||
| 1062 | "__________________________________________________________________________________________________\n", |
||
| 1063 | "stem_conv (Conv2D) (None, 128, 128, 32) 864 input_1[0][0] \n", |
||
| 1064 | "__________________________________________________________________________________________________\n", |
||
| 1065 | "stem_bn (BatchNormalization) (None, 128, 128, 32) 128 stem_conv[0][0] \n", |
||
| 1066 | "__________________________________________________________________________________________________\n", |
||
| 1067 | "stem_activation (Activation) (None, 128, 128, 32) 0 stem_bn[0][0] \n", |
||
| 1068 | "__________________________________________________________________________________________________\n", |
||
| 1069 | "block1a_dwconv (DepthwiseConv2D (None, 128, 128, 32) 288 stem_activation[0][0] \n", |
||
| 1070 | "__________________________________________________________________________________________________\n", |
||
| 1071 | "block1a_bn (BatchNormalization) (None, 128, 128, 32) 128 block1a_dwconv[0][0] \n", |
||
| 1072 | "__________________________________________________________________________________________________\n", |
||
| 1073 | "block1a_activation (Activation) (None, 128, 128, 32) 0 block1a_bn[0][0] \n", |
||
| 1074 | "__________________________________________________________________________________________________\n", |
||
| 1075 | "block1a_se_squeeze (GlobalAvera (None, 32) 0 block1a_activation[0][0] \n", |
||
| 1076 | "__________________________________________________________________________________________________\n", |
||
| 1077 | "block1a_se_reshape (Reshape) (None, 1, 1, 32) 0 block1a_se_squeeze[0][0] \n", |
||
| 1078 | "__________________________________________________________________________________________________\n", |
||
| 1079 | "block1a_se_reduce (Conv2D) (None, 1, 1, 8) 264 block1a_se_reshape[0][0] \n", |
||
| 1080 | "__________________________________________________________________________________________________\n", |
||
| 1081 | "block1a_se_expand (Conv2D) (None, 1, 1, 32) 288 block1a_se_reduce[0][0] \n", |
||
| 1082 | "__________________________________________________________________________________________________\n", |
||
| 1083 | "block1a_se_excite (Multiply) (None, 128, 128, 32) 0 block1a_activation[0][0] \n", |
||
| 1084 | " block1a_se_expand[0][0] \n", |
||
| 1085 | "__________________________________________________________________________________________________\n", |
||
| 1086 | "block1a_project_conv (Conv2D) (None, 128, 128, 16) 512 block1a_se_excite[0][0] \n", |
||
| 1087 | "__________________________________________________________________________________________________\n", |
||
| 1088 | "block1a_project_bn (BatchNormal (None, 128, 128, 16) 64 block1a_project_conv[0][0] \n", |
||
| 1089 | "__________________________________________________________________________________________________\n", |
||
| 1090 | "block2a_expand_conv (Conv2D) (None, 128, 128, 96) 1536 block1a_project_bn[0][0] \n", |
||
| 1091 | "__________________________________________________________________________________________________\n", |
||
| 1092 | "block2a_expand_bn (BatchNormali (None, 128, 128, 96) 384 block2a_expand_conv[0][0] \n", |
||
| 1093 | "__________________________________________________________________________________________________\n", |
||
| 1094 | "block2a_expand_activation (Acti (None, 128, 128, 96) 0 block2a_expand_bn[0][0] \n", |
||
| 1095 | "__________________________________________________________________________________________________\n", |
||
| 1096 | "block2a_dwconv (DepthwiseConv2D (None, 64, 64, 96) 864 block2a_expand_activation[0][0] \n", |
||
| 1097 | "__________________________________________________________________________________________________\n", |
||
| 1098 | "block2a_bn (BatchNormalization) (None, 64, 64, 96) 384 block2a_dwconv[0][0] \n", |
||
| 1099 | "__________________________________________________________________________________________________\n", |
||
| 1100 | "block2a_activation (Activation) (None, 64, 64, 96) 0 block2a_bn[0][0] \n", |
||
| 1101 | "__________________________________________________________________________________________________\n", |
||
| 1102 | "block2a_se_squeeze (GlobalAvera (None, 96) 0 block2a_activation[0][0] \n", |
||
| 1103 | "__________________________________________________________________________________________________\n", |
||
| 1104 | "block2a_se_reshape (Reshape) (None, 1, 1, 96) 0 block2a_se_squeeze[0][0] \n", |
||
| 1105 | "__________________________________________________________________________________________________\n", |
||
| 1106 | "block2a_se_reduce (Conv2D) (None, 1, 1, 4) 388 block2a_se_reshape[0][0] \n", |
||
| 1107 | "__________________________________________________________________________________________________\n", |
||
| 1108 | "block2a_se_expand (Conv2D) (None, 1, 1, 96) 480 block2a_se_reduce[0][0] \n", |
||
| 1109 | "__________________________________________________________________________________________________\n", |
||
| 1110 | "block2a_se_excite (Multiply) (None, 64, 64, 96) 0 block2a_activation[0][0] \n", |
||
| 1111 | " block2a_se_expand[0][0] \n", |
||
| 1112 | "__________________________________________________________________________________________________\n", |
||
| 1113 | "block2a_project_conv (Conv2D) (None, 64, 64, 24) 2304 block2a_se_excite[0][0] \n", |
||
| 1114 | "__________________________________________________________________________________________________\n", |
||
| 1115 | "block2a_project_bn (BatchNormal (None, 64, 64, 24) 96 block2a_project_conv[0][0] \n", |
||
| 1116 | "__________________________________________________________________________________________________\n", |
||
| 1117 | "block2b_expand_conv (Conv2D) (None, 64, 64, 144) 3456 block2a_project_bn[0][0] \n", |
||
| 1118 | "__________________________________________________________________________________________________\n", |
||
| 1119 | "block2b_expand_bn (BatchNormali (None, 64, 64, 144) 576 block2b_expand_conv[0][0] \n", |
||
| 1120 | "__________________________________________________________________________________________________\n", |
||
| 1121 | "block2b_expand_activation (Acti (None, 64, 64, 144) 0 block2b_expand_bn[0][0] \n", |
||
| 1122 | "__________________________________________________________________________________________________\n", |
||
| 1123 | "block2b_dwconv (DepthwiseConv2D (None, 64, 64, 144) 1296 block2b_expand_activation[0][0] \n", |
||
| 1124 | "__________________________________________________________________________________________________\n", |
||
| 1125 | "block2b_bn (BatchNormalization) (None, 64, 64, 144) 576 block2b_dwconv[0][0] \n", |
||
| 1126 | "__________________________________________________________________________________________________\n", |
||
| 1127 | "block2b_activation (Activation) (None, 64, 64, 144) 0 block2b_bn[0][0] \n", |
||
| 1128 | "__________________________________________________________________________________________________\n", |
||
| 1129 | "block2b_se_squeeze (GlobalAvera (None, 144) 0 block2b_activation[0][0] \n", |
||
| 1130 | "__________________________________________________________________________________________________\n", |
||
| 1131 | "block2b_se_reshape (Reshape) (None, 1, 1, 144) 0 block2b_se_squeeze[0][0] \n", |
||
| 1132 | "__________________________________________________________________________________________________\n", |
||
| 1133 | "block2b_se_reduce (Conv2D) (None, 1, 1, 6) 870 block2b_se_reshape[0][0] \n", |
||
| 1134 | "__________________________________________________________________________________________________\n", |
||
| 1135 | "block2b_se_expand (Conv2D) (None, 1, 1, 144) 1008 block2b_se_reduce[0][0] \n", |
||
| 1136 | "__________________________________________________________________________________________________\n", |
||
| 1137 | "block2b_se_excite (Multiply) (None, 64, 64, 144) 0 block2b_activation[0][0] \n", |
||
| 1138 | " block2b_se_expand[0][0] \n", |
||
| 1139 | "__________________________________________________________________________________________________\n", |
||
| 1140 | "block2b_project_conv (Conv2D) (None, 64, 64, 24) 3456 block2b_se_excite[0][0] \n", |
||
| 1141 | "__________________________________________________________________________________________________\n", |
||
| 1142 | "block2b_project_bn (BatchNormal (None, 64, 64, 24) 96 block2b_project_conv[0][0] \n", |
||
| 1143 | "__________________________________________________________________________________________________\n", |
||
| 1144 | "block2b_drop (FixedDropout) (None, 64, 64, 24) 0 block2b_project_bn[0][0] \n", |
||
| 1145 | "__________________________________________________________________________________________________\n", |
||
| 1146 | "block2b_add (Add) (None, 64, 64, 24) 0 block2b_drop[0][0] \n", |
||
| 1147 | " block2a_project_bn[0][0] \n", |
||
| 1148 | "__________________________________________________________________________________________________\n", |
||
| 1149 | "block3a_expand_conv (Conv2D) (None, 64, 64, 144) 3456 block2b_add[0][0] \n", |
||
| 1150 | "__________________________________________________________________________________________________\n", |
||
| 1151 | "block3a_expand_bn (BatchNormali (None, 64, 64, 144) 576 block3a_expand_conv[0][0] \n", |
||
| 1152 | "__________________________________________________________________________________________________\n", |
||
| 1153 | "block3a_expand_activation (Acti (None, 64, 64, 144) 0 block3a_expand_bn[0][0] \n", |
||
| 1154 | "__________________________________________________________________________________________________\n", |
||
| 1155 | "block3a_dwconv (DepthwiseConv2D (None, 32, 32, 144) 3600 block3a_expand_activation[0][0] \n", |
||
| 1156 | "__________________________________________________________________________________________________\n", |
||
| 1157 | "block3a_bn (BatchNormalization) (None, 32, 32, 144) 576 block3a_dwconv[0][0] \n", |
||
| 1158 | "__________________________________________________________________________________________________\n", |
||
| 1159 | "block3a_activation (Activation) (None, 32, 32, 144) 0 block3a_bn[0][0] \n", |
||
| 1160 | "__________________________________________________________________________________________________\n", |
||
| 1161 | "block3a_se_squeeze (GlobalAvera (None, 144) 0 block3a_activation[0][0] \n", |
||
| 1162 | "__________________________________________________________________________________________________\n", |
||
| 1163 | "block3a_se_reshape (Reshape) (None, 1, 1, 144) 0 block3a_se_squeeze[0][0] \n", |
||
| 1164 | "__________________________________________________________________________________________________\n", |
||
| 1165 | "block3a_se_reduce (Conv2D) (None, 1, 1, 6) 870 block3a_se_reshape[0][0] \n", |
||
| 1166 | "__________________________________________________________________________________________________\n", |
||
| 1167 | "block3a_se_expand (Conv2D) (None, 1, 1, 144) 1008 block3a_se_reduce[0][0] \n", |
||
| 1168 | "__________________________________________________________________________________________________\n", |
||
| 1169 | "block3a_se_excite (Multiply) (None, 32, 32, 144) 0 block3a_activation[0][0] \n", |
||
| 1170 | " block3a_se_expand[0][0] \n", |
||
| 1171 | "__________________________________________________________________________________________________\n", |
||
| 1172 | "block3a_project_conv (Conv2D) (None, 32, 32, 40) 5760 block3a_se_excite[0][0] \n", |
||
| 1173 | "__________________________________________________________________________________________________\n", |
||
| 1174 | "block3a_project_bn (BatchNormal (None, 32, 32, 40) 160 block3a_project_conv[0][0] \n", |
||
| 1175 | "__________________________________________________________________________________________________\n", |
||
| 1176 | "block3b_expand_conv (Conv2D) (None, 32, 32, 240) 9600 block3a_project_bn[0][0] \n", |
||
| 1177 | "__________________________________________________________________________________________________\n", |
||
| 1178 | "block3b_expand_bn (BatchNormali (None, 32, 32, 240) 960 block3b_expand_conv[0][0] \n", |
||
| 1179 | "__________________________________________________________________________________________________\n", |
||
| 1180 | "block3b_expand_activation (Acti (None, 32, 32, 240) 0 block3b_expand_bn[0][0] \n", |
||
| 1181 | "__________________________________________________________________________________________________\n", |
||
| 1182 | "block3b_dwconv (DepthwiseConv2D (None, 32, 32, 240) 6000 block3b_expand_activation[0][0] \n", |
||
| 1183 | "__________________________________________________________________________________________________\n", |
||
| 1184 | "block3b_bn (BatchNormalization) (None, 32, 32, 240) 960 block3b_dwconv[0][0] \n", |
||
| 1185 | "__________________________________________________________________________________________________\n", |
||
| 1186 | "block3b_activation (Activation) (None, 32, 32, 240) 0 block3b_bn[0][0] \n", |
||
| 1187 | "__________________________________________________________________________________________________\n", |
||
| 1188 | "block3b_se_squeeze (GlobalAvera (None, 240) 0 block3b_activation[0][0] \n", |
||
| 1189 | "__________________________________________________________________________________________________\n", |
||
| 1190 | "block3b_se_reshape (Reshape) (None, 1, 1, 240) 0 block3b_se_squeeze[0][0] \n", |
||
| 1191 | "__________________________________________________________________________________________________\n", |
||
| 1192 | "block3b_se_reduce (Conv2D) (None, 1, 1, 10) 2410 block3b_se_reshape[0][0] \n", |
||
| 1193 | "__________________________________________________________________________________________________\n", |
||
| 1194 | "block3b_se_expand (Conv2D) (None, 1, 1, 240) 2640 block3b_se_reduce[0][0] \n", |
||
| 1195 | "__________________________________________________________________________________________________\n", |
||
| 1196 | "block3b_se_excite (Multiply) (None, 32, 32, 240) 0 block3b_activation[0][0] \n", |
||
| 1197 | " block3b_se_expand[0][0] \n", |
||
| 1198 | "__________________________________________________________________________________________________\n", |
||
| 1199 | "block3b_project_conv (Conv2D) (None, 32, 32, 40) 9600 block3b_se_excite[0][0] \n", |
||
| 1200 | "__________________________________________________________________________________________________\n", |
||
| 1201 | "block3b_project_bn (BatchNormal (None, 32, 32, 40) 160 block3b_project_conv[0][0] \n", |
||
| 1202 | "__________________________________________________________________________________________________\n", |
||
| 1203 | "block3b_drop (FixedDropout) (None, 32, 32, 40) 0 block3b_project_bn[0][0] \n", |
||
| 1204 | "__________________________________________________________________________________________________\n", |
||
| 1205 | "block3b_add (Add) (None, 32, 32, 40) 0 block3b_drop[0][0] \n", |
||
| 1206 | " block3a_project_bn[0][0] \n", |
||
| 1207 | "__________________________________________________________________________________________________\n", |
||
| 1208 | "block4a_expand_conv (Conv2D) (None, 32, 32, 240) 9600 block3b_add[0][0] \n", |
||
| 1209 | "__________________________________________________________________________________________________\n", |
||
| 1210 | "block4a_expand_bn (BatchNormali (None, 32, 32, 240) 960 block4a_expand_conv[0][0] \n", |
||
| 1211 | "__________________________________________________________________________________________________\n", |
||
| 1212 | "block4a_expand_activation (Acti (None, 32, 32, 240) 0 block4a_expand_bn[0][0] \n", |
||
| 1213 | "__________________________________________________________________________________________________\n", |
||
| 1214 | "block4a_dwconv (DepthwiseConv2D (None, 16, 16, 240) 2160 block4a_expand_activation[0][0] \n", |
||
| 1215 | "__________________________________________________________________________________________________\n", |
||
| 1216 | "block4a_bn (BatchNormalization) (None, 16, 16, 240) 960 block4a_dwconv[0][0] \n", |
||
| 1217 | "__________________________________________________________________________________________________\n", |
||
| 1218 | "block4a_activation (Activation) (None, 16, 16, 240) 0 block4a_bn[0][0] \n", |
||
| 1219 | "__________________________________________________________________________________________________\n", |
||
| 1220 | "block4a_se_squeeze (GlobalAvera (None, 240) 0 block4a_activation[0][0] \n", |
||
| 1221 | "__________________________________________________________________________________________________\n", |
||
| 1222 | "block4a_se_reshape (Reshape) (None, 1, 1, 240) 0 block4a_se_squeeze[0][0] \n", |
||
| 1223 | "__________________________________________________________________________________________________\n", |
||
| 1224 | "block4a_se_reduce (Conv2D) (None, 1, 1, 10) 2410 block4a_se_reshape[0][0] \n", |
||
| 1225 | "__________________________________________________________________________________________________\n", |
||
| 1226 | "block4a_se_expand (Conv2D) (None, 1, 1, 240) 2640 block4a_se_reduce[0][0] \n", |
||
| 1227 | "__________________________________________________________________________________________________\n", |
||
| 1228 | "block4a_se_excite (Multiply) (None, 16, 16, 240) 0 block4a_activation[0][0] \n", |
||
| 1229 | " block4a_se_expand[0][0] \n", |
||
| 1230 | "__________________________________________________________________________________________________\n", |
||
| 1231 | "block4a_project_conv (Conv2D) (None, 16, 16, 80) 19200 block4a_se_excite[0][0] \n", |
||
| 1232 | "__________________________________________________________________________________________________\n", |
||
| 1233 | "block4a_project_bn (BatchNormal (None, 16, 16, 80) 320 block4a_project_conv[0][0] \n", |
||
| 1234 | "__________________________________________________________________________________________________\n", |
||
| 1235 | "block4b_expand_conv (Conv2D) (None, 16, 16, 480) 38400 block4a_project_bn[0][0] \n", |
||
| 1236 | "__________________________________________________________________________________________________\n", |
||
| 1237 | "block4b_expand_bn (BatchNormali (None, 16, 16, 480) 1920 block4b_expand_conv[0][0] \n", |
||
| 1238 | "__________________________________________________________________________________________________\n", |
||
| 1239 | "block4b_expand_activation (Acti (None, 16, 16, 480) 0 block4b_expand_bn[0][0] \n", |
||
| 1240 | "__________________________________________________________________________________________________\n", |
||
| 1241 | "block4b_dwconv (DepthwiseConv2D (None, 16, 16, 480) 4320 block4b_expand_activation[0][0] \n", |
||
| 1242 | "__________________________________________________________________________________________________\n", |
||
| 1243 | "block4b_bn (BatchNormalization) (None, 16, 16, 480) 1920 block4b_dwconv[0][0] \n", |
||
| 1244 | "__________________________________________________________________________________________________\n", |
||
| 1245 | "block4b_activation (Activation) (None, 16, 16, 480) 0 block4b_bn[0][0] \n", |
||
| 1246 | "__________________________________________________________________________________________________\n", |
||
| 1247 | "block4b_se_squeeze (GlobalAvera (None, 480) 0 block4b_activation[0][0] \n", |
||
| 1248 | "__________________________________________________________________________________________________\n", |
||
| 1249 | "block4b_se_reshape (Reshape) (None, 1, 1, 480) 0 block4b_se_squeeze[0][0] \n", |
||
| 1250 | "__________________________________________________________________________________________________\n", |
||
| 1251 | "block4b_se_reduce (Conv2D) (None, 1, 1, 20) 9620 block4b_se_reshape[0][0] \n", |
||
| 1252 | "__________________________________________________________________________________________________\n", |
||
| 1253 | "block4b_se_expand (Conv2D) (None, 1, 1, 480) 10080 block4b_se_reduce[0][0] \n", |
||
| 1254 | "__________________________________________________________________________________________________\n", |
||
| 1255 | "block4b_se_excite (Multiply) (None, 16, 16, 480) 0 block4b_activation[0][0] \n", |
||
| 1256 | " block4b_se_expand[0][0] \n", |
||
| 1257 | "__________________________________________________________________________________________________\n", |
||
| 1258 | "block4b_project_conv (Conv2D) (None, 16, 16, 80) 38400 block4b_se_excite[0][0] \n", |
||
| 1259 | "__________________________________________________________________________________________________\n", |
||
| 1260 | "block4b_project_bn (BatchNormal (None, 16, 16, 80) 320 block4b_project_conv[0][0] \n", |
||
| 1261 | "__________________________________________________________________________________________________\n", |
||
| 1262 | "block4b_drop (FixedDropout) (None, 16, 16, 80) 0 block4b_project_bn[0][0] \n", |
||
| 1263 | "__________________________________________________________________________________________________\n", |
||
| 1264 | "block4b_add (Add) (None, 16, 16, 80) 0 block4b_drop[0][0] \n", |
||
| 1265 | " block4a_project_bn[0][0] \n", |
||
| 1266 | "__________________________________________________________________________________________________\n", |
||
| 1267 | "block4c_expand_conv (Conv2D) (None, 16, 16, 480) 38400 block4b_add[0][0] \n", |
||
| 1268 | "__________________________________________________________________________________________________\n", |
||
| 1269 | "block4c_expand_bn (BatchNormali (None, 16, 16, 480) 1920 block4c_expand_conv[0][0] \n", |
||
| 1270 | "__________________________________________________________________________________________________\n", |
||
| 1271 | "block4c_expand_activation (Acti (None, 16, 16, 480) 0 block4c_expand_bn[0][0] \n", |
||
| 1272 | "__________________________________________________________________________________________________\n", |
||
| 1273 | "block4c_dwconv (DepthwiseConv2D (None, 16, 16, 480) 4320 block4c_expand_activation[0][0] \n", |
||
| 1274 | "__________________________________________________________________________________________________\n", |
||
| 1275 | "block4c_bn (BatchNormalization) (None, 16, 16, 480) 1920 block4c_dwconv[0][0] \n", |
||
| 1276 | "__________________________________________________________________________________________________\n", |
||
| 1277 | "block4c_activation (Activation) (None, 16, 16, 480) 0 block4c_bn[0][0] \n", |
||
| 1278 | "__________________________________________________________________________________________________\n", |
||
| 1279 | "block4c_se_squeeze (GlobalAvera (None, 480) 0 block4c_activation[0][0] \n", |
||
| 1280 | "__________________________________________________________________________________________________\n", |
||
| 1281 | "block4c_se_reshape (Reshape) (None, 1, 1, 480) 0 block4c_se_squeeze[0][0] \n", |
||
| 1282 | "__________________________________________________________________________________________________\n", |
||
| 1283 | "block4c_se_reduce (Conv2D) (None, 1, 1, 20) 9620 block4c_se_reshape[0][0] \n", |
||
| 1284 | "__________________________________________________________________________________________________\n", |
||
| 1285 | "block4c_se_expand (Conv2D) (None, 1, 1, 480) 10080 block4c_se_reduce[0][0] \n", |
||
| 1286 | "__________________________________________________________________________________________________\n", |
||
| 1287 | "block4c_se_excite (Multiply) (None, 16, 16, 480) 0 block4c_activation[0][0] \n", |
||
| 1288 | " block4c_se_expand[0][0] \n", |
||
| 1289 | "__________________________________________________________________________________________________\n", |
||
| 1290 | "block4c_project_conv (Conv2D) (None, 16, 16, 80) 38400 block4c_se_excite[0][0] \n", |
||
| 1291 | "__________________________________________________________________________________________________\n", |
||
| 1292 | "block4c_project_bn (BatchNormal (None, 16, 16, 80) 320 block4c_project_conv[0][0] \n", |
||
| 1293 | "__________________________________________________________________________________________________\n", |
||
| 1294 | "block4c_drop (FixedDropout) (None, 16, 16, 80) 0 block4c_project_bn[0][0] \n", |
||
| 1295 | "__________________________________________________________________________________________________\n", |
||
| 1296 | "block4c_add (Add) (None, 16, 16, 80) 0 block4c_drop[0][0] \n", |
||
| 1297 | " block4b_add[0][0] \n", |
||
| 1298 | "__________________________________________________________________________________________________\n", |
||
| 1299 | "block5a_expand_conv (Conv2D) (None, 16, 16, 480) 38400 block4c_add[0][0] \n", |
||
| 1300 | "__________________________________________________________________________________________________\n", |
||
| 1301 | "block5a_expand_bn (BatchNormali (None, 16, 16, 480) 1920 block5a_expand_conv[0][0] \n", |
||
| 1302 | "__________________________________________________________________________________________________\n", |
||
| 1303 | "block5a_expand_activation (Acti (None, 16, 16, 480) 0 block5a_expand_bn[0][0] \n", |
||
| 1304 | "__________________________________________________________________________________________________\n", |
||
| 1305 | "block5a_dwconv (DepthwiseConv2D (None, 16, 16, 480) 12000 block5a_expand_activation[0][0] \n", |
||
| 1306 | "__________________________________________________________________________________________________\n", |
||
| 1307 | "block5a_bn (BatchNormalization) (None, 16, 16, 480) 1920 block5a_dwconv[0][0] \n", |
||
| 1308 | "__________________________________________________________________________________________________\n", |
||
| 1309 | "block5a_activation (Activation) (None, 16, 16, 480) 0 block5a_bn[0][0] \n", |
||
| 1310 | "__________________________________________________________________________________________________\n", |
||
| 1311 | "block5a_se_squeeze (GlobalAvera (None, 480) 0 block5a_activation[0][0] \n", |
||
| 1312 | "__________________________________________________________________________________________________\n", |
||
| 1313 | "block5a_se_reshape (Reshape) (None, 1, 1, 480) 0 block5a_se_squeeze[0][0] \n", |
||
| 1314 | "__________________________________________________________________________________________________\n", |
||
| 1315 | "block5a_se_reduce (Conv2D) (None, 1, 1, 20) 9620 block5a_se_reshape[0][0] \n", |
||
| 1316 | "__________________________________________________________________________________________________\n", |
||
| 1317 | "block5a_se_expand (Conv2D) (None, 1, 1, 480) 10080 block5a_se_reduce[0][0] \n", |
||
| 1318 | "__________________________________________________________________________________________________\n", |
||
| 1319 | "block5a_se_excite (Multiply) (None, 16, 16, 480) 0 block5a_activation[0][0] \n", |
||
| 1320 | " block5a_se_expand[0][0] \n", |
||
| 1321 | "__________________________________________________________________________________________________\n", |
||
| 1322 | "block5a_project_conv (Conv2D) (None, 16, 16, 112) 53760 block5a_se_excite[0][0] \n", |
||
| 1323 | "__________________________________________________________________________________________________\n", |
||
| 1324 | "block5a_project_bn (BatchNormal (None, 16, 16, 112) 448 block5a_project_conv[0][0] \n", |
||
| 1325 | "__________________________________________________________________________________________________\n", |
||
| 1326 | "block5b_expand_conv (Conv2D) (None, 16, 16, 672) 75264 block5a_project_bn[0][0] \n", |
||
| 1327 | "__________________________________________________________________________________________________\n", |
||
| 1328 | "block5b_expand_bn (BatchNormali (None, 16, 16, 672) 2688 block5b_expand_conv[0][0] \n", |
||
| 1329 | "__________________________________________________________________________________________________\n", |
||
| 1330 | "block5b_expand_activation (Acti (None, 16, 16, 672) 0 block5b_expand_bn[0][0] \n", |
||
| 1331 | "__________________________________________________________________________________________________\n", |
||
| 1332 | "block5b_dwconv (DepthwiseConv2D (None, 16, 16, 672) 16800 block5b_expand_activation[0][0] \n", |
||
| 1333 | "__________________________________________________________________________________________________\n", |
||
| 1334 | "block5b_bn (BatchNormalization) (None, 16, 16, 672) 2688 block5b_dwconv[0][0] \n", |
||
| 1335 | "__________________________________________________________________________________________________\n", |
||
| 1336 | "block5b_activation (Activation) (None, 16, 16, 672) 0 block5b_bn[0][0] \n", |
||
| 1337 | "__________________________________________________________________________________________________\n", |
||
| 1338 | "block5b_se_squeeze (GlobalAvera (None, 672) 0 block5b_activation[0][0] \n", |
||
| 1339 | "__________________________________________________________________________________________________\n", |
||
| 1340 | "block5b_se_reshape (Reshape) (None, 1, 1, 672) 0 block5b_se_squeeze[0][0] \n", |
||
| 1341 | "__________________________________________________________________________________________________\n", |
||
| 1342 | "block5b_se_reduce (Conv2D) (None, 1, 1, 28) 18844 block5b_se_reshape[0][0] \n", |
||
| 1343 | "__________________________________________________________________________________________________\n", |
||
| 1344 | "block5b_se_expand (Conv2D) (None, 1, 1, 672) 19488 block5b_se_reduce[0][0] \n", |
||
| 1345 | "__________________________________________________________________________________________________\n", |
||
| 1346 | "block5b_se_excite (Multiply) (None, 16, 16, 672) 0 block5b_activation[0][0] \n", |
||
| 1347 | " block5b_se_expand[0][0] \n", |
||
| 1348 | "__________________________________________________________________________________________________\n", |
||
| 1349 | "block5b_project_conv (Conv2D) (None, 16, 16, 112) 75264 block5b_se_excite[0][0] \n", |
||
| 1350 | "__________________________________________________________________________________________________\n", |
||
| 1351 | "block5b_project_bn (BatchNormal (None, 16, 16, 112) 448 block5b_project_conv[0][0] \n", |
||
| 1352 | "__________________________________________________________________________________________________\n", |
||
| 1353 | "block5b_drop (FixedDropout) (None, 16, 16, 112) 0 block5b_project_bn[0][0] \n", |
||
| 1354 | "__________________________________________________________________________________________________\n", |
||
| 1355 | "block5b_add (Add) (None, 16, 16, 112) 0 block5b_drop[0][0] \n", |
||
| 1356 | " block5a_project_bn[0][0] \n", |
||
| 1357 | "__________________________________________________________________________________________________\n", |
||
| 1358 | "block5c_expand_conv (Conv2D) (None, 16, 16, 672) 75264 block5b_add[0][0] \n", |
||
| 1359 | "__________________________________________________________________________________________________\n", |
||
| 1360 | "block5c_expand_bn (BatchNormali (None, 16, 16, 672) 2688 block5c_expand_conv[0][0] \n", |
||
| 1361 | "__________________________________________________________________________________________________\n", |
||
| 1362 | "block5c_expand_activation (Acti (None, 16, 16, 672) 0 block5c_expand_bn[0][0] \n", |
||
| 1363 | "__________________________________________________________________________________________________\n", |
||
| 1364 | "block5c_dwconv (DepthwiseConv2D (None, 16, 16, 672) 16800 block5c_expand_activation[0][0] \n", |
||
| 1365 | "__________________________________________________________________________________________________\n", |
||
| 1366 | "block5c_bn (BatchNormalization) (None, 16, 16, 672) 2688 block5c_dwconv[0][0] \n", |
||
| 1367 | "__________________________________________________________________________________________________\n", |
||
| 1368 | "block5c_activation (Activation) (None, 16, 16, 672) 0 block5c_bn[0][0] \n", |
||
| 1369 | "__________________________________________________________________________________________________\n", |
||
| 1370 | "block5c_se_squeeze (GlobalAvera (None, 672) 0 block5c_activation[0][0] \n", |
||
| 1371 | "__________________________________________________________________________________________________\n", |
||
| 1372 | "block5c_se_reshape (Reshape) (None, 1, 1, 672) 0 block5c_se_squeeze[0][0] \n", |
||
| 1373 | "__________________________________________________________________________________________________\n", |
||
| 1374 | "block5c_se_reduce (Conv2D) (None, 1, 1, 28) 18844 block5c_se_reshape[0][0] \n", |
||
| 1375 | "__________________________________________________________________________________________________\n", |
||
| 1376 | "block5c_se_expand (Conv2D) (None, 1, 1, 672) 19488 block5c_se_reduce[0][0] \n", |
||
| 1377 | "__________________________________________________________________________________________________\n", |
||
| 1378 | "block5c_se_excite (Multiply) (None, 16, 16, 672) 0 block5c_activation[0][0] \n", |
||
| 1379 | " block5c_se_expand[0][0] \n", |
||
| 1380 | "__________________________________________________________________________________________________\n", |
||
| 1381 | "block5c_project_conv (Conv2D) (None, 16, 16, 112) 75264 block5c_se_excite[0][0] \n", |
||
| 1382 | "__________________________________________________________________________________________________\n", |
||
| 1383 | "block5c_project_bn (BatchNormal (None, 16, 16, 112) 448 block5c_project_conv[0][0] \n", |
||
| 1384 | "__________________________________________________________________________________________________\n", |
||
| 1385 | "block5c_drop (FixedDropout) (None, 16, 16, 112) 0 block5c_project_bn[0][0] \n", |
||
| 1386 | "__________________________________________________________________________________________________\n", |
||
| 1387 | "block5c_add (Add) (None, 16, 16, 112) 0 block5c_drop[0][0] \n", |
||
| 1388 | " block5b_add[0][0] \n", |
||
| 1389 | "__________________________________________________________________________________________________\n", |
||
| 1390 | "block6a_expand_conv (Conv2D) (None, 16, 16, 672) 75264 block5c_add[0][0] \n", |
||
| 1391 | "__________________________________________________________________________________________________\n", |
||
| 1392 | "block6a_expand_bn (BatchNormali (None, 16, 16, 672) 2688 block6a_expand_conv[0][0] \n", |
||
| 1393 | "__________________________________________________________________________________________________\n", |
||
| 1394 | "block6a_expand_activation (Acti (None, 16, 16, 672) 0 block6a_expand_bn[0][0] \n", |
||
| 1395 | "__________________________________________________________________________________________________\n", |
||
| 1396 | "block6a_dwconv (DepthwiseConv2D (None, 8, 8, 672) 16800 block6a_expand_activation[0][0] \n", |
||
| 1397 | "__________________________________________________________________________________________________\n", |
||
| 1398 | "block6a_bn (BatchNormalization) (None, 8, 8, 672) 2688 block6a_dwconv[0][0] \n", |
||
| 1399 | "__________________________________________________________________________________________________\n", |
||
| 1400 | "block6a_activation (Activation) (None, 8, 8, 672) 0 block6a_bn[0][0] \n", |
||
| 1401 | "__________________________________________________________________________________________________\n", |
||
| 1402 | "block6a_se_squeeze (GlobalAvera (None, 672) 0 block6a_activation[0][0] \n", |
||
| 1403 | "__________________________________________________________________________________________________\n", |
||
| 1404 | "block6a_se_reshape (Reshape) (None, 1, 1, 672) 0 block6a_se_squeeze[0][0] \n", |
||
| 1405 | "__________________________________________________________________________________________________\n", |
||
| 1406 | "block6a_se_reduce (Conv2D) (None, 1, 1, 28) 18844 block6a_se_reshape[0][0] \n", |
||
| 1407 | "__________________________________________________________________________________________________\n", |
||
| 1408 | "block6a_se_expand (Conv2D) (None, 1, 1, 672) 19488 block6a_se_reduce[0][0] \n", |
||
| 1409 | "__________________________________________________________________________________________________\n", |
||
| 1410 | "block6a_se_excite (Multiply) (None, 8, 8, 672) 0 block6a_activation[0][0] \n", |
||
| 1411 | " block6a_se_expand[0][0] \n", |
||
| 1412 | "__________________________________________________________________________________________________\n", |
||
| 1413 | "block6a_project_conv (Conv2D) (None, 8, 8, 192) 129024 block6a_se_excite[0][0] \n", |
||
| 1414 | "__________________________________________________________________________________________________\n", |
||
| 1415 | "block6a_project_bn (BatchNormal (None, 8, 8, 192) 768 block6a_project_conv[0][0] \n", |
||
| 1416 | "__________________________________________________________________________________________________\n", |
||
| 1417 | "block6b_expand_conv (Conv2D) (None, 8, 8, 1152) 221184 block6a_project_bn[0][0] \n", |
||
| 1418 | "__________________________________________________________________________________________________\n", |
||
| 1419 | "block6b_expand_bn (BatchNormali (None, 8, 8, 1152) 4608 block6b_expand_conv[0][0] \n", |
||
| 1420 | "__________________________________________________________________________________________________\n", |
||
| 1421 | "block6b_expand_activation (Acti (None, 8, 8, 1152) 0 block6b_expand_bn[0][0] \n", |
||
| 1422 | "__________________________________________________________________________________________________\n", |
||
| 1423 | "block6b_dwconv (DepthwiseConv2D (None, 8, 8, 1152) 28800 block6b_expand_activation[0][0] \n", |
||
| 1424 | "__________________________________________________________________________________________________\n", |
||
| 1425 | "block6b_bn (BatchNormalization) (None, 8, 8, 1152) 4608 block6b_dwconv[0][0] \n", |
||
| 1426 | "__________________________________________________________________________________________________\n", |
||
| 1427 | "block6b_activation (Activation) (None, 8, 8, 1152) 0 block6b_bn[0][0] \n", |
||
| 1428 | "__________________________________________________________________________________________________\n", |
||
| 1429 | "block6b_se_squeeze (GlobalAvera (None, 1152) 0 block6b_activation[0][0] \n", |
||
| 1430 | "__________________________________________________________________________________________________\n", |
||
| 1431 | "block6b_se_reshape (Reshape) (None, 1, 1, 1152) 0 block6b_se_squeeze[0][0] \n", |
||
| 1432 | "__________________________________________________________________________________________________\n", |
||
| 1433 | "block6b_se_reduce (Conv2D) (None, 1, 1, 48) 55344 block6b_se_reshape[0][0] \n", |
||
| 1434 | "__________________________________________________________________________________________________\n", |
||
| 1435 | "block6b_se_expand (Conv2D) (None, 1, 1, 1152) 56448 block6b_se_reduce[0][0] \n", |
||
| 1436 | "__________________________________________________________________________________________________\n", |
||
| 1437 | "block6b_se_excite (Multiply) (None, 8, 8, 1152) 0 block6b_activation[0][0] \n", |
||
| 1438 | " block6b_se_expand[0][0] \n", |
||
| 1439 | "__________________________________________________________________________________________________\n", |
||
| 1440 | "block6b_project_conv (Conv2D) (None, 8, 8, 192) 221184 block6b_se_excite[0][0] \n", |
||
| 1441 | "__________________________________________________________________________________________________\n", |
||
| 1442 | "block6b_project_bn (BatchNormal (None, 8, 8, 192) 768 block6b_project_conv[0][0] \n", |
||
| 1443 | "__________________________________________________________________________________________________\n", |
||
| 1444 | "block6b_drop (FixedDropout) (None, 8, 8, 192) 0 block6b_project_bn[0][0] \n", |
||
| 1445 | "__________________________________________________________________________________________________\n", |
||
| 1446 | "block6b_add (Add) (None, 8, 8, 192) 0 block6b_drop[0][0] \n", |
||
| 1447 | " block6a_project_bn[0][0] \n", |
||
| 1448 | "__________________________________________________________________________________________________\n", |
||
| 1449 | "block6c_expand_conv (Conv2D) (None, 8, 8, 1152) 221184 block6b_add[0][0] \n", |
||
| 1450 | "__________________________________________________________________________________________________\n", |
||
| 1451 | "block6c_expand_bn (BatchNormali (None, 8, 8, 1152) 4608 block6c_expand_conv[0][0] \n", |
||
| 1452 | "__________________________________________________________________________________________________\n", |
||
| 1453 | "block6c_expand_activation (Acti (None, 8, 8, 1152) 0 block6c_expand_bn[0][0] \n", |
||
| 1454 | "__________________________________________________________________________________________________\n", |
||
| 1455 | "block6c_dwconv (DepthwiseConv2D (None, 8, 8, 1152) 28800 block6c_expand_activation[0][0] \n", |
||
| 1456 | "__________________________________________________________________________________________________\n", |
||
| 1457 | "block6c_bn (BatchNormalization) (None, 8, 8, 1152) 4608 block6c_dwconv[0][0] \n", |
||
| 1458 | "__________________________________________________________________________________________________\n", |
||
| 1459 | "block6c_activation (Activation) (None, 8, 8, 1152) 0 block6c_bn[0][0] \n", |
||
| 1460 | "__________________________________________________________________________________________________\n", |
||
| 1461 | "block6c_se_squeeze (GlobalAvera (None, 1152) 0 block6c_activation[0][0] \n", |
||
| 1462 | "__________________________________________________________________________________________________\n", |
||
| 1463 | "block6c_se_reshape (Reshape) (None, 1, 1, 1152) 0 block6c_se_squeeze[0][0] \n", |
||
| 1464 | "__________________________________________________________________________________________________\n", |
||
| 1465 | "block6c_se_reduce (Conv2D) (None, 1, 1, 48) 55344 block6c_se_reshape[0][0] \n", |
||
| 1466 | "__________________________________________________________________________________________________\n", |
||
| 1467 | "block6c_se_expand (Conv2D) (None, 1, 1, 1152) 56448 block6c_se_reduce[0][0] \n", |
||
| 1468 | "__________________________________________________________________________________________________\n", |
||
| 1469 | "block6c_se_excite (Multiply) (None, 8, 8, 1152) 0 block6c_activation[0][0] \n", |
||
| 1470 | " block6c_se_expand[0][0] \n", |
||
| 1471 | "__________________________________________________________________________________________________\n", |
||
| 1472 | "block6c_project_conv (Conv2D) (None, 8, 8, 192) 221184 block6c_se_excite[0][0] \n", |
||
| 1473 | "__________________________________________________________________________________________________\n", |
||
| 1474 | "block6c_project_bn (BatchNormal (None, 8, 8, 192) 768 block6c_project_conv[0][0] \n", |
||
| 1475 | "__________________________________________________________________________________________________\n", |
||
| 1476 | "block6c_drop (FixedDropout) (None, 8, 8, 192) 0 block6c_project_bn[0][0] \n", |
||
| 1477 | "__________________________________________________________________________________________________\n", |
||
| 1478 | "block6c_add (Add) (None, 8, 8, 192) 0 block6c_drop[0][0] \n", |
||
| 1479 | " block6b_add[0][0] \n", |
||
| 1480 | "__________________________________________________________________________________________________\n", |
||
| 1481 | "block6d_expand_conv (Conv2D) (None, 8, 8, 1152) 221184 block6c_add[0][0] \n", |
||
| 1482 | "__________________________________________________________________________________________________\n", |
||
| 1483 | "block6d_expand_bn (BatchNormali (None, 8, 8, 1152) 4608 block6d_expand_conv[0][0] \n", |
||
| 1484 | "__________________________________________________________________________________________________\n", |
||
| 1485 | "block6d_expand_activation (Acti (None, 8, 8, 1152) 0 block6d_expand_bn[0][0] \n", |
||
| 1486 | "__________________________________________________________________________________________________\n", |
||
| 1487 | "block6d_dwconv (DepthwiseConv2D (None, 8, 8, 1152) 28800 block6d_expand_activation[0][0] \n", |
||
| 1488 | "__________________________________________________________________________________________________\n", |
||
| 1489 | "block6d_bn (BatchNormalization) (None, 8, 8, 1152) 4608 block6d_dwconv[0][0] \n", |
||
| 1490 | "__________________________________________________________________________________________________\n", |
||
| 1491 | "block6d_activation (Activation) (None, 8, 8, 1152) 0 block6d_bn[0][0] \n", |
||
| 1492 | "__________________________________________________________________________________________________\n", |
||
| 1493 | "block6d_se_squeeze (GlobalAvera (None, 1152) 0 block6d_activation[0][0] \n", |
||
| 1494 | "__________________________________________________________________________________________________\n", |
||
| 1495 | "block6d_se_reshape (Reshape) (None, 1, 1, 1152) 0 block6d_se_squeeze[0][0] \n", |
||
| 1496 | "__________________________________________________________________________________________________\n", |
||
| 1497 | "block6d_se_reduce (Conv2D) (None, 1, 1, 48) 55344 block6d_se_reshape[0][0] \n", |
||
| 1498 | "__________________________________________________________________________________________________\n", |
||
| 1499 | "block6d_se_expand (Conv2D) (None, 1, 1, 1152) 56448 block6d_se_reduce[0][0] \n", |
||
| 1500 | "__________________________________________________________________________________________________\n", |
||
| 1501 | "block6d_se_excite (Multiply) (None, 8, 8, 1152) 0 block6d_activation[0][0] \n", |
||
| 1502 | " block6d_se_expand[0][0] \n", |
||
| 1503 | "__________________________________________________________________________________________________\n", |
||
| 1504 | "block6d_project_conv (Conv2D) (None, 8, 8, 192) 221184 block6d_se_excite[0][0] \n", |
||
| 1505 | "__________________________________________________________________________________________________\n", |
||
| 1506 | "block6d_project_bn (BatchNormal (None, 8, 8, 192) 768 block6d_project_conv[0][0] \n", |
||
| 1507 | "__________________________________________________________________________________________________\n", |
||
| 1508 | "block6d_drop (FixedDropout) (None, 8, 8, 192) 0 block6d_project_bn[0][0] \n", |
||
| 1509 | "__________________________________________________________________________________________________\n", |
||
| 1510 | "block6d_add (Add) (None, 8, 8, 192) 0 block6d_drop[0][0] \n", |
||
| 1511 | " block6c_add[0][0] \n", |
||
| 1512 | "__________________________________________________________________________________________________\n", |
||
| 1513 | "block7a_expand_conv (Conv2D) (None, 8, 8, 1152) 221184 block6d_add[0][0] \n", |
||
| 1514 | "__________________________________________________________________________________________________\n", |
||
| 1515 | "block7a_expand_bn (BatchNormali (None, 8, 8, 1152) 4608 block7a_expand_conv[0][0] \n", |
||
| 1516 | "__________________________________________________________________________________________________\n", |
||
| 1517 | "block7a_expand_activation (Acti (None, 8, 8, 1152) 0 block7a_expand_bn[0][0] \n", |
||
| 1518 | "__________________________________________________________________________________________________\n", |
||
| 1519 | "block7a_dwconv (DepthwiseConv2D (None, 8, 8, 1152) 10368 block7a_expand_activation[0][0] \n", |
||
| 1520 | "__________________________________________________________________________________________________\n", |
||
| 1521 | "block7a_bn (BatchNormalization) (None, 8, 8, 1152) 4608 block7a_dwconv[0][0] \n", |
||
| 1522 | "__________________________________________________________________________________________________\n", |
||
| 1523 | "block7a_activation (Activation) (None, 8, 8, 1152) 0 block7a_bn[0][0] \n", |
||
| 1524 | "__________________________________________________________________________________________________\n", |
||
| 1525 | "block7a_se_squeeze (GlobalAvera (None, 1152) 0 block7a_activation[0][0] \n", |
||
| 1526 | "__________________________________________________________________________________________________\n", |
||
| 1527 | "block7a_se_reshape (Reshape) (None, 1, 1, 1152) 0 block7a_se_squeeze[0][0] \n", |
||
| 1528 | "__________________________________________________________________________________________________\n", |
||
| 1529 | "block7a_se_reduce (Conv2D) (None, 1, 1, 48) 55344 block7a_se_reshape[0][0] \n", |
||
| 1530 | "__________________________________________________________________________________________________\n", |
||
| 1531 | "block7a_se_expand (Conv2D) (None, 1, 1, 1152) 56448 block7a_se_reduce[0][0] \n", |
||
| 1532 | "__________________________________________________________________________________________________\n", |
||
| 1533 | "block7a_se_excite (Multiply) (None, 8, 8, 1152) 0 block7a_activation[0][0] \n", |
||
| 1534 | " block7a_se_expand[0][0] \n", |
||
| 1535 | "__________________________________________________________________________________________________\n", |
||
| 1536 | "block7a_project_conv (Conv2D) (None, 8, 8, 320) 368640 block7a_se_excite[0][0] \n", |
||
| 1537 | "__________________________________________________________________________________________________\n", |
||
| 1538 | "block7a_project_bn (BatchNormal (None, 8, 8, 320) 1280 block7a_project_conv[0][0] \n", |
||
| 1539 | "__________________________________________________________________________________________________\n", |
||
| 1540 | "top_conv (Conv2D) (None, 8, 8, 1280) 409600 block7a_project_bn[0][0] \n", |
||
| 1541 | "__________________________________________________________________________________________________\n", |
||
| 1542 | "top_bn (BatchNormalization) (None, 8, 8, 1280) 5120 top_conv[0][0] \n", |
||
| 1543 | "__________________________________________________________________________________________________\n", |
||
| 1544 | "top_activation (Activation) (None, 8, 8, 1280) 0 top_bn[0][0] \n", |
||
| 1545 | "__________________________________________________________________________________________________\n", |
||
| 1546 | "avg_pool (GlobalAveragePooling2 (None, 1280) 0 top_activation[0][0] \n", |
||
| 1547 | "__________________________________________________________________________________________________\n", |
||
| 1548 | "dropout_1 (Dropout) (None, 1280) 0 avg_pool[0][0] \n", |
||
| 1549 | "__________________________________________________________________________________________________\n", |
||
| 1550 | "dense_1 (Dense) (None, 6) 7686 dropout_1[0][0] \n", |
||
| 1551 | "==================================================================================================\n", |
||
| 1552 | "Total params: 4,057,250\n", |
||
| 1553 | "Trainable params: 4,015,234\n", |
||
| 1554 | "Non-trainable params: 42,016\n", |
||
| 1555 | "__________________________________________________________________________________________________\n" |
||
| 1556 | ] |
||
| 1557 | } |
||
| 1558 | ], |
||
| 1559 | "source": [ |
||
| 1560 | "base_model = efn.EfficientNetB0(weights = 'imagenet', include_top = False, \\\n", |
||
| 1561 | " pooling = 'avg', input_shape = (HEIGHT, WIDTH, 3))\n", |
||
| 1562 | "x = base_model.output\n", |
||
| 1563 | "x = Dropout(0.125)(x)\n", |
||
| 1564 | "output_layer = Dense(6, activation = 'sigmoid')(x)\n", |
||
| 1565 | "model = Model(inputs=base_model.input, outputs=output_layer)\n", |
||
| 1566 | "model.compile(optimizer = Adam(learning_rate = 0.0001), \n", |
||
| 1567 | " loss = 'binary_crossentropy',\n", |
||
| 1568 | " metrics = ['acc', tf.keras.metrics.AUC()])\n", |
||
| 1569 | "model.summary()" |
||
| 1570 | ] |
||
| 1571 | }, |
||
| 1572 | { |
||
| 1573 | "cell_type": "code", |
||
| 1574 | "execution_count": 25, |
||
| 1575 | "metadata": {}, |
||
| 1576 | "outputs": [ |
||
| 1577 | { |
||
| 1578 | "data": { |
||
| 1579 | "text/plain": [ |
||
| 1580 | "(636396, 40622)" |
||
| 1581 | ] |
||
| 1582 | }, |
||
| 1583 | "execution_count": 25, |
||
| 1584 | "metadata": {}, |
||
| 1585 | "output_type": "execute_result" |
||
| 1586 | } |
||
| 1587 | ], |
||
| 1588 | "source": [ |
||
| 1589 | "# https://github.com/trent-b/iterative-stratification\n", |
||
| 1590 | "# Mutlilabel stratification\n", |
||
| 1591 | "splits = MultilabelStratifiedShuffleSplit(n_splits = 2, test_size = TEST_SIZE, random_state = SEED)\n", |
||
| 1592 | "file_names = train_final_df.index\n", |
||
| 1593 | "labels = train_final_df.values\n", |
||
| 1594 | "# Lets take only the first split\n", |
||
| 1595 | "split = next(splits.split(file_names, labels))\n", |
||
| 1596 | "train_idx = split[0]\n", |
||
| 1597 | "valid_idx = split[1]\n", |
||
| 1598 | "submission_predictions = []\n", |
||
| 1599 | "len(train_idx), len(valid_idx)" |
||
| 1600 | ] |
||
| 1601 | }, |
||
| 1602 | { |
||
| 1603 | "cell_type": "code", |
||
| 1604 | "execution_count": 26, |
||
| 1605 | "metadata": {}, |
||
| 1606 | "outputs": [], |
||
| 1607 | "source": [ |
||
| 1608 | "# train data generator\n", |
||
| 1609 | "data_generator_train = TrainDataGenerator(train_final_df.iloc[train_idx], \n", |
||
| 1610 | " train_final_df.iloc[train_idx], \n", |
||
| 1611 | " TRAIN_BATCH_SIZE, \n", |
||
| 1612 | " (WIDTH, HEIGHT),\n", |
||
| 1613 | " augment = True)\n", |
||
| 1614 | "\n", |
||
| 1615 | "# validation data generator\n", |
||
| 1616 | "data_generator_val = TrainDataGenerator(train_final_df.iloc[valid_idx], \n", |
||
| 1617 | " train_final_df.iloc[valid_idx], \n", |
||
| 1618 | " VALID_BATCH_SIZE, \n", |
||
| 1619 | " (WIDTH, HEIGHT),\n", |
||
| 1620 | " augment = False)" |
||
| 1621 | ] |
||
| 1622 | }, |
||
| 1623 | { |
||
| 1624 | "cell_type": "code", |
||
| 1625 | "execution_count": 27, |
||
| 1626 | "metadata": {}, |
||
| 1627 | "outputs": [ |
||
| 1628 | { |
||
| 1629 | "data": { |
||
| 1630 | "text/plain": [ |
||
| 1631 | "(19888, 635)" |
||
| 1632 | ] |
||
| 1633 | }, |
||
| 1634 | "execution_count": 27, |
||
| 1635 | "metadata": {}, |
||
| 1636 | "output_type": "execute_result" |
||
| 1637 | } |
||
| 1638 | ], |
||
| 1639 | "source": [ |
||
| 1640 | "len(data_generator_train), len(data_generator_val)" |
||
| 1641 | ] |
||
| 1642 | }, |
||
| 1643 | { |
||
| 1644 | "cell_type": "markdown", |
||
| 1645 | "metadata": {}, |
||
| 1646 | "source": [ |
||
| 1647 | "Competition evaluation metric is evaluated based on weighted log loss but we haven't given weights for each subtype but as per discussion from this thread https://www.kaggle.com/c/rsna-intracranial-hemorrhage-detection/discussion/109526#latest-630190 any has a wieght of 2 than other types below sample is taken from the discussion threas" |
||
| 1648 | ] |
||
| 1649 | }, |
||
| 1650 | { |
||
| 1651 | "cell_type": "code", |
||
| 1652 | "execution_count": 28, |
||
| 1653 | "metadata": {}, |
||
| 1654 | "outputs": [], |
||
| 1655 | "source": [ |
||
| 1656 | "from keras import backend as K\n", |
||
| 1657 | "\n", |
||
| 1658 | "def weighted_log_loss(y_true, y_pred):\n", |
||
| 1659 | " \"\"\"\n", |
||
| 1660 | " Can be used as the loss function in model.compile()\n", |
||
| 1661 | " ---------------------------------------------------\n", |
||
| 1662 | " \"\"\"\n", |
||
| 1663 | " \n", |
||
| 1664 | " class_weights = np.array([2., 1., 1., 1., 1., 1.])\n", |
||
| 1665 | " \n", |
||
| 1666 | " eps = K.epsilon()\n", |
||
| 1667 | " \n", |
||
| 1668 | " y_pred = K.clip(y_pred, eps, 1.0-eps)\n", |
||
| 1669 | "\n", |
||
| 1670 | " out = -( y_true * K.log( y_pred) * class_weights\n", |
||
| 1671 | " + (1.0 - y_true) * K.log(1.0 - y_pred) * class_weights)\n", |
||
| 1672 | " \n", |
||
| 1673 | " return K.mean(out, axis=-1)\n", |
||
| 1674 | "\n", |
||
| 1675 | "\n", |
||
| 1676 | "def _normalized_weighted_average(arr, weights=None):\n", |
||
| 1677 | " \"\"\"\n", |
||
| 1678 | " A simple Keras implementation that mimics that of \n", |
||
| 1679 | " numpy.average(), specifically for this competition\n", |
||
| 1680 | " \"\"\"\n", |
||
| 1681 | " \n", |
||
| 1682 | " if weights is not None:\n", |
||
| 1683 | " scl = K.sum(weights)\n", |
||
| 1684 | " weights = K.expand_dims(weights, axis=1)\n", |
||
| 1685 | " return K.sum(K.dot(arr, weights), axis=1) / scl\n", |
||
| 1686 | " return K.mean(arr, axis=1)\n", |
||
| 1687 | "\n", |
||
| 1688 | "\n", |
||
| 1689 | "def weighted_loss(y_true, y_pred):\n", |
||
| 1690 | " \"\"\"\n", |
||
| 1691 | " Will be used as the metric in model.compile()\n", |
||
| 1692 | " ---------------------------------------------\n", |
||
| 1693 | " \n", |
||
| 1694 | " Similar to the custom loss function 'weighted_log_loss()' above\n", |
||
| 1695 | " but with normalized weights, which should be very similar \n", |
||
| 1696 | " to the official competition metric:\n", |
||
| 1697 | " https://www.kaggle.com/kambarakun/lb-probe-weights-n-of-positives-scoring\n", |
||
| 1698 | " and hence:\n", |
||
| 1699 | " sklearn.metrics.log_loss with sample weights\n", |
||
| 1700 | " \"\"\"\n", |
||
| 1701 | " \n", |
||
| 1702 | " class_weights = K.variable([2., 1., 1., 1., 1., 1.])\n", |
||
| 1703 | " \n", |
||
| 1704 | " eps = K.epsilon()\n", |
||
| 1705 | " \n", |
||
| 1706 | " y_pred = K.clip(y_pred, eps, 1.0-eps)\n", |
||
| 1707 | "\n", |
||
| 1708 | " loss = -( y_true * K.log( y_pred)\n", |
||
| 1709 | " + (1.0 - y_true) * K.log(1.0 - y_pred))\n", |
||
| 1710 | " \n", |
||
| 1711 | " loss_samples = _normalized_weighted_average(loss, class_weights)\n", |
||
| 1712 | " \n", |
||
| 1713 | " return K.mean(loss_samples)\n", |
||
| 1714 | "\n", |
||
| 1715 | "\n", |
||
| 1716 | "def weighted_log_loss_metric(trues, preds):\n", |
||
| 1717 | " \"\"\"\n", |
||
| 1718 | " Will be used to calculate the log loss \n", |
||
| 1719 | " of the validation set in PredictionCheckpoint()\n", |
||
| 1720 | " ------------------------------------------\n", |
||
| 1721 | " \"\"\"\n", |
||
| 1722 | " class_weights = [2., 1., 1., 1., 1., 1.]\n", |
||
| 1723 | " \n", |
||
| 1724 | " epsilon = 1e-7\n", |
||
| 1725 | " \n", |
||
| 1726 | " preds = np.clip(preds, epsilon, 1-epsilon)\n", |
||
| 1727 | " loss = trues * np.log(preds) + (1 - trues) * np.log(1 - preds)\n", |
||
| 1728 | " loss_samples = np.average(loss, axis=1, weights=class_weights)\n", |
||
| 1729 | "\n", |
||
| 1730 | " return - loss_samples.mean()" |
||
| 1731 | ] |
||
| 1732 | }, |
||
| 1733 | { |
||
| 1734 | "cell_type": "code", |
||
| 1735 | "execution_count": 29, |
||
| 1736 | "metadata": {}, |
||
| 1737 | "outputs": [], |
||
| 1738 | "source": [ |
||
| 1739 | "filepath=\"model.h5\"\n", |
||
| 1740 | "checkpoint = ModelCheckpoint(filepath, monitor='val_loss', verbose=1, \\\n", |
||
| 1741 | " save_best_only=True, mode='min')\n", |
||
| 1742 | "\n", |
||
| 1743 | "callbacks_list = [checkpoint]" |
||
| 1744 | ] |
||
| 1745 | }, |
||
| 1746 | { |
||
| 1747 | "cell_type": "markdown", |
||
| 1748 | "metadata": {}, |
||
| 1749 | "source": [ |
||
| 1750 | "For a single epoch we are going to train only last 5 layers of Efficient. Since we have a large number of images around 600k so its better to train the all the layers on the whole train dataset but due its high computation resources required to train we only goin to train last five layers on whole dataset and for rest of epochs we only train on a sample of dataset but will train all the layers." |
||
| 1751 | ] |
||
| 1752 | }, |
||
| 1753 | { |
||
| 1754 | "cell_type": "code", |
||
| 1755 | "execution_count": 30, |
||
| 1756 | "metadata": {}, |
||
| 1757 | "outputs": [ |
||
| 1758 | { |
||
| 1759 | "data": { |
||
| 1760 | "text/plain": [ |
||
| 1761 | "False" |
||
| 1762 | ] |
||
| 1763 | }, |
||
| 1764 | "execution_count": 30, |
||
| 1765 | "metadata": {}, |
||
| 1766 | "output_type": "execute_result" |
||
| 1767 | } |
||
| 1768 | ], |
||
| 1769 | "source": [ |
||
| 1770 | "os.path.isfile('../input/orginal-087-eff/model.h5')" |
||
| 1771 | ] |
||
| 1772 | }, |
||
| 1773 | { |
||
| 1774 | "cell_type": "code", |
||
| 1775 | "execution_count": 32, |
||
| 1776 | "metadata": {}, |
||
| 1777 | "outputs": [], |
||
| 1778 | "source": [ |
||
| 1779 | "train=False\n", |
||
| 1780 | "\n", |
||
| 1781 | "if train:\n", |
||
| 1782 | " if not os.path.isfile('../input/orginal-087-eff/model.h5'):\n", |
||
| 1783 | " for layer in model.layers[:-5]:\n", |
||
| 1784 | " layer.trainable = False\n", |
||
| 1785 | " model.compile(optimizer = Adam(learning_rate = 0.0001), \n", |
||
| 1786 | " loss = 'binary_crossentropy',\n", |
||
| 1787 | " metrics = ['acc'])\n", |
||
| 1788 | "\n", |
||
| 1789 | " model.fit_generator(generator = data_generator_train,\n", |
||
| 1790 | " validation_data = data_generator_val,\n", |
||
| 1791 | " epochs = 1,\n", |
||
| 1792 | " callbacks = callbacks_list,\n", |
||
| 1793 | " verbose = 1)" |
||
| 1794 | ] |
||
| 1795 | }, |
||
| 1796 | { |
||
| 1797 | "cell_type": "code", |
||
| 1798 | "execution_count": 33, |
||
| 1799 | "metadata": {}, |
||
| 1800 | "outputs": [], |
||
| 1801 | "source": [ |
||
| 1802 | "if train:\n", |
||
| 1803 | " for base_layer in model.layers[:-1]:\n", |
||
| 1804 | " base_layer.trainable = True\n", |
||
| 1805 | "\n", |
||
| 1806 | " model.load_weights('model.h5')\n", |
||
| 1807 | "\n", |
||
| 1808 | " model.compile(optimizer = Adam(learning_rate = 0.0004), \n", |
||
| 1809 | " loss = 'binary_crossentropy',\n", |
||
| 1810 | " metrics = ['acc'])\n", |
||
| 1811 | " model.fit_generator(generator = data_generator_train,\n", |
||
| 1812 | " validation_data = data_generator_val,\n", |
||
| 1813 | " steps_per_epoch=len(data_generator_train)/6,\n", |
||
| 1814 | " epochs = 10,\n", |
||
| 1815 | " callbacks = callbacks_list,\n", |
||
| 1816 | " verbose = 1)" |
||
| 1817 | ] |
||
| 1818 | }, |
||
| 1819 | { |
||
| 1820 | "cell_type": "code", |
||
| 1821 | "execution_count": 34, |
||
| 1822 | "metadata": {}, |
||
| 1823 | "outputs": [ |
||
| 1824 | { |
||
| 1825 | "name": "stdout", |
||
| 1826 | "output_type": "stream", |
||
| 1827 | "text": [ |
||
| 1828 | "Collecting gdown\n", |
||
| 1829 | " Downloading https://files.pythonhosted.org/packages/b0/b4/a8e9d0b02bca6aa53087001abf064cc9992bda11bd6840875b8098d93573/gdown-3.8.3.tar.gz\n", |
||
| 1830 | "Requirement already satisfied: filelock in /opt/conda/lib/python3.6/site-packages (from gdown) (3.0.12)\n", |
||
| 1831 | "Requirement already satisfied: requests in /opt/conda/lib/python3.6/site-packages (from gdown) (2.22.0)\n", |
||
| 1832 | "Requirement already satisfied: six in /opt/conda/lib/python3.6/site-packages (from gdown) (1.12.0)\n", |
||
| 1833 | "Requirement already satisfied: tqdm in /opt/conda/lib/python3.6/site-packages (from gdown) (4.36.1)\n", |
||
| 1834 | "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.6/site-packages (from requests->gdown) (2019.9.11)\n", |
||
| 1835 | "Requirement already satisfied: idna<2.9,>=2.5 in /opt/conda/lib/python3.6/site-packages (from requests->gdown) (2.8)\n", |
||
| 1836 | "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /opt/conda/lib/python3.6/site-packages (from requests->gdown) (1.24.2)\n", |
||
| 1837 | "Requirement already satisfied: chardet<3.1.0,>=3.0.2 in /opt/conda/lib/python3.6/site-packages (from requests->gdown) (3.0.4)\n", |
||
| 1838 | "Building wheels for collected packages: gdown\n", |
||
| 1839 | " Building wheel for gdown (setup.py) ... \u001b[?25ldone\n", |
||
| 1840 | "\u001b[?25h Created wheel for gdown: filename=gdown-3.8.3-cp36-none-any.whl size=8850 sha256=14a9f4491a378ff8119e1ca63a536cd63c0f8e914ac97133e046e690719cf217\n", |
||
| 1841 | " Stored in directory: /tmp/.cache/pip/wheels/a7/9d/16/9e0bda9a327ff2cddaee8de48a27553fb1efce73133593d066\n", |
||
| 1842 | "Successfully built gdown\n", |
||
| 1843 | "Installing collected packages: gdown\n", |
||
| 1844 | "Successfully installed gdown-3.8.3\n" |
||
| 1845 | ] |
||
| 1846 | } |
||
| 1847 | ], |
||
| 1848 | "source": [ |
||
| 1849 | "!pip install gdown" |
||
| 1850 | ] |
||
| 1851 | }, |
||
| 1852 | { |
||
| 1853 | "cell_type": "code", |
||
| 1854 | "execution_count": 35, |
||
| 1855 | "metadata": {}, |
||
| 1856 | "outputs": [ |
||
| 1857 | { |
||
| 1858 | "name": "stdout", |
||
| 1859 | "output_type": "stream", |
||
| 1860 | "text": [ |
||
| 1861 | "Downloading...\n", |
||
| 1862 | "From: https://drive.google.com/uc?id=1OMWQjtnVkMKLQ3jG4RpUQR2IyimO-W46\n", |
||
| 1863 | "To: /kaggle/working/model (2).h5\n", |
||
| 1864 | "49.2MB [00:00, 135MB/s] \n" |
||
| 1865 | ] |
||
| 1866 | } |
||
| 1867 | ], |
||
| 1868 | "source": [ |
||
| 1869 | "!gdown \"https://drive.google.com/uc?id=1OMWQjtnVkMKLQ3jG4RpUQR2IyimO-W46\"" |
||
| 1870 | ] |
||
| 1871 | }, |
||
| 1872 | { |
||
| 1873 | "cell_type": "code", |
||
| 1874 | "execution_count": 36, |
||
| 1875 | "metadata": {}, |
||
| 1876 | "outputs": [], |
||
| 1877 | "source": [ |
||
| 1878 | "!cp \"model (2).h5\" model.h5" |
||
| 1879 | ] |
||
| 1880 | }, |
||
| 1881 | { |
||
| 1882 | "cell_type": "code", |
||
| 1883 | "execution_count": 37, |
||
| 1884 | "metadata": {}, |
||
| 1885 | "outputs": [ |
||
| 1886 | { |
||
| 1887 | "name": "stdout", |
||
| 1888 | "output_type": "stream", |
||
| 1889 | "text": [ |
||
| 1890 | "1228/1228 [==============================] - 1048s 854ms/step\n" |
||
| 1891 | ] |
||
| 1892 | }, |
||
| 1893 | { |
||
| 1894 | "data": { |
||
| 1895 | "text/plain": [ |
||
| 1896 | "(78592, 6)" |
||
| 1897 | ] |
||
| 1898 | }, |
||
| 1899 | "execution_count": 37, |
||
| 1900 | "metadata": {}, |
||
| 1901 | "output_type": "execute_result" |
||
| 1902 | } |
||
| 1903 | ], |
||
| 1904 | "source": [ |
||
| 1905 | "model.load_weights('model.h5')\n", |
||
| 1906 | "\n", |
||
| 1907 | "preds = model.predict_generator(TestDataGenerator(test_df.index, None, VALID_BATCH_SIZE, \\\n", |
||
| 1908 | " (WIDTH, HEIGHT), path_test_img), \n", |
||
| 1909 | " verbose=1)\n", |
||
| 1910 | "preds.shape" |
||
| 1911 | ] |
||
| 1912 | }, |
||
| 1913 | { |
||
| 1914 | "cell_type": "code", |
||
| 1915 | "execution_count": 38, |
||
| 1916 | "metadata": {}, |
||
| 1917 | "outputs": [], |
||
| 1918 | "source": [ |
||
| 1919 | "from tqdm import tqdm" |
||
| 1920 | ] |
||
| 1921 | }, |
||
| 1922 | { |
||
| 1923 | "cell_type": "code", |
||
| 1924 | "execution_count": 39, |
||
| 1925 | "metadata": {}, |
||
| 1926 | "outputs": [], |
||
| 1927 | "source": [ |
||
| 1928 | "cols = list(train_final_df.columns)" |
||
| 1929 | ] |
||
| 1930 | }, |
||
| 1931 | { |
||
| 1932 | "cell_type": "code", |
||
| 1933 | "execution_count": 40, |
||
| 1934 | "metadata": {}, |
||
| 1935 | "outputs": [ |
||
| 1936 | { |
||
| 1937 | "name": "stderr", |
||
| 1938 | "output_type": "stream", |
||
| 1939 | "text": [ |
||
| 1940 | "100%|█████████▉| 78545/78592 [00:01<00:00, 48430.17it/s]\n" |
||
| 1941 | ] |
||
| 1942 | } |
||
| 1943 | ], |
||
| 1944 | "source": [ |
||
| 1945 | "# We have preditions for each of the image\n", |
||
| 1946 | "# We need to make 6 rows for each of file according to the subtype\n", |
||
| 1947 | "ids = []\n", |
||
| 1948 | "values = []\n", |
||
| 1949 | "for i, j in tqdm(zip(preds, test_df.index.to_list()), total=preds.shape[0]):\n", |
||
| 1950 | "# print(i, j)\n", |
||
| 1951 | " # i=[any_prob, epidural_prob, intraparenchymal_prob, intraventricular_prob, subarachnoid_prob, subdural_prob]\n", |
||
| 1952 | " # j = filename ==> ID_xyz.dcm\n", |
||
| 1953 | " for k in range(i.shape[0]):\n", |
||
| 1954 | " ids.append([j.replace('.dcm', '_' + cols[k])])\n", |
||
| 1955 | " values.append(i[k]) " |
||
| 1956 | ] |
||
| 1957 | }, |
||
| 1958 | { |
||
| 1959 | "cell_type": "code", |
||
| 1960 | "execution_count": 41, |
||
| 1961 | "metadata": {}, |
||
| 1962 | "outputs": [ |
||
| 1963 | { |
||
| 1964 | "data": { |
||
| 1965 | "text/html": [ |
||
| 1966 | "<div>\n", |
||
| 1967 | "<style scoped>\n", |
||
| 1968 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 1969 | " vertical-align: middle;\n", |
||
| 1970 | " }\n", |
||
| 1971 | "\n", |
||
| 1972 | " .dataframe tbody tr th {\n", |
||
| 1973 | " vertical-align: top;\n", |
||
| 1974 | " }\n", |
||
| 1975 | "\n", |
||
| 1976 | " .dataframe thead th {\n", |
||
| 1977 | " text-align: right;\n", |
||
| 1978 | " }\n", |
||
| 1979 | "</style>\n", |
||
| 1980 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 1981 | " <thead>\n", |
||
| 1982 | " <tr style=\"text-align: right;\">\n", |
||
| 1983 | " <th></th>\n", |
||
| 1984 | " <th>0</th>\n", |
||
| 1985 | " </tr>\n", |
||
| 1986 | " </thead>\n", |
||
| 1987 | " <tbody>\n", |
||
| 1988 | " <tr>\n", |
||
| 1989 | " <th>0</th>\n", |
||
| 1990 | " <td>ID_000012eaf_any</td>\n", |
||
| 1991 | " </tr>\n", |
||
| 1992 | " <tr>\n", |
||
| 1993 | " <th>1</th>\n", |
||
| 1994 | " <td>ID_000012eaf_epidural</td>\n", |
||
| 1995 | " </tr>\n", |
||
| 1996 | " <tr>\n", |
||
| 1997 | " <th>2</th>\n", |
||
| 1998 | " <td>ID_000012eaf_intraparenchymal</td>\n", |
||
| 1999 | " </tr>\n", |
||
| 2000 | " <tr>\n", |
||
| 2001 | " <th>3</th>\n", |
||
| 2002 | " <td>ID_000012eaf_intraventricular</td>\n", |
||
| 2003 | " </tr>\n", |
||
| 2004 | " <tr>\n", |
||
| 2005 | " <th>4</th>\n", |
||
| 2006 | " <td>ID_000012eaf_subarachnoid</td>\n", |
||
| 2007 | " </tr>\n", |
||
| 2008 | " </tbody>\n", |
||
| 2009 | "</table>\n", |
||
| 2010 | "</div>" |
||
| 2011 | ], |
||
| 2012 | "text/plain": [ |
||
| 2013 | " 0\n", |
||
| 2014 | "0 ID_000012eaf_any\n", |
||
| 2015 | "1 ID_000012eaf_epidural\n", |
||
| 2016 | "2 ID_000012eaf_intraparenchymal\n", |
||
| 2017 | "3 ID_000012eaf_intraventricular\n", |
||
| 2018 | "4 ID_000012eaf_subarachnoid" |
||
| 2019 | ] |
||
| 2020 | }, |
||
| 2021 | "execution_count": 41, |
||
| 2022 | "metadata": {}, |
||
| 2023 | "output_type": "execute_result" |
||
| 2024 | } |
||
| 2025 | ], |
||
| 2026 | "source": [ |
||
| 2027 | "df = pd.DataFrame(data=ids)\n", |
||
| 2028 | "df.head()" |
||
| 2029 | ] |
||
| 2030 | }, |
||
| 2031 | { |
||
| 2032 | "cell_type": "code", |
||
| 2033 | "execution_count": 42, |
||
| 2034 | "metadata": {}, |
||
| 2035 | "outputs": [ |
||
| 2036 | { |
||
| 2037 | "data": { |
||
| 2038 | "text/html": [ |
||
| 2039 | "<div>\n", |
||
| 2040 | "<style scoped>\n", |
||
| 2041 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 2042 | " vertical-align: middle;\n", |
||
| 2043 | " }\n", |
||
| 2044 | "\n", |
||
| 2045 | " .dataframe tbody tr th {\n", |
||
| 2046 | " vertical-align: top;\n", |
||
| 2047 | " }\n", |
||
| 2048 | "\n", |
||
| 2049 | " .dataframe thead th {\n", |
||
| 2050 | " text-align: right;\n", |
||
| 2051 | " }\n", |
||
| 2052 | "</style>\n", |
||
| 2053 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 2054 | " <thead>\n", |
||
| 2055 | " <tr style=\"text-align: right;\">\n", |
||
| 2056 | " <th></th>\n", |
||
| 2057 | " <th>ID</th>\n", |
||
| 2058 | " <th>Label</th>\n", |
||
| 2059 | " </tr>\n", |
||
| 2060 | " </thead>\n", |
||
| 2061 | " <tbody>\n", |
||
| 2062 | " <tr>\n", |
||
| 2063 | " <th>0</th>\n", |
||
| 2064 | " <td>ID_28fbab7eb_epidural</td>\n", |
||
| 2065 | " <td>0.5</td>\n", |
||
| 2066 | " </tr>\n", |
||
| 2067 | " <tr>\n", |
||
| 2068 | " <th>1</th>\n", |
||
| 2069 | " <td>ID_28fbab7eb_intraparenchymal</td>\n", |
||
| 2070 | " <td>0.5</td>\n", |
||
| 2071 | " </tr>\n", |
||
| 2072 | " <tr>\n", |
||
| 2073 | " <th>2</th>\n", |
||
| 2074 | " <td>ID_28fbab7eb_intraventricular</td>\n", |
||
| 2075 | " <td>0.5</td>\n", |
||
| 2076 | " </tr>\n", |
||
| 2077 | " <tr>\n", |
||
| 2078 | " <th>3</th>\n", |
||
| 2079 | " <td>ID_28fbab7eb_subarachnoid</td>\n", |
||
| 2080 | " <td>0.5</td>\n", |
||
| 2081 | " </tr>\n", |
||
| 2082 | " <tr>\n", |
||
| 2083 | " <th>4</th>\n", |
||
| 2084 | " <td>ID_28fbab7eb_subdural</td>\n", |
||
| 2085 | " <td>0.5</td>\n", |
||
| 2086 | " </tr>\n", |
||
| 2087 | " </tbody>\n", |
||
| 2088 | "</table>\n", |
||
| 2089 | "</div>" |
||
| 2090 | ], |
||
| 2091 | "text/plain": [ |
||
| 2092 | " ID Label\n", |
||
| 2093 | "0 ID_28fbab7eb_epidural 0.5\n", |
||
| 2094 | "1 ID_28fbab7eb_intraparenchymal 0.5\n", |
||
| 2095 | "2 ID_28fbab7eb_intraventricular 0.5\n", |
||
| 2096 | "3 ID_28fbab7eb_subarachnoid 0.5\n", |
||
| 2097 | "4 ID_28fbab7eb_subdural 0.5" |
||
| 2098 | ] |
||
| 2099 | }, |
||
| 2100 | "execution_count": 42, |
||
| 2101 | "metadata": {}, |
||
| 2102 | "output_type": "execute_result" |
||
| 2103 | } |
||
| 2104 | ], |
||
| 2105 | "source": [ |
||
| 2106 | "sample_df = pd.read_csv(input_folder + 'stage_1_sample_submission.csv')\n", |
||
| 2107 | "sample_df.head()" |
||
| 2108 | ] |
||
| 2109 | }, |
||
| 2110 | { |
||
| 2111 | "cell_type": "code", |
||
| 2112 | "execution_count": 43, |
||
| 2113 | "metadata": {}, |
||
| 2114 | "outputs": [ |
||
| 2115 | { |
||
| 2116 | "data": { |
||
| 2117 | "text/html": [ |
||
| 2118 | "<div>\n", |
||
| 2119 | "<style scoped>\n", |
||
| 2120 | " .dataframe tbody tr th:only-of-type {\n", |
||
| 2121 | " vertical-align: middle;\n", |
||
| 2122 | " }\n", |
||
| 2123 | "\n", |
||
| 2124 | " .dataframe tbody tr th {\n", |
||
| 2125 | " vertical-align: top;\n", |
||
| 2126 | " }\n", |
||
| 2127 | "\n", |
||
| 2128 | " .dataframe thead th {\n", |
||
| 2129 | " text-align: right;\n", |
||
| 2130 | " }\n", |
||
| 2131 | "</style>\n", |
||
| 2132 | "<table border=\"1\" class=\"dataframe\">\n", |
||
| 2133 | " <thead>\n", |
||
| 2134 | " <tr style=\"text-align: right;\">\n", |
||
| 2135 | " <th></th>\n", |
||
| 2136 | " <th>ID</th>\n", |
||
| 2137 | " <th>Label</th>\n", |
||
| 2138 | " </tr>\n", |
||
| 2139 | " </thead>\n", |
||
| 2140 | " <tbody>\n", |
||
| 2141 | " <tr>\n", |
||
| 2142 | " <th>0</th>\n", |
||
| 2143 | " <td>ID_000012eaf_any</td>\n", |
||
| 2144 | " <td>0.029101</td>\n", |
||
| 2145 | " </tr>\n", |
||
| 2146 | " <tr>\n", |
||
| 2147 | " <th>1</th>\n", |
||
| 2148 | " <td>ID_000012eaf_epidural</td>\n", |
||
| 2149 | " <td>0.001475</td>\n", |
||
| 2150 | " </tr>\n", |
||
| 2151 | " <tr>\n", |
||
| 2152 | " <th>2</th>\n", |
||
| 2153 | " <td>ID_000012eaf_intraparenchymal</td>\n", |
||
| 2154 | " <td>0.001740</td>\n", |
||
| 2155 | " </tr>\n", |
||
| 2156 | " <tr>\n", |
||
| 2157 | " <th>3</th>\n", |
||
| 2158 | " <td>ID_000012eaf_intraventricular</td>\n", |
||
| 2159 | " <td>0.001194</td>\n", |
||
| 2160 | " </tr>\n", |
||
| 2161 | " <tr>\n", |
||
| 2162 | " <th>4</th>\n", |
||
| 2163 | " <td>ID_000012eaf_subarachnoid</td>\n", |
||
| 2164 | " <td>0.001531</td>\n", |
||
| 2165 | " </tr>\n", |
||
| 2166 | " </tbody>\n", |
||
| 2167 | "</table>\n", |
||
| 2168 | "</div>" |
||
| 2169 | ], |
||
| 2170 | "text/plain": [ |
||
| 2171 | " ID Label\n", |
||
| 2172 | "0 ID_000012eaf_any 0.029101\n", |
||
| 2173 | "1 ID_000012eaf_epidural 0.001475\n", |
||
| 2174 | "2 ID_000012eaf_intraparenchymal 0.001740\n", |
||
| 2175 | "3 ID_000012eaf_intraventricular 0.001194\n", |
||
| 2176 | "4 ID_000012eaf_subarachnoid 0.001531" |
||
| 2177 | ] |
||
| 2178 | }, |
||
| 2179 | "execution_count": 43, |
||
| 2180 | "metadata": {}, |
||
| 2181 | "output_type": "execute_result" |
||
| 2182 | } |
||
| 2183 | ], |
||
| 2184 | "source": [ |
||
| 2185 | "df['Label'] = values\n", |
||
| 2186 | "df.columns = sample_df.columns\n", |
||
| 2187 | "df.head()" |
||
| 2188 | ] |
||
| 2189 | }, |
||
| 2190 | { |
||
| 2191 | "cell_type": "code", |
||
| 2192 | "execution_count": 44, |
||
| 2193 | "metadata": {}, |
||
| 2194 | "outputs": [], |
||
| 2195 | "source": [ |
||
| 2196 | "df.to_csv('submission.csv', index=False)" |
||
| 2197 | ] |
||
| 2198 | }, |
||
| 2199 | { |
||
| 2200 | "cell_type": "code", |
||
| 2201 | "execution_count": 45, |
||
| 2202 | "metadata": {}, |
||
| 2203 | "outputs": [ |
||
| 2204 | { |
||
| 2205 | "data": { |
||
| 2206 | "text/html": [ |
||
| 2207 | "<a href=submission.csv>Download CSV file</a>" |
||
| 2208 | ], |
||
| 2209 | "text/plain": [ |
||
| 2210 | "<IPython.core.display.HTML object>" |
||
| 2211 | ] |
||
| 2212 | }, |
||
| 2213 | "execution_count": 45, |
||
| 2214 | "metadata": {}, |
||
| 2215 | "output_type": "execute_result" |
||
| 2216 | } |
||
| 2217 | ], |
||
| 2218 | "source": [ |
||
| 2219 | "create_download_link(filename='submission.csv')" |
||
| 2220 | ] |
||
| 2221 | } |
||
| 2222 | ], |
||
| 2223 | "metadata": { |
||
| 2224 | "kernelspec": { |
||
| 2225 | "display_name": "Python 3", |
||
| 2226 | "language": "python", |
||
| 2227 | "name": "python3" |
||
| 2228 | }, |
||
| 2229 | "language_info": { |
||
| 2230 | "codemirror_mode": { |
||
| 2231 | "name": "ipython", |
||
| 2232 | "version": 3 |
||
| 2233 | }, |
||
| 2234 | "file_extension": ".py", |
||
| 2235 | "mimetype": "text/x-python", |
||
| 2236 | "name": "python", |
||
| 2237 | "nbconvert_exporter": "python", |
||
| 2238 | "pygments_lexer": "ipython3", |
||
| 2239 | "version": "3.6.5" |
||
| 2240 | } |
||
| 2241 | }, |
||
| 2242 | "nbformat": 4, |
||
| 2243 | "nbformat_minor": 1 |
||
| 2244 | } |