File size: 30,682 Bytes
4ba4ba4 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 | {
"cells": [
{
"cell_type": "markdown",
"id": "ab23242e-3c5e-4ba6-9cb9-dfc4d8c22321",
"metadata": {},
"source": [
"<img src=\"./figs/IOAI-Logo.png\" alt=\"IOAI Logo\" width=\"200\" height=\"auto\">\n",
"\n",
"[IOAI 2025 (Beijing, China), Individual Contest](https://ioai-official.org/china-2025)\n",
"\n",
"[](https://colab.research.google.com/github/IOAI-official/IOAI-2025/blob/main/Individual-Contest/Radar/Solution/Radar_Solution.ipynb)"
]
},
{
"cell_type": "markdown",
"id": "f2ea67c2-ec81-4ced-b8a4-a19b591376bb",
"metadata": {},
"source": [
"# Radar: Reference Solution"
]
},
{
"cell_type": "markdown",
"id": "97ae6685",
"metadata": {},
"source": [
"## Configs"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "4243ca5d",
"metadata": {
"jupyter": {
"source_hidden": false
}
},
"outputs": [],
"source": [
"radar_configs = {\n",
" 'c': 3e8, \n",
" 'startFreq': 77e9, \n",
" 'Tr': 60e-6, # Sweeping frequency time\n",
" 'Idle_time': 100e-6, # free time\n",
" 'Fs': 10e6, # Sampling frequency\n",
" 'Slope': 29.982e12, # chirp slope\n",
" 'Bandwidth': 60e-6 * 29.982e12, # Transmission signal bandwidth\n",
" 'BandwidthValid': 0.767539200e9,# Effective bandwidth of the transmitted signal\n",
" 'range_size': 256, # range size\n",
" 'azimuth_size': 181, # azimuth size\n",
" 'elevation_size': 181, # elevation size\n",
" 'crop_num': 3, # crop some indices in range domain\n",
" 'n_chirps': 128, # number of chirps in one frame\n",
" 'min_azimuth': -90, # min radar azimuth\n",
" 'max_azimuth': 90, # max radar azimuth\n",
" 'min_elevation': -90, # min radar elevation\n",
" 'max_elevation': 90, # max radar elevation \n",
" 'min_range': 1.0, # min radar range\n",
" 'max_range': 25.0, # max radar range\n",
" 'range_res': 3e8/(2*0.767539200e9), \n",
" 'angle_res': 1\n",
"}\n",
"\n",
"dimssnet_configs = {\n",
" 'n_epoch': 10,\n",
" 'batch_size': 2,\n",
" 'learning_rate': 1e-5,\n",
" 'range_size': 50,\n",
" 'azimuth_size': 181,\n",
" 'elevation_size': 181,\n",
" 'doppler_size': 181, \n",
" 'min_azimuth': -90, # min radar azimuth\n",
" 'max_azimuth': 90, # max radar azimuth\n",
" 'min_elevation': -90, # min radar elevation\n",
" 'max_elevation': 90, # max radar elevation \n",
" 'min_range': 1.0, # min radar range\n",
" 'max_range': 25.0, # max radar range\n",
"}\n",
"\n",
"n_class = 5\n",
"class_table = {\n",
" -1: 'background', \n",
" 0: 'suicase', \n",
" 1: 'chair', \n",
" 2: 'person',\n",
" 3: 'wall'\n",
"}"
]
},
{
"cell_type": "markdown",
"id": "c39a54f5",
"metadata": {},
"source": [
"## Models"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "10817ce8",
"metadata": {
"jupyter": {
"source_hidden": false
}
},
"outputs": [],
"source": [
"import torch\n",
"import torch.nn as nn\n",
"import torch.nn.functional as F\n",
"\n",
"class BaseEncoder(nn.Module):\n",
" def __init__(self):\n",
" super(BaseEncoder, self).__init__()\n",
" self.conv1a = nn.Conv2d(in_channels=1, out_channels=64, \n",
" kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" self.conv1b = nn.Conv2d(in_channels=64, out_channels=64, \n",
" kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" self.conv2a = nn.Conv2d(in_channels=64, out_channels=128, \n",
" kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" self.conv2b = nn.Conv2d(in_channels=128, out_channels=128, \n",
" kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" self.conv3a = nn.Conv2d(in_channels=128, out_channels=256, \n",
" kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" self.conv3b = nn.Conv2d(in_channels=256, out_channels=256, \n",
" kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
"\n",
" self.bn1a = nn.BatchNorm2d(num_features=64)\n",
" self.bn1b = nn.BatchNorm2d(num_features=64)\n",
" self.bn2a = nn.BatchNorm2d(num_features=128)\n",
" self.bn2b = nn.BatchNorm2d(num_features=128)\n",
" self.bn3a = nn.BatchNorm2d(num_features=256)\n",
" self.bn3b = nn.BatchNorm2d(num_features=256)\n",
"\n",
" self.relu = nn.ReLU()\n",
" \n",
" def forward(self, x):\n",
" x = self.relu(self.bn1a(self.conv1a(x)))\n",
" x = self.relu(self.bn1b(self.conv1b(x)))\n",
" x = self.relu(self.bn2a(self.conv2a(x)))\n",
" x = self.relu(self.bn2b(self.conv2b(x)))\n",
" x = self.relu(self.bn3a(self.conv3a(x)))\n",
" x = self.relu(self.bn3b(self.conv3b(x)))\n",
" \n",
" return x\n",
"\n",
"\n",
"class BaseDecoder(nn.Module):\n",
" def __init__(self):\n",
" super(BaseDecoder, self).__init__()\n",
" self.convt1 = nn.ConvTranspose2d(in_channels=256, out_channels=128,\n",
" kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" self.convt2 = nn.ConvTranspose2d(in_channels=128, out_channels=64,\n",
" kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" self.convt3 = nn.ConvTranspose2d(in_channels=64, out_channels=32,\n",
" kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
" self.prelu = nn.PReLU()\n",
"\n",
" def forward(self, x):\n",
" x = self.prelu(self.convt1(x))\n",
" x = self.prelu(self.convt2(x))\n",
" x = self.prelu(self.convt3(x))\n",
"\n",
" return x"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "ddb913e6",
"metadata": {
"jupyter": {
"source_hidden": false
}
},
"outputs": [],
"source": [
"range_size=radar_configs['range_size']\n",
"azimuth_size=radar_configs['azimuth_size']\n",
"elevation_size=radar_configs['elevation_size']\n",
"\n",
"class Fuse_fea(nn.Module):\n",
" def __init__(self):\n",
" super(Fuse_fea, self).__init__()\n",
" self.convt1 = nn.Conv2d(in_channels=192, out_channels=96,\n",
" kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" self.convt2 = nn.Conv2d(in_channels=96, out_channels=48,\n",
" kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" self.convt3 = nn.Conv2d(in_channels=48, out_channels=24,\n",
" kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
" self.convt4 = nn.Conv2d(in_channels=24, out_channels=n_class,\n",
" kernel_size=(3, 3), stride=(1, 1), padding=(1, 1)) \n",
" self.relu = nn.ReLU()\n",
" self.sigmoid = nn.Sigmoid()\n",
"\n",
" def forward(self, feas_sra, feas_dra, feas_sre, feas_dre, feas_srd, feas_drd):\n",
" feas_sre = torch.sum(feas_sre, 3, keepdim=True) # (B, 32, 50, 181) -> (B, 32, 50, 1)\n",
" feas_sre = feas_sre.expand(-1, -1, -1, azimuth_size)\n",
" \n",
" feas_dre = torch.sum(feas_dre, 3, keepdim=True) # (B, 32, 50, 181) -> (B, 32, 50, 1)\n",
" feas_dre = feas_dre.expand(-1, -1, -1, azimuth_size)\n",
"\n",
" feas_srd = torch.sum(feas_srd, 3, keepdim=True) # (B, 32, 50, 181) -> (B, 32, 50, 1)\n",
" feas_srd = feas_srd.expand(-1, -1, -1, azimuth_size)\n",
"\n",
" feas_drd = torch.sum(feas_drd, 3, keepdim=True) # (B, 32, 50, 181) -> (B, 32, 50, 1)\n",
" feas_drd = feas_drd.expand(-1, -1, -1, azimuth_size)\n",
"\n",
" # Resize using bilinear interpolation\n",
" feas_sra = F.interpolate(feas_sra, size=(range_size, azimuth_size), mode='bilinear', align_corners=False)\n",
" feas_dra = F.interpolate(feas_dra, size=(range_size, azimuth_size), mode='bilinear', align_corners=False)\n",
" feas_sre = F.interpolate(feas_sre, size=(range_size, azimuth_size), mode='bilinear', align_corners=False)\n",
" feas_dre = F.interpolate(feas_dre, size=(range_size, azimuth_size), mode='bilinear', align_corners=False)\n",
" feas_srd = F.interpolate(feas_srd, size=(range_size, azimuth_size), mode='bilinear', align_corners=False)\n",
" feas_drd = F.interpolate(feas_drd, size=(range_size, azimuth_size), mode='bilinear', align_corners=False)\n",
" \n",
" feas_ra = torch.cat((feas_sra, feas_dra, feas_sre, feas_dre, feas_srd, feas_drd), 1) # 6*(B, 32, 50, 181) -> (B, 192, 50, 181)\n",
"\n",
" x = self.relu(self.convt1(feas_ra)) # (B, 192, 50, 181) -> (B, 96, 50, 181)\n",
" x = self.relu(self.convt2(x)) # (B, 96, 50, 181) -> (B, 48, 50, 181)\n",
" x = self.relu(self.convt3(x)) # (B, 48, 50, 181) -> (B, 24, 50, 181) \n",
" x = self.sigmoid(self.convt4(x)) # (B, 24, 50, 181) -> (B, 5, 50, 181)\n",
" \n",
" return x\n",
" "
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "483a5ba6",
"metadata": {
"jupyter": {
"source_hidden": false
}
},
"outputs": [],
"source": [
"class DIMSSNet(nn.Module):\n",
" def __init__(self):\n",
" super(DIMSSNet, self).__init__()\n",
" self.encode_sra, self.decode_sra = BaseEncoder(), BaseDecoder()\n",
" self.encode_dra, self.decode_dra = BaseEncoder(), BaseDecoder()\n",
" self.encode_sre, self.decode_sre = BaseEncoder(), BaseDecoder()\n",
" self.encode_dre, self.decode_dre = BaseEncoder(), BaseDecoder()\n",
" self.encode_srd, self.decode_srd = BaseEncoder(), BaseDecoder()\n",
" self.encode_drd, self.decode_drd = BaseEncoder(), BaseDecoder()\n",
" \n",
" self.fuse_fea = Fuse_fea()\n",
" \n",
" def forward(self, x_sra, x_dra, x_sre, x_dre, x_srd, x_drd):\n",
" x_sra = self.encode_sra(x_sra)\n",
" feas_sra = self.decode_sra(x_sra) # (B, 32, 50, 181)\n",
" \n",
" x_dra = self.encode_dra(x_dra)\n",
" feas_dra = self.decode_dra(x_dra) # (B, 32, 50, 181)\n",
" \n",
" x_sre = self.encode_sre(x_sre)\n",
" feas_sre = self.decode_sre(x_sre) # (B, 32, 50, 181)\n",
" \n",
" x_dre = self.encode_dre(x_dre)\n",
" feas_dre = self.decode_dre(x_dre) # (B, 32, 50, 181)\n",
"\n",
" x_srd = self.encode_sre(x_srd)\n",
" feas_srd = self.decode_sre(x_srd) # (B, 32, 50, 181)\n",
" \n",
" x_drd = self.encode_dre(x_drd)\n",
" feas_drd = self.decode_dre(x_drd) # (B, 32, 50, 181) \n",
" \n",
" feas = self.fuse_fea(feas_sra, feas_dra, feas_sre, feas_dre, feas_srd, feas_drd) # (B, 32, 50, 181)\n",
"\n",
" return feas"
]
},
{
"cell_type": "markdown",
"id": "b8d86ece",
"metadata": {},
"source": [
"## Utils"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "49705a8f",
"metadata": {
"jupyter": {
"source_hidden": false
}
},
"outputs": [],
"source": [
"import scipy.io\n",
"import numpy as np\n",
"\n",
"# Define the target size\n",
"TARGET_HEIGHT = 50\n",
"TARGET_WIDTH = 181\n",
"\n",
"def load_raw_data(file_path):\n",
" data = scipy.io.loadmat(file_path)\n",
"\n",
" # Range-Azimuth Static\n",
" range_az_static = data['range_az_static']\n",
" range_az_static_prob = data['range_az_music_average_static_probability']\n",
" range_az_static_class2_prob = range_az_static_prob[:, :, 1]\n",
"\n",
" # Range-Elevation Static\n",
" range_ele_static = data['range_ele_static_all']\n",
"\n",
" # Range-Azimuth Dynamic\n",
" range_az_dynamic = data['range_az_dynamic']\n",
" range_az_dynamic_prob = data['range_az_music_average_dynamic_probability']\n",
" range_az_dynamic_class1_prob = range_az_dynamic_prob[:, :, 0]\n",
"\n",
" # Range-Elevation Dynamic\n",
" range_ele_dynamic = data['range_ele_dynamic_all'] \n",
"\n",
" return [range_az_static, range_az_dynamic, range_ele_static,\n",
" range_ele_dynamic, range_az_dynamic_class1_prob, range_az_static_class2_prob]\n",
"\n",
"def generate_parameter():\n",
" parameter = {\n",
" 'c': 3e8, # the speed of light\n",
" 'start_freq': 77e9, # starting frequency\n",
" 'tr': 60e-6, # Sweeping time, that is, the cycle\n",
" 'samples': 256, # sampling point\n",
" 'fs': 10e6, # Sampling rate\n",
" 'tframe_set': 80e-3, # Frame period\n",
" 'range_bin': 256, # range bin\n",
" 'chirps': 128, # chirp number\n",
" 'doppler_bin': 128, # doppler bin\n",
" 'slope': 29.982e12, # chirp slope\n",
" 'bandwidth': 29.982e12 * 60e-6, # Effective bandwidth of the transmitted signal\n",
" 'bandwidth_valid': 256 / 10e6 * 29.982e12, # Transmission signal bandwidth\n",
" 'center_freq': 77e9 + (29.982e12 * 60e-6) / 2, # center frequency\n",
" 'lambda': 3e8 / (77e9 + (29.982e12 * 60e-6) / 2), # wavelength\n",
" 'tx_antenna': [1, 1, 1], # The number of transmitting antennas\n",
" 'rx_antenna': [1, 1, 1, 1], # Number of receiving antennas\n",
" 'tx_num': 3, # The number of transmitting antennas\n",
" 'rx_num': 4, # The number of receiving antennas\n",
" 'virtual_antenna': 12, # Number of virtual antennas\n",
" 'dz': (3e8 / (77e9 + (29.982e12 * 60e-6) / 2)) / 2, # Pitch spacing of the receiving antenna\n",
" 'dx': (3e8 / (77e9 + (29.982e12 * 60e-6) / 2)) / 2, # Horizontal spacing of receiving antennas\n",
" 'num_cpi': 50, # frames\n",
" }\n",
" return parameter\n",
"\n"
]
},
{
"cell_type": "markdown",
"id": "c266f24b",
"metadata": {},
"source": [
"## Dataloaders"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "892cfb3a",
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"from torch.utils.data import Dataset, DataLoader\n",
"from sklearn.model_selection import train_test_split\n",
"\n",
"class CustomDataset(Dataset):\n",
" def __init__(self, file_paths, transform=None):\n",
" self.file_paths = file_paths\n",
" self.transform = transform\n",
" self.file_names = [os.path.basename(path) for path in file_paths]\n",
"\n",
" def __len__(self):\n",
" return len(self.file_paths)\n",
"\n",
" def __getitem__(self, idx):\n",
" data = torch.load(self.file_paths[idx], weights_only=True)\n",
" \n",
" images = data[:6] #first 6 channels are the heatmaps\n",
" labels = data[6] #last channel is the label\n",
" \n",
" images = images.float() \n",
" labels = labels.long() \n",
" labels = labels + 1 #labels are -1 to 3, add 1 to make them 0 to 4\n",
"\n",
" if self.transform:\n",
" images = self.transform(images)\n",
" labels = self.transform(labels)\n",
" \n",
" return images, labels, self.file_names[idx]\n",
"\n",
"class CustomDataset_test(Dataset):\n",
" def __init__(self, file_paths, transform=None):\n",
" self.file_paths = file_paths\n",
" self.transform = transform\n",
" self.file_names = [os.path.basename(path) for path in file_paths]\n",
"\n",
" def __len__(self):\n",
" return len(self.file_paths)\n",
"\n",
" def __getitem__(self, idx):\n",
" data = torch.load(self.file_paths[idx], weights_only=True)\n",
" \n",
" images = data[:6] #only load first 6 channels, labels are not in val_set and test_set\n",
" \n",
" images = images.float() \n",
"\n",
" if self.transform:\n",
" images = self.transform(images)\n",
" \n",
" return images, self.file_names[idx]\n",
"\n",
"# extend base_path to file_path\n",
"def generate_file_paths(base_path):\n",
" file_paths = []\n",
" for frame in os.listdir(base_path):\n",
" frame_path = os.path.join(base_path, frame)\n",
" if frame_path.endswith('.mat.pt'):\n",
" file_paths.append(frame_path)\n",
" return [path for path in file_paths if os.path.exists(path)]\n",
"\n",
"def load_data(base_path, batch_size=4, num_workers=2, test_size=0.2):\n",
" file_paths = generate_file_paths(base_path)\n",
" \n",
" train_paths, test_paths = train_test_split(file_paths, test_size=test_size, random_state=42)\n",
" \n",
" train_dataset = CustomDataset(file_paths=train_paths)\n",
" test_dataset = CustomDataset(file_paths=test_paths)\n",
" \n",
" train_loader = DataLoader(\n",
" train_dataset, \n",
" batch_size=batch_size, \n",
" shuffle=True, \n",
" num_workers=num_workers, \n",
" drop_last=True\n",
" )\n",
" \n",
" test_loader = DataLoader(\n",
" test_dataset, \n",
" batch_size=batch_size, \n",
" shuffle=False, \n",
" num_workers=num_workers, \n",
" drop_last=True\n",
" )\n",
" \n",
" return train_loader, test_loader\n"
]
},
{
"cell_type": "markdown",
"id": "67cd6c15",
"metadata": {},
"source": [
"## Loss Function"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "4c573803",
"metadata": {},
"outputs": [],
"source": [
"#Used to handle the problem of class imbalance\n",
"\n",
"def to_one_hot(tensor, nClasses):\n",
" n, h, w = tensor.size()\n",
" one_hot = torch.zeros(n, nClasses, h, w, device=tensor.device)\n",
" index_tensor = tensor.long().view(n, 1, h, w)\n",
" one_hot = one_hot.scatter_(1, index_tensor, 1)\n",
" return one_hot\n",
"\n",
"def _neg_loss(pred, gt):\n",
" pred = torch.clamp(pred, 1e-5, 1-1e-5)\n",
"\n",
" pos_inds = gt.eq(1).float()\n",
" neg_inds = gt.eq(0).float()\n",
"\n",
" balance_cof = 1.0\n",
" pos_loss = torch.log(pred) * torch.pow(1 - pred, 2) * pos_inds * balance_cof\n",
" neg_loss = torch.log(1 - pred) * torch.pow(pred, 2) * neg_inds\n",
" num_pos = pos_inds.float().sum()\n",
" pos_loss_sum = pos_loss.sum()\n",
" neg_loss_sum = neg_loss.sum()\n",
" if num_pos == 0:\n",
" loss = -neg_loss_sum\n",
" else:\n",
" loss = -(pos_loss_sum + neg_loss_sum) / num_pos\n",
" return loss\n",
"\n",
"class FocalLoss(nn.Module):\n",
" def __init__(self):\n",
" super(FocalLoss, self).__init__()\n",
" self.neg_loss = _neg_loss\n",
"\n",
" def forward(self, out, target):\n",
" return self.neg_loss(out, target)"
]
},
{
"cell_type": "markdown",
"id": "dec2b582",
"metadata": {},
"source": [
"## Train"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "7c2f0301",
"metadata": {},
"outputs": [],
"source": [
"import torch.optim as optim\n",
"import matplotlib.pyplot as plt\n",
"import time\n",
"\n",
"device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
"\n",
"num_epochs = 20\n",
"\n",
"base_path = '../training_set'\n",
"train_loader, test_loader = load_data(\n",
" base_path=base_path,\n",
" batch_size=4,\n",
" num_workers=2,\n",
" test_size=0.2\n",
")\n",
"\n",
"dimssnet = DIMSSNet().to(device)\n",
"criterion = FocalLoss()\n",
"optimizer = optim.Adam(dimssnet.parameters(), lr=0.001)\n",
"start_time = time.time()\n",
"test_loss_record = []\n",
"train_loss_record = []\n",
"\n",
"for epoch in range(num_epochs):\n",
" dimssnet.train()\n",
" running_loss = 0.0\n",
" for images, labels, _ in train_loader:\n",
" images = images.to(device)\n",
" labels = labels.to(device)\n",
" \n",
" images = [image.unsqueeze(1) for image in images]\n",
" images = torch.stack(images, dim=0)\n",
" images = images.float()\n",
" \n",
" outputs = dimssnet(images[:, 0, :, :], images[:, 1, :, :], images[:, 2, :, :],\n",
" images[:, 3, :, :], images[:, 4, :, :], images[:, 5, :, :])\n",
"\n",
" outputs_resized = F.interpolate(outputs, size=(TARGET_HEIGHT, TARGET_WIDTH), mode='bilinear', align_corners=False)\n",
" labels = to_one_hot(labels, n_class)\n",
" \n",
" loss = criterion(outputs_resized, labels.long())\n",
" optimizer.zero_grad()\n",
" loss.backward()\n",
" optimizer.step()\n",
" \n",
" running_loss += loss.item()\n",
"\n",
" test_loss = 0.0\n",
" dimssnet.eval()\n",
" with torch.no_grad():\n",
" for images, labels, _ in test_loader:\n",
" images = images.to(device)\n",
" labels = labels.to(device)\n",
" \n",
" images = [image.unsqueeze(1) for image in images]\n",
" images = torch.stack(images, dim=0)\n",
" images = images.float()\n",
" \n",
" outputs = dimssnet(images[:, 0, :, :], images[:, 1, :, :], images[:, 2, :, :],\n",
" images[:, 3, :, :], images[:, 4, :, :], images[:, 5, :, :])\n",
" \n",
" outputs_resized = F.interpolate(outputs, size=(TARGET_HEIGHT, TARGET_WIDTH), mode='bilinear', align_corners=False)\n",
" labels = to_one_hot(labels, n_class)\n",
" \n",
" loss = criterion(outputs_resized, labels.long())\n",
" test_loss += loss.item()\n",
" \n",
" test_loss_record.append(test_loss/len(test_loader))\n",
" train_loss_record.append(running_loss/len(train_loader))\n",
" print(f\"Epoch {epoch+1}/{num_epochs}, Training Loss: {running_loss/len(train_loader):.4f}, Test Loss: {test_loss/len(test_loader):.4f}\")\n",
"\n",
"# Create a new state_dict and add the \"model\" prefix, in order to match the loading method in metrics\n",
"new_state_dict = {}\n",
"for k, v in dimssnet.state_dict().items():\n",
" new_key = f\"model.{k}\" \n",
" new_state_dict[new_key] = v\n",
"\n",
"print(\"dimssnet successfully saved!\") \n",
"print(\"Training finished.\")\n",
"end_time = time.time()\n",
"total_time = end_time - start_time\n",
"print(f\"Training completed in {total_time} seconds\")"
]
},
{
"cell_type": "markdown",
"id": "f001aee8",
"metadata": {},
"source": [
"## Inference"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "8b330819",
"metadata": {},
"outputs": [],
"source": [
"# Run inference on validation set and testing set\n",
"import pandas as pd\n",
"\n",
"def run_inference(model, data_loader):\n",
" model.eval()\n",
" predictions = []\n",
" filenames = []\n",
" \n",
" with torch.no_grad():\n",
" for images, file_names in data_loader:\n",
" images = images.cuda() if torch.cuda.is_available() else images\n",
" \n",
" x = [image.unsqueeze(1) for image in images]\n",
" x = torch.stack(x, dim=0)\n",
" x = x.float()\n",
" \n",
" # Forward propagation\n",
" outputs = model(x[:, 0, :, :], x[:, 1, :, :], x[:, 2, :, :],\n",
" x[:, 3, :, :], x[:, 4, :, :], x[:, 5, :, :])\n",
" \n",
" # Adjust the output size\n",
" outputs = F.interpolate(outputs, size=(50, 181), mode='bilinear', align_corners=False)\n",
"\n",
" preds = torch.argmax(outputs, dim=1)\n",
" \n",
" preds = preds - 1\n",
" \n",
" for i, pred in enumerate(preds):\n",
" predictions.append(pred.cpu().numpy().flatten())\n",
" filenames.append(file_names[i])\n",
" \n",
" return predictions, filenames\n",
"\n",
"# Load validation set\n",
"if os.environ.get('DATA_PATH'):\n",
" DATA_PATH = os.environ.get(\"DATA_PATH\") + \"/\" \n",
"else:\n",
" DATA_PATH = \"\" # Fallback for local testing\n",
"# Load validation set\n",
"val_paths = generate_file_paths(DATA_PATH + 'validation_set')\n",
"val_dataset = CustomDataset_test(file_paths=val_paths)\n",
"val_loader = DataLoader(\n",
" val_dataset,\n",
" batch_size=1,\n",
" shuffle=False,\n",
" num_workers=0\n",
")\n",
"\n",
"# Load testing set\n",
"if os.environ.get('DATA_PATH'):\n",
" DATA_PATH = os.environ.get(\"DATA_PATH\") + \"/\" \n",
"else:\n",
" DATA_PATH = \"\" # Fallback for local testing\n",
"# Load testing set\n",
"test_paths = generate_file_paths(DATA_PATH + 'testing_set')\n",
"test_dataset = CustomDataset_test(file_paths=test_paths)\n",
"test_loader = DataLoader(\n",
" test_dataset,\n",
" batch_size=1,\n",
" shuffle=False,\n",
" num_workers=0\n",
")\n",
"\n",
"# Run inference on validation set\n",
"print(\"Running inference on validation set...\")\n",
"val_predictions, val_filenames = run_inference(dimssnet, val_loader)\n",
"\n",
"# Save validation results to CSV\n",
"val_results = []\n",
"for filename, pred in zip(val_filenames, val_predictions):\n",
" # Create a row with filename and flattened predictions\n",
" row = {'filename': filename}\n",
" for i, p in enumerate(pred):\n",
" row[f'pixel_{i}'] = p\n",
" val_results.append(row)\n",
"\n",
"val_df = pd.DataFrame(val_results)\n",
"val_df.to_csv('submission_val.csv', index=False)\n",
"print(f\"Validation results saved to output_validation.csv with shape: {val_df.shape}\")\n",
"\n",
"# Run inference on testing set\n",
"print(\"Running inference on testing set...\")\n",
"test_predictions, test_filenames = run_inference(dimssnet, test_loader)\n",
"\n",
"# Save testing results to CSV\n",
"test_results = []\n",
"for filename, pred in zip(test_filenames, test_predictions):\n",
" # Create a row with filename and flattened predictions\n",
" row = {'filename': filename}\n",
" for i, p in enumerate(pred):\n",
" row[f'pixel_{i}'] = p\n",
" test_results.append(row)\n",
"\n",
"test_df = pd.DataFrame(test_results)\n",
"test_df.to_csv('submission_test.csv', index=False)\n",
"print(f\"Testing results saved to output_testing.csv with shape: {test_df.shape}\")\n",
"\n",
"print(\"\\nInference completed! Results saved to:\")\n",
"print(\"- submission_val_ref.csv (for validation set leaderboard)\")\n",
"print(\"- submission_test_ref.csv (for testing set leaderboard)\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "72fde72d",
"metadata": {},
"outputs": [],
"source": [
"import zipfile\n",
"import os\n",
"\n",
"# Define the files to zip and the zip file name.\n",
"files_to_zip = ['submission_val.csv', 'submission_test.csv']\n",
"zip_filename = 'submission.zip'\n",
"\n",
"# Create a zip file\n",
"with zipfile.ZipFile(zip_filename, 'w') as zipf:\n",
" for file in files_to_zip:\n",
" # Add the file to the zip fil\n",
" zipf.write(file, os.path.basename(file))\n",
"\n",
"print(f'{zip_filename} Created successfully!')"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "ioai",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.11"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|