File size: 10,865 Bytes
fad2e05 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 | {
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "kFEwutXUvYCs",
"outputId": "e883733e-f919-401a-e6d3-4edf7f0d012e"
},
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"Mounted at /content/drive\n"
]
}
],
"source": [
"# Accessing our google drive since the data is rather heavy\n",
"from google.colab import drive\n",
"drive.mount('/content/drive', force_remount=True)"
]
},
{
"cell_type": "markdown",
"source": [
"# Subsetting and resizing our images"
],
"metadata": {
"id": "uBoCyPh5B5P6"
}
},
{
"cell_type": "markdown",
"metadata": {
"id": "OUxBi_iSvYCu"
},
"source": [
"Our computer and the free version of Google Colab cannot handle taking care of all our data. To be able to do anything, we need to reduce the total amount of images and their size."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "wtG9723wvYCv"
},
"outputs": [],
"source": [
"!unzip /content/drive/MyDrive/Projet_Artefact_Memes/data.zip -d data"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "tZtMuK37vYCw"
},
"outputs": [],
"source": [
"# Importing packages\n",
"import os\n",
"import numpy as np\n",
"import pandas as pd\n",
"import cv2\n",
"from tqdm import tqdm"
]
},
{
"cell_type": "markdown",
"source": [
"We are going to reduce drastically the number of images: from 9000 usable images, we are going to keep only 5000. Therefore, only loading the content of *train.json* is more than enough for taking care of our images.\n",
"\n",
"Howerver, we are going to also load our *dev.json* file and concatenate them for later use."
],
"metadata": {
"id": "rHop57d3C7fJ"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "I3QOZrAWvYCw",
"outputId": "85764aed-b306-4ed0-b8a9-5bda9929e1ba"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"(8500, 4)"
]
},
"metadata": {},
"execution_count": 3
}
],
"source": [
"# We read our data\n",
"filepath = './drive/MyDrive/Projet_Artefact_Memes'\n",
"\n",
"with open(f'{filepath}/train.jsonl') as f:\n",
" df = pd.read_json(f, lines=True)\n",
"df.shape"
]
},
{
"cell_type": "code",
"source": [
"with open(f'{filepath}/dev.jsonl') as f:\n",
" df_dev = pd.read_json(f, lines=True)\n",
"df_dev.shape"
],
"metadata": {
"id": "_-_5bzGGUbJe"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# We concat our datasets\n",
"df_all = pd.concat([df, df_dev])\n",
"\n",
"# We save our dataframe as a csv file for later use\n",
"df_all.to_csv(f'{filepath}/data_all.csv', index=False)"
],
"metadata": {
"id": "UDBCk8j1Uqtx"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "markdown",
"source": [
"We decide on having a balanced dataset, so we keep a random subset of 2500 hateful memes and 2500 non-hateful memes."
],
"metadata": {
"id": "v5iZxRfODLrs"
}
},
{
"cell_type": "code",
"source": [
"df_subset_hate = df[df['label']==1]\n",
"df_subset_hate = df_subset_hate.sample(2500, random_state=24)\n",
"\n",
"df_subset_no_hate = df[df['label']==0]\n",
"df_subset_no_hate = df_subset_no_hate.sample(2500, random_state=24)"
],
"metadata": {
"id": "cWH0Zw__dG4h"
},
"execution_count": null,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# We concat the result to have a single dataframe\n",
"df = pd.concat([df_subset_hate, df_subset_no_hate])\n",
"df.shape"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "LW7y8Ky3dG7D",
"outputId": "3d60e251-e08d-415f-a59c-746de04350dd"
},
"execution_count": null,
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"(5000, 4)"
]
},
"metadata": {},
"execution_count": 6
}
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "tPAPf0O5vYCw",
"outputId": "7c308aed-c06a-4a16-e42d-034edea29752"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"1 2500\n",
"0 2500\n",
"Name: label, dtype: int64"
]
},
"metadata": {},
"execution_count": 7
}
],
"source": [
"# We check the balancing of our classes\n",
"df['label'].value_counts()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "p_irx94gvYCx"
},
"outputs": [],
"source": [
"# We split to only keep the image name\n",
"df['img'] = df['img'].apply(lambda x: x.split('/')[1])\n",
"\n",
"# We save our dataframe as a csv file for later use\n",
"df.to_csv(f'{filepath}/data_5K_balanced.csv', index=False)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "5c6rwhv1vYCx"
},
"outputs": [],
"source": [
"# We order our images just in case\n",
"# (not necessary in this case but better safe than sorry)\n",
"images_ordered = []\n",
"for item in os.listdir('./data/data/img'):\n",
" if item in df['img'].values:\n",
" images_ordered.append(item)\n",
"\n",
"images_ordered = sorted(images_ordered)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "20yYghIcvYCx",
"outputId": "b8f0eea4-10f0-4240-a427-1b39b391b4ed"
},
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"5000\n"
]
}
],
"source": [
"# We check that we have indeed 5000 images\n",
"print(len(images_ordered))"
]
},
{
"cell_type": "markdown",
"source": [
"Now, we can resize our images to a reduce shape (120 x 120)."
],
"metadata": {
"id": "vKJjg-BOE-ur"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "ndez6r98vYCy",
"outputId": "47eaf76d-9886-47bc-fe95-f6c8f19d57b5"
},
"outputs": [
{
"output_type": "stream",
"name": "stderr",
"text": [
"100%|██████████| 5000/5000 [01:11<00:00, 69.94it/s]\n"
]
}
],
"source": [
"images = []\n",
"for i in tqdm(range(5000)):\n",
" if images_ordered[i].endswith('.png'):\n",
"\n",
" # On lit l'image\n",
" img = cv2.imread(f'./data/data/img/{images_ordered[i]}')\n",
"\n",
" # On remet en RGB et on ne laisse pas en BGR\n",
" img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)\n",
"\n",
" # On resize nos images\n",
" resized_img = cv2.resize(img, (120, 120), interpolation = cv2.INTER_AREA)\n",
"\n",
" images.append(resized_img)"
]
},
{
"cell_type": "markdown",
"source": [
"We then save our list of images as a numpy array."
],
"metadata": {
"id": "76bsZNbrHRFR"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "7vl-XZKWvYCy",
"outputId": "e76ae26c-f0bf-4691-952c-18ca52731924"
},
"outputs": [
{
"output_type": "execute_result",
"data": {
"text/plain": [
"(5000, 120, 120, 3)"
]
},
"metadata": {},
"execution_count": 13
}
],
"source": [
"X = np.array(images)\n",
"X.shape"
]
},
{
"cell_type": "markdown",
"source": [
"Then, we scale our images."
],
"metadata": {
"id": "nGvpcCEfHVnh"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "Bb3CVTV-vYC0"
},
"outputs": [],
"source": [
"X = X / 255"
]
},
{
"cell_type": "markdown",
"source": [
"Finally, we save our array of scaled images in a .npz file."
],
"metadata": {
"id": "R-cXQclYHaiA"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "DHithg4566m-"
},
"outputs": [],
"source": [
"np.savez_compressed('./drive/MyDrive/Projet_Artefact_Memes/data_scaled_120_balanced.npz', a=X)"
]
},
{
"cell_type": "code",
"source": [],
"metadata": {
"id": "Q2yaXobsRLrj"
},
"execution_count": null,
"outputs": []
}
],
"metadata": {
"accelerator": "GPU",
"colab": {
"gpuType": "T4",
"provenance": []
},
"gpuClass": "standard",
"kernelspec": {
"display_name": "artefact",
"language": "python",
"name": "python3"
},
"language_info": {
"name": "python",
"version": "3.10.4"
},
"orig_nbformat": 4
},
"nbformat": 4,
"nbformat_minor": 0
} |