File size: 56,082 Bytes
fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 a6a2fc4 fba52a7 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 | import json
import time
import numpy as np
from scipy import ndimage
from scipy.stats import mode
from src.utils import matrix2answer
def find_grid(image, frame=False, possible_colors=None):
"""Looks for the grid in image and returns color and size"""
grid_color = -1
size = [1, 1]
if possible_colors is None:
possible_colors = list(range(10))
for color in possible_colors:
for i in range(size[0] + 1, image.shape[0] // 2 + 1):
if (image.shape[0] + 1) % i == 0:
step = (image.shape[0] + 1) // i
if (image[(step - 1) :: step] == color).all():
size[0] = i
grid_color = color
for i in range(size[1] + 1, image.shape[1] // 2 + 1):
if (image.shape[1] + 1) % i == 0:
step = (image.shape[1] + 1) // i
if (image[:, (step - 1) :: step] == color).all():
size[1] = i
grid_color = color
if grid_color == -1 and not frame:
color_candidate = image[0, 0]
if (
(image[0] == color_candidate).all()
and (image[-1] == color_candidate).all()
and (image[:, -1] == color_candidate).all()
and (image[:, 0] == color_candidate).all()
):
grid_color, size, _ = find_grid(
image[1 : image.shape[0] - 1, 1 : image.shape[1] - 1], frame=True, possible_colors=[color_candidate]
)
return grid_color, size, frame
else:
return grid_color, size, frame
return grid_color, size, frame
def find_color_boundaries(array, color):
"""Looks for the boundaries of any color and returns them"""
if not (array == color).any():
return None
ind_0 = np.arange(array.shape[0])
ind_1 = np.arange(array.shape[1])
temp_0 = ind_0[(array == color).max(axis=1)] # axis 0
min_0, max_0 = temp_0.min(), temp_0.max()
temp_1 = ind_1[(array == color).max(axis=0)] # axis 1
min_1, max_1 = temp_1.min(), temp_1.max()
return min_0, max_0, min_1, max_1
def get_color_max(image, color):
"""Returns the part of the image inside the color boundaries"""
boundaries = find_color_boundaries(image, color)
if boundaries:
return (0, image[boundaries[0] : boundaries[1] + 1, boundaries[2] : boundaries[3] + 1])
else:
return 1, None
def get_pixel(image, i, j):
"""Returns the pixel by coordinates"""
if i >= image.shape[0] or j >= image.shape[1]:
return 1, None
return 0, image[i : i + 1, j : j + 1]
def get_pixel_fixed(image, i):
return 0, np.array([[i]])
def get_grid(image, grid_size, cell, frame=False):
""" returns the particular cell form the image with grid"""
if frame:
return get_grid(image[1 : image.shape[0] - 1, 1 : image.shape[1] - 1], grid_size, cell, frame=False)
if cell[0] >= grid_size[0] or cell[1] >= grid_size[1]:
return 1, None
steps = ((image.shape[0] + 1) // grid_size[0], (image.shape[1] + 1) // grid_size[1])
block = image[steps[0] * cell[0] : steps[0] * (cell[0] + 1) - 1, steps[1] * cell[1] : steps[1] * (cell[1] + 1) - 1]
return 0, block
def get_half(image, side):
""" returns the half of the image"""
if side not in ["l", "r", "t", "b", "long1", "long2"]:
return 1, None
if side == "l":
return 0, image[:, : (image.shape[1]) // 2]
elif side == "r":
return 0, image[:, -((image.shape[1]) // 2) :]
elif side == "b":
return 0, image[-((image.shape[0]) // 2) :, :]
elif side == "t":
return 0, image[: (image.shape[0]) // 2, :]
elif side == "long1":
if image.shape[0] >= image.shape[1]:
return get_half(image, "t")
else:
return get_half(image, "l")
elif side == "long2":
if image.shape[0] >= image.shape[1]:
return get_half(image, "b")
else:
return get_half(image, "r")
def get_corner(image, side):
"""Return one quadrant of the image; tl/tr/bl/br follow row-0-at-top conventions."""
if side not in ["tl", "tr", "bl", "br"]:
return 1, None
size = (image.shape[0]) // 2, (image.shape[1]) // 2
if side == "tl":
return 0, image[: size[0], : size[1]]
if side == "tr":
return 0, image[: size[0], -size[1] :]
if side == "bl":
return 0, image[-size[0] :, : size[1]]
if side == "br":
return 0, image[-size[0] :, -size[1] :]
def get_k_part(image, num, k):
if image.shape[0] > image.shape[1]:
max_axis = 0
max_shape = image.shape[0]
else:
max_axis = 1
max_shape = image.shape[1]
if max_shape % num != 0:
return 1, None
size = max_shape // num
if max_axis == 0:
return 0, image[k * size : (k + 1) * size]
else:
return 0, image[:, k * size : (k + 1) * size]
def get_rotation(image, k):
return 0, np.rot90(image, k)
def get_transpose(image):
return 0, np.transpose(image)
def get_roll(image, shift, axis):
return 0, np.roll(image, shift=shift, axis=axis)
def get_cut_edge(image, l, r, t, b):
"""deletes pixels from some sided of an image"""
return 0, image[t : image.shape[0] - b, l : image.shape[1] - r]
def get_resize(image, scale):
""" resizes image according to scale"""
if isinstance(scale, int):
if image.shape[0] % scale != 0 or image.shape[1] % scale != 0:
return 1, None
if image.shape[0] < scale or image.shape[1] < scale:
return 2, None
arrays = []
size = image.shape[0] // scale, image.shape[1] // scale
for i in range(scale):
for j in range(scale):
arrays.append(image[i::scale, j::scale])
# keepdims=True restores the pre-SciPy-1.11 shape that the rest of the
# pipeline assumes (mode shape == (1, H, W), so .mode[0] is the 2D block).
result = mode(np.stack(arrays), axis=0, keepdims=True).mode[0]
else:
size = int(image.shape[0] / scale), int(image.shape[1] / scale)
result = []
for i in range(size[0]):
result.append([])
for j in range(size[1]):
result[-1].append(image[int(i * scale), int(j * scale)])
result = np.uint8(result)
return 0, result
def get_resize_to(image, size_x, size_y):
""" resizes image according to scale"""
scale_x = image.shape[0] // size_x
scale_y = image.shape[1] // size_y
if scale_x == 0 or scale_y == 0:
return 3, None
if image.shape[0] % scale_x != 0 or image.shape[1] % scale_y != 0:
return 1, None
if image.shape[0] < scale_x or image.shape[1] < scale_y:
return 2, None
arrays = []
for i in range(scale_x):
for j in range(scale_y):
arrays.append(image[i::scale_x, j::scale_y])
result = mode(np.stack(arrays), axis=0, keepdims=True).mode[0]
return 0, result
def get_reflect(image, side):
""" returns images generated by reflections of the input"""
if side not in ["r", "l", "t", "b", "rt", "rb", "lt", "lb"]:
return 1, None
try:
if side == "r":
result = np.zeros((image.shape[0], image.shape[1] * 2 - 1))
result[:, : image.shape[1]] = image
result[:, -image.shape[1] :] = image[:, ::-1]
elif side == "l":
result = np.zeros((image.shape[0], image.shape[1] * 2 - 1))
result[:, : image.shape[1]] = image[:, ::-1]
result[:, -image.shape[1] :] = image
elif side == "b":
result = np.zeros((image.shape[0] * 2 - 1, image.shape[1]))
result[: image.shape[0], :] = image
result[-image.shape[0] :, :] = image[::-1]
elif side == "t":
result = np.zeros((image.shape[0] * 2 - 1, image.shape[1]))
result[: image.shape[0], :] = image[::-1]
result[-image.shape[0] :, :] = image
elif side == "rb":
result = np.zeros((image.shape[0] * 2 - 1, image.shape[1] * 2 - 1))
result[: image.shape[0], : image.shape[1]] = image
result[: image.shape[0], -image.shape[1] :] = image[:, ::-1]
result[-image.shape[0] :, : image.shape[1]] = image[::-1, :]
result[-image.shape[0] :, -image.shape[1] :] = image[::-1, ::-1]
elif side == "rt":
result = np.zeros((image.shape[0] * 2 - 1, image.shape[1] * 2 - 1))
result[: image.shape[0], : image.shape[1]] = image[::-1, :]
result[: image.shape[0], -image.shape[1] :] = image[::-1, ::-1]
result[-image.shape[0] :, : image.shape[1]] = image
result[-image.shape[0] :, -image.shape[1] :] = image[:, ::-1]
elif side == "lt":
result = np.zeros((image.shape[0] * 2 - 1, image.shape[1] * 2 - 1))
result[: image.shape[0], : image.shape[1]] = image[::-1, ::-1]
result[: image.shape[0], -image.shape[1] :] = image[::-1, :]
result[-image.shape[0] :, : image.shape[1]] = image[:, ::-1]
result[-image.shape[0] :, -image.shape[1] :] = image
elif side == "lb":
result = np.zeros((image.shape[0] * 2 - 1, image.shape[1] * 2 - 1))
result[: image.shape[0], : image.shape[1]] = image[:, ::-1]
result[: image.shape[0], -image.shape[1] :] = image
result[-image.shape[0] :, : image.shape[1]] = image[::-1, ::-1]
result[-image.shape[0] :, -image.shape[1] :] = image[::-1, :]
except:
return 2, None
return 0, result
def get_color_swap(image, color_1, color_2):
"""swapping two colors"""
if not (image == color_1).any() and not (image == color_2).any():
return 1, None
result = image.copy()
result[image == color_1] = color_2
result[image == color_2] = color_1
return 0, result
def get_cut(image, x1, y1, x2, y2):
if x1 >= x2 or y1 >= y2:
return 1, None
else:
return 0, image[x1:x2, y1:y2]
def get_min_block(image, full=True):
if full:
structure = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]
else:
structure = [[0, 1, 0], [1, 1, 1], [0, 1, 0]]
masks, n_masks = ndimage.label(image, structure=structure)
sizes = [(masks == i).sum() for i in range(1, n_masks + 1)]
if n_masks == 0:
return 2, None
min_n = np.argmin(sizes) + 1
boundaries = find_color_boundaries(masks, min_n)
if boundaries:
return (0, image[boundaries[0] : boundaries[1] + 1, boundaries[2] : boundaries[3] + 1])
else:
return 1, None
def get_min_block_mask(image, full=True):
if full:
structure = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]
else:
structure = [[0, 1, 0], [1, 1, 1], [0, 1, 0]]
masks, n_masks = ndimage.label(image, structure=structure)
sizes = [(masks == i).sum() for i in range(1, n_masks + 1)]
if n_masks == 0:
return 2, None
min_n = np.argmin(sizes) + 1
return 0, masks == min_n
def get_max_block_mask(image, full=True):
if full:
structure = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]
else:
structure = [[0, 1, 0], [1, 1, 1], [0, 1, 0]]
masks, n_masks = ndimage.label(image, structure=structure)
sizes = [(masks == i).sum() for i in range(1, n_masks + 1)]
if n_masks == 0:
return 2, None
min_n = np.argmax(sizes) + 1
return 0, masks == min_n
def get_max_block(image, full=True):
if full:
structure = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]
else:
structure = [[0, 1, 0], [1, 1, 1], [0, 1, 0]]
masks, n_masks = ndimage.label(image, structure=structure)
sizes = [(masks == i).sum() for i in range(1, n_masks + 1)]
if n_masks == 0:
return 2, None
max_n = np.argmax(sizes) + 1
boundaries = find_color_boundaries(masks, max_n)
if boundaries:
return (0, image[boundaries[0] : boundaries[1] + 1, boundaries[2] : boundaries[3] + 1])
else:
return 1, None
def get_block_with_side_colors(image, block_type="min", structure=0):
if structure == 0:
structure = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]
else:
structure = [[0, 1, 0], [1, 1, 1], [0, 1, 0]]
masks, n_masks = ndimage.label(image, structure=structure)
if n_masks == 0:
return 2, None
unique_nums = []
for i in range(1, n_masks + 1):
unique = np.unique(image[masks == i])
unique_nums.append(len(unique))
if block_type == "min":
n = np.argmin(unique_nums) + 1
else:
n = np.argmax(unique_nums) + 1
boundaries = find_color_boundaries(masks, n)
if boundaries:
return (0, image[boundaries[0] : boundaries[1] + 1, boundaries[2] : boundaries[3] + 1])
else:
return 1, None
def get_block_with_side_colors_count(image, block_type="min", structure=0):
if structure == 0:
structure = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]
else:
structure = [[0, 1, 0], [1, 1, 1], [0, 1, 0]]
masks, n_masks = ndimage.label(image, structure=structure)
if n_masks == 0:
return 2, None
unique_nums = []
for i in range(1, n_masks + 1):
unique, counts = np.unique(image[masks == i], return_counts=True)
unique_nums.append(min(counts))
if block_type == "min":
n = np.argmin(unique_nums) + 1
else:
n = np.argmax(unique_nums) + 1
boundaries = find_color_boundaries(masks, n)
if boundaries:
return (0, image[boundaries[0] : boundaries[1] + 1, boundaries[2] : boundaries[3] + 1])
else:
return 1, None
def get_color(color_dict, colors):
""" retrive the absolute number corresponding a color set by color_dict"""
for i, color in enumerate(colors):
for data in color:
equal = True
for k, v in data.items():
if k not in color_dict or v != color_dict[k]:
equal = False
break
if equal:
return i
return -1
def get_mask_from_block(image, color):
if color in np.unique(image, return_counts=False):
return 0, image == color
else:
return 1, None
def get_background(image, color):
return 0, np.uint8(np.ones_like(image) * color)
def get_mask_from_max_color_coverage(image, color):
if color in np.unique(image, return_counts=False):
boundaries = find_color_boundaries(image, color)
result = (image.copy() * 0).astype(bool)
result[boundaries[0] : boundaries[1] + 1, boundaries[2] : boundaries[3] + 1] = True
return 0, result
else:
return 1, None
def add_unique_colors(image, result, colors=None):
"""adds information about colors unique for some parts of the image"""
if colors is None:
colors = np.unique(image)
unique_side = [False for i in range(10)]
unique_corner = [False for i in range(10)]
half_size = (((image.shape[0] + 1) // 2), ((image.shape[1] + 1) // 2))
for (image_part, side, unique_list) in [
(image[: half_size[0]], "bottom", unique_side),
(image[-half_size[0] :], "top", unique_side),
(image[:, : half_size[1]], "right", unique_side),
(image[:, -half_size[1] :], "left", unique_side),
(image[: half_size[0], : half_size[1]], "tl", unique_corner),
(image[: half_size[0], -half_size[1] :], "tr", unique_corner),
(image[-half_size[0] :, : half_size[1]], "bl", unique_corner),
(image[-half_size[0] :, -half_size[1] :], "br", unique_corner),
]:
unique = np.uint8(np.unique(image_part))
if len(unique) == len(colors) - 1:
color = [x for x in colors if x not in unique][0]
unique_list[color] = True
result["colors"][color].append({"type": "unique", "side": side})
for i in range(10):
if unique_corner[i]:
result["colors"][i].append({"type": "unique", "side": "corner"})
if unique_side[i]:
result["colors"][i].append({"type": "unique", "side": "side"})
if unique_side[i] or unique_corner[i]:
result["colors"][i].append({"type": "unique", "side": "any"})
return
def add_center_color(image, result, colors=None):
i = image.shape[0] // 4
j = image.shape[1] // 4
center = image[i : image.shape[0] - i, j : image.shape[1] - j]
values, counts = np.unique(center, return_counts=True)
if len(counts) > 0:
ind = np.argmax(counts)
color = values[ind]
result["colors"][color].append({"type": "center"})
def get_color_scheme(image, target_image=None, params=None):
"""processes original image and returns dict color scheme"""
result = {
"grid_color": -1,
"colors": [[], [], [], [], [], [], [], [], [], []],
"colors_sorted": [],
"grid_size": [1, 1],
}
if params is None:
params = ["coverage", "unique", "corners", "top", "grid"]
# preparing colors info
unique, counts = np.unique(image, return_counts=True)
colors = [unique[i] for i in np.argsort(counts)]
result["colors_sorted"] = colors
result["colors_num"] = len(colors)
for color in range(10):
# use abs color value - same for any image
result["colors"][color].append({"type": "abs", "k": color})
if len(colors) == 2 and 0 in colors:
result["colors"][[x for x in colors if x != 0][0]].append({"type": "non_zero"})
if "coverage" in params:
for k, color in enumerate(colors):
# use k-th colour (sorted by presence on image)
result["colors"][color].append({"type": "min", "k": k})
# use k-th colour (sorted by presence on image)
result["colors"][color].append({"type": "max", "k": len(colors) - k - 1})
if "unique" in params:
add_unique_colors(image, result, colors=None)
add_center_color(image, result)
if "corners" in params:
# colors in the corners of images
result["colors"][image[0, 0]].append({"type": "corner", "side": "tl"})
result["colors"][image[0, -1]].append({"type": "corner", "side": "tr"})
result["colors"][image[-1, 0]].append({"type": "corner", "side": "bl"})
result["colors"][image[-1, -1]].append({"type": "corner", "side": "br"})
if "top" in params:
# colors that are on top of other and have full vertical on horizontal line
for k in range(10):
mask = image == k
is_on_top0 = mask.min(axis=0).any()
is_on_top1 = mask.min(axis=1).any()
if is_on_top0:
result["colors"][k].append({"type": "on_top", "side": "0"})
if is_on_top1:
result["colors"][k].append({"type": "on_top", "side": "1"})
if is_on_top1 or is_on_top0:
result["colors"][k].append({"type": "on_top", "side": "any"})
if "grid" in params:
grid_color, grid_size, frame = find_grid(image)
if grid_color >= 0:
result["grid_color"] = grid_color
result["grid_size"] = grid_size
result["grid_frame"] = frame
result["colors"][grid_color].append({"type": "grid"})
return result
def add_block(target_dict, image, params_list):
array_hash = hash(matrix2answer(image))
if array_hash not in target_dict["arrays"]:
target_dict["arrays"][array_hash] = {"array": image, "params": []}
for params in params_list:
params_hash = get_dict_hash(params)
target_dict["arrays"][array_hash]["params"].append(params)
target_dict["params"][params_hash] = array_hash
def get_original(image):
return 0, image
def get_inversed_colors(image):
unique = np.unique(image)
if len(unique) != 2:
return 1, None
result = image.copy()
result[image == unique[0]] = unique[1]
result[image == unique[1]] = unique[0]
return 0, result
def generate_blocks(image, result, max_time=600, max_blocks=200000, max_masks=200000, target_image=None, params=None):
all_params = [
"initial",
"background",
"min_max_blocks",
"block_with_side_colors",
"max_area_covered",
"grid_cells",
"halves",
"corners",
"rotate",
"transpose",
"cut_edges",
"resize",
"reflect",
"cut_parts",
"swap_colors",
"k_part",
]
if not params:
params = all_params
start_time = time.time()
result["blocks"] = {"arrays": {}, "params": {}}
if "initial" in params:
# starting with the original image
add_block(result["blocks"], image, [[{"type": "original"}]])
# inverse colors
status, block = get_inversed_colors(image)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(result["blocks"], block, [[{"type": "inversed_colors"}]])
# adding min and max blocks
if (
("min_max_blocks" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
# print("min_max_blocks")
for full in [True, False]:
status, block = get_max_block(image, full)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(result["blocks"], block, [[{"type": "max_block", "full": full}]])
if (
("block_with_side_colors" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
# print("min_max_blocks")
for block_type in ["min", "max"]:
for structure in [0, 1]:
status, block = get_block_with_side_colors(image, block_type, structure)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(
result["blocks"],
block,
[[{"type": "block_with_side_colors", "block_type": block_type, "structure": structure}]],
)
for block_type in ["min", "max"]:
for structure in [0, 1]:
status, block = get_block_with_side_colors_count(image, block_type, structure)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(
result["blocks"],
block,
[[{"type": "block_with_side_colors_count", "block_type": block_type, "structure": structure}]],
)
# print(sum([len(x['params']) for x in result['blocks']['arrays'].values()]))
# adding background
if ("background" in params) and (time.time() - start_time < max_time):
# print("background")
for color in range(10):
status, block = get_background(image, color)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = []
for color_dict in result["colors"][color].copy():
params_list.append([{"type": "background", "color": color_dict}])
add_block(result["blocks"], block, params_list)
# adding the max area covered by each color
if ("max_area_covered" in params) and (time.time() - start_time < max_time):
# print("max_area_covered")
for color in result["colors_sorted"]:
status, block = get_color_max(image, color)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = []
for color_dict in result["colors"][color].copy():
params_list.append([{"type": "color_max", "color": color_dict}])
add_block(result["blocks"], block, params_list)
# adding grid cells
if (
("grid_cells" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
if result["grid_color"] > 0:
for i in range(result["grid_size"][0]):
for j in range(result["grid_size"][1]):
status, block = get_grid(image, result["grid_size"], (i, j), frame=result["grid_frame"])
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(
result["blocks"],
block,
[
[
{
"type": "grid",
"grid_size": result["grid_size"],
"cell": [i, j],
"frame": result["grid_frame"],
}
]
],
)
# adding halves of the images
if ("halves" in params) and (time.time() - start_time < max_time) and (len(result["blocks"]["arrays"]) < max_blocks):
for side in ["l", "r", "t", "b", "long1", "long2"]:
status, block = get_half(image, side=side)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(result["blocks"], block, [[{"type": "half", "side": side}]])
# extracting pixels from image
if ("pixels" in params) and (time.time() - start_time < max_time) and (len(result["blocks"]["arrays"]) < max_blocks):
stop = False
for i in range(image.shape[0]):
if stop:
break
for j in range(image.shape[1]):
status, block = get_pixel(image, i=i, j=j)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(result["blocks"], block, [[{"type": "pixel", "i": i, "j": j}]])
if len(result["blocks"]["arrays"]) >= max_blocks:
stop = True
break
# extracting pixels from image
if (
("pixel_fixed" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
for i in range(10):
status, block = get_pixel_fixed(image, i=i)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(result["blocks"], block, [[{"type": "pixel_fixed", "i": i}]])
# adding halves of the images
if ("k_part" in params) and (time.time() - start_time < max_time) and (len(result["blocks"]["arrays"]) < max_blocks):
for num in [3, 4]:
for k in range(num):
status, block = get_k_part(image, num=num, k=k)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(result["blocks"], block, [[{"type": "k_part", "num": num, "k": k}]])
# adding corners of the images
if (
("corners" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
for side in ["tl", "tr", "bl", "br"]:
status, block = get_corner(image, side=side)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
add_block(result["blocks"], block, [[{"type": "corner", "side": side}]])
main_blocks_num = len(result["blocks"])
# rotate all blocks
if ("rotate" in params) and (time.time() - start_time < max_time) and (len(result["blocks"]["arrays"]) < max_blocks):
current_blocks = result["blocks"]["arrays"].copy()
for k in range(1, 4):
for key, data in current_blocks.items():
status, block = get_rotation(data["array"], k=k)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = [i + [{"type": "rotation", "k": k}] for i in data["params"]]
add_block(result["blocks"], block, params_list)
# transpose all blocks
if (
("transpose" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
current_blocks = result["blocks"]["arrays"].copy()
for key, data in current_blocks.items():
status, block = get_transpose(data["array"])
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = [i + [{"type": "transpose"}] for i in data["params"]]
add_block(result["blocks"], block, params_list)
# cut edges for all blocks
if (
("cut_edges" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
current_blocks = result["blocks"]["arrays"].copy()
for l, r, t, b in [
(1, 1, 1, 1),
(1, 0, 0, 0),
(0, 1, 0, 0),
(0, 0, 1, 0),
(0, 0, 0, 1),
(1, 1, 0, 0),
(1, 0, 0, 1),
(0, 0, 1, 1),
(0, 1, 1, 0),
]:
if time.time() - start_time < max_time:
for key, data in current_blocks.items():
status, block = get_cut_edge(data["array"], l=l, r=r, t=t, b=b)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = [
i + [{"type": "cut_edge", "l": l, "r": r, "t": t, "b": b}] for i in data["params"]
]
add_block(result["blocks"], block, params_list)
# resize all blocks
if ("resize" in params) and (time.time() - start_time < max_time) and (len(result["blocks"]["arrays"]) < max_blocks):
current_blocks = result["blocks"]["arrays"].copy()
for scale in [2, 3, 1 / 2, 1 / 3]:
for key, data in current_blocks.items():
status, block = get_resize(data["array"], scale)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = [i + [{"type": "resize", "scale": scale}] for i in data["params"]]
add_block(result["blocks"], block, params_list)
for size_x, size_y in [(2, 2), (3, 3)]:
for key, data in current_blocks.items():
status, block = get_resize_to(data["array"], size_x, size_y)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = [
i + [{"type": "resize_to", "size_x": size_x, "size_y": size_y}] for i in data["params"]
]
add_block(result["blocks"], block, params_list)
# reflect all blocks
if (
("reflect" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
current_blocks = result["blocks"]["arrays"].copy()
for side in ["r", "l", "t", "b", "rt", "rb", "lt", "lb"]:
if time.time() - start_time < max_time:
for key, data in current_blocks.items():
status, block = get_reflect(data["array"], side)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
params_list = [i + [{"type": "reflect", "side": side}] for i in data["params"]]
add_block(result["blocks"], block, params_list)
# cut some parts of images
if (
("cut_parts" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
max_x = image.shape[0]
max_y = image.shape[1]
min_block_size = 2
stop = False
for x1 in range(0, max_x - min_block_size):
if stop:
break
if time.time() - start_time < max_time:
if max_x - x1 <= min_block_size:
continue
for x2 in range(x1 + min_block_size, max_x):
if stop:
break
for y1 in range(0, max_y - min_block_size):
if stop:
break
if max_y - y1 <= min_block_size:
continue
for y2 in range(y1 + min_block_size, max_y):
status, block = get_cut(image, x1, y1, x2, y2)
if status == 0:
add_block(
result["blocks"], block, [[{"type": "cut", "x1": x1, "x2": x2, "y1": y1, "y2": y2}]]
)
if len(result["blocks"]["arrays"]) >= max_blocks:
stop = True
break
list_param_list = []
list_blocks = []
# swap some colors
if (
("swap_colors" in params)
and (time.time() - start_time < max_time)
and (len(result["blocks"]["arrays"]) < max_blocks)
):
current_blocks = result["blocks"]["arrays"].copy()
# Budget is measured against current block count + pending to-be-added,
# so the soft cap actually holds after the subsequent add_block loop.
budget = max_blocks - len(result["blocks"]["arrays"])
stop = False
for color_1 in range(9):
if stop:
break
if time.time() - start_time < max_time:
for color_2 in range(color_1 + 1, 10):
if stop:
break
for key, data in current_blocks.items():
if stop:
break
status, block = get_color_swap(data["array"], color_1, color_2)
if status == 0 and block.shape[0] > 0 and block.shape[1] > 0:
for color_dict_1 in result["colors"][color_1].copy():
if stop:
break
for color_dict_2 in result["colors"][color_2].copy():
list_param_list.append(
[
j
+ [{"type": "color_swap", "color_1": color_dict_1, "color_2": color_dict_2}]
for j in data["params"]
]
)
list_blocks.append(block)
if len(list_blocks) >= budget:
stop = True
break
for block, params_list in zip(list_blocks, list_param_list):
add_block(result["blocks"], block, params_list)
if time.time() - start_time > max_time:
print("Time is over")
if len(result["blocks"]["arrays"]) >= max_blocks:
print("Max number of blocks exceeded")
return result
def generate_masks(image, result, max_time=600, max_blocks=200000, max_masks=200000, target_image=None, params=None):
start_time = time.time()
all_params = ["initial_masks", "additional_masks", "coverage_masks", "min_max_masks"]
if not params:
params = all_params
result["masks"] = {"arrays": {}, "params": {}}
# making one mask for each generated block
current_blocks = result["blocks"]["arrays"].copy()
if ("initial_masks" in params) and (time.time() - start_time < max_time * 2):
stop = False
for key, data in current_blocks.items():
if stop:
break
for color in result["colors_sorted"]:
status, mask = get_mask_from_block(data["array"], color)
if status == 0 and mask.shape[0] > 0 and mask.shape[1] > 0:
params_list = [
{"operation": "none", "params": {"block": i, "color": color_dict}}
for i in data["params"]
for color_dict in result["colors"][color]
]
add_block(result["masks"], mask, params_list)
if len(result["masks"]["arrays"]) >= max_masks:
stop = True
break
initial_masks = result["masks"]["arrays"].copy()
if (
("initial_masks" in params)
and (time.time() - start_time < max_time * 2)
and (len(result["masks"]["arrays"]) < max_masks)
):
for key, mask in initial_masks.items():
add_block(
result["masks"],
np.logical_not(mask["array"]),
[{"operation": "not", "params": param["params"]} for param in mask["params"]],
)
if len(result["masks"]["arrays"]) >= max_masks:
break
initial_masks = result["masks"]["arrays"].copy()
masks_to_add = []
processed = []
if ("additional_masks" in params) and (time.time() - start_time < max_time * 2):
# `additional_masks` is O(n^2) in number of existing masks and each
# iteration emits up to 3 new masks, so it is the classic explosion
# point. Budget against max_masks so we stop early.
budget = max(0, max_masks - len(result["masks"]["arrays"]))
stop = False
for key1, mask1 in initial_masks.items():
if stop:
break
processed.append(key1)
if time.time() - start_time < max_time * 2 and (
target_image is None
or (target_image.shape == mask1["array"].shape)
or (target_image.shape == mask1["array"].T.shape)
):
for key2, mask2 in initial_masks.items():
if stop:
break
if key2 in processed:
continue
if (mask1["array"].shape[0] == mask2["array"].shape[0]) and (
mask1["array"].shape[1] == mask2["array"].shape[1]
):
params_list_and = []
params_list_or = []
params_list_xor = []
for param1 in mask1["params"]:
for param2 in mask2["params"]:
params_list_and.append(
{"operation": "and", "params": {"mask1": param1, "mask2": param2}}
)
params_list_or.append({"operation": "or", "params": {"mask1": param1, "mask2": param2}})
params_list_xor.append(
{"operation": "xor", "params": {"mask1": param1, "mask2": param2}}
)
masks_to_add.append(
(result["masks"], np.logical_and(mask1["array"], mask2["array"]), params_list_and)
)
masks_to_add.append(
(result["masks"], np.logical_or(mask1["array"], mask2["array"]), params_list_or)
)
masks_to_add.append(
(result["masks"], np.logical_xor(mask1["array"], mask2["array"]), params_list_xor)
)
if len(masks_to_add) >= budget:
stop = True
break
for path, array, params_list in masks_to_add:
add_block(path, array, params_list)
# coverage_masks
if ("coverage_masks" in params) and (time.time() - start_time < max_time * 2):
for color in result["colors_sorted"][1:]:
status, mask = get_mask_from_max_color_coverage(image, color)
if status == 0 and mask.shape[0] > 0 and mask.shape[1] > 0:
params_list = [
{"operation": "coverage", "params": {"color": color_dict}}
for color_dict in result["colors"][color].copy()
]
add_block(result["masks"], mask, params_list)
# coverage_masks
if ("min_max_masks" in params) and (time.time() - start_time < max_time * 2):
status, mask = get_min_block_mask(image)
if status == 0 and mask.shape[0] > 0 and mask.shape[1] > 0:
params_list = [{"operation": "min_block"}]
add_block(result["masks"], mask, params_list)
status, mask = get_max_block_mask(image)
if status == 0 and mask.shape[0] > 0 and mask.shape[1] > 0:
params_list = [{"operation": "max_block"}]
add_block(result["masks"], mask, params_list)
if time.time() - start_time > max_time:
print("Time is over")
if len(result["masks"]["arrays"]) >= max_masks:
print("Max number of masks exceeded")
return result
def process_image(
image, max_time=600, max_blocks=200000, max_masks=200000, target_image=None, params=None, color_params=None
):
"""processes the original image and returns dict with structured image blocks"""
result = get_color_scheme(image, target_image=target_image, params=color_params)
result = generate_blocks(image, result, max_time, max_blocks, max_masks, target_image, params)
result = generate_masks(image, result, max_time, max_blocks, max_masks, target_image, params)
return result
def get_mask_from_block_params(image, params, block_cache=None, mask_cache=None, color_scheme=None):
if mask_cache is None:
mask_cache = {"arrays": {}, "params": {}}
dict_hash = get_dict_hash(params)
if dict_hash in mask_cache:
mask = mask_cache["arrays"][mask_cache["params"][dict_hash]]["array"]
if len(mask) == 0:
return 1, None
else:
return 0, mask
if params["operation"] == "none":
status, block = get_predict(image, params["params"]["block"], block_cache, color_scheme)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 1, None
if not color_scheme:
color_scheme = get_color_scheme(image)
color_num = get_color(params["params"]["color"], color_scheme["colors"])
if color_num < 0:
add_block(mask_cache, np.array([[]]), [params])
return 2, None
status, mask = get_mask_from_block(block, color_num)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 6, None
add_block(mask_cache, mask, [params])
return 0, mask
elif params["operation"] == "not":
new_params = params.copy()
new_params["operation"] = "none"
status, mask = get_mask_from_block_params(
image, new_params, block_cache=block_cache, color_scheme=color_scheme, mask_cache=mask_cache
)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 3, None
mask = np.logical_not(mask)
add_block(mask_cache, mask, [params])
return 0, mask
elif params["operation"] in ["and", "or", "xor"]:
new_params = params["params"]["mask1"]
status, mask1 = get_mask_from_block_params(
image, new_params, block_cache=block_cache, color_scheme=color_scheme, mask_cache=mask_cache
)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 4, None
new_params = params["params"]["mask2"]
status, mask2 = get_mask_from_block_params(
image, new_params, block_cache=block_cache, color_scheme=color_scheme, mask_cache=mask_cache
)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 5, None
if mask1.shape[0] != mask2.shape[0] or mask1.shape[1] != mask2.shape[1]:
add_block(mask_cache, np.array([[]]), [params])
return 6, None
if params["operation"] == "and":
mask = np.logical_and(mask1, mask2)
elif params["operation"] == "or":
mask = np.logical_or(mask1, mask2)
elif params["operation"] == "xor":
mask = np.logical_xor(mask1, mask2)
add_block(mask_cache, mask, [params])
return 0, mask
elif params["operation"] == "coverage":
if not color_scheme:
color_scheme = get_color_scheme(image)
color_num = get_color(params["params"]["color"], color_scheme["colors"])
if color_num < 0:
add_block(mask_cache, np.array([[]]), [params])
return 2, None
status, mask = get_mask_from_max_color_coverage(image, color_num)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 6, None
add_block(mask_cache, mask, [params])
return 0, mask
elif params["operation"] == "min_block":
status, mask = get_min_block_mask(image)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 6, None
add_block(mask_cache, mask, [params])
return 0, mask
elif params["operation"] == "max_block":
status, mask = get_max_block_mask(image)
if status != 0:
add_block(mask_cache, np.array([[]]), [params])
return 6, None
add_block(mask_cache, mask, [params])
return 0, mask
def get_dict_hash(d):
return hash(json.dumps(d, sort_keys=True))
def get_predict(image, transforms, block_cache=None, color_scheme=None):
""" applies the list of transforms to the image"""
params_hash = get_dict_hash(transforms)
if params_hash in block_cache["params"]:
if block_cache["params"][params_hash] is None:
return 1, None
else:
return 0, block_cache["arrays"][block_cache["params"][params_hash]]["array"]
if not color_scheme:
color_scheme = get_color_scheme(image)
if len(transforms) > 1:
status, previous_image = get_predict(image, transforms[:-1], block_cache=block_cache, color_scheme=color_scheme)
if status != 0:
return status, None
else:
previous_image = image
transform = transforms[-1]
function = globals()["get_" + transform["type"]]
params = transform.copy()
params.pop("type")
for color_name in ["color", "color_1", "color_2"]:
if color_name in params:
params[color_name] = get_color(params[color_name], color_scheme["colors"])
if params[color_name] < 0:
return 2, None
status, result = function(previous_image, **params)
if status != 0 or len(result) == 0 or len(result[0]) == 0:
block_cache["params"][params_hash] = None
return 1, None
add_block(block_cache, result, [transforms])
return 0, result
def filter_colors(sample):
# filtering colors, that are not present in at least one of the images
all_colors = []
for color_scheme1 in sample["train"]:
list_of_colors = [get_dict_hash(color_dict) for i in range(10) for color_dict in color_scheme1["colors"][i]]
all_colors.append(list_of_colors)
for j in range(1, len(sample["train"])):
all_colors[0] = [x for x in all_colors[0] if x in all_colors[j]]
keep_colors = set(all_colors[0])
for color_scheme1 in sample["train"]:
for i in range(10):
j = 0
while j < len(color_scheme1["colors"][i]):
if get_dict_hash(color_scheme1["colors"][i][j]) in keep_colors:
j += 1
else:
del color_scheme1["colors"][i][j]
delete_colors = []
color_scheme0 = sample["train"][0]
for i in range(10):
if len(color_scheme0["colors"][i]) > 1:
for j, color_dict1 in enumerate(color_scheme0["colors"][i][::-1][:-1]):
hash1 = get_dict_hash(color_dict1)
delete = True
for color_dict2 in color_scheme0["colors"][i][::-1][j + 1 :]:
hash2 = get_dict_hash(color_dict2)
for color_scheme1 in list(sample["train"][1:]) + list(sample["test"]):
found = False
for k in range(10):
hash_array = [get_dict_hash(color_dict) for color_dict in color_scheme1["colors"][k]]
if hash1 in hash_array and hash2 in hash_array:
found = True
break
if not found:
delete = False
break
if delete:
delete_colors.append(hash1)
break
for color_scheme1 in sample["train"]:
for i in range(10):
j = 0
while j < len(color_scheme1["colors"][i]):
if get_dict_hash(color_scheme1["colors"][i][j]) in delete_colors:
del color_scheme1["colors"][i][j]
else:
j += 1
return
def filter_blocks(sample, arrays_type="blocks", max_time=60):
"""De-duplicate equivalent param lists across images.
This is O(n_blocks * m_params^2 * n_images * k_blocks_per_image) in the worst case,
which explodes on heavy preprocessor sets. A wall-clock budget (`max_time`) bounds
the filtering step so a pathological input can't wedge the worker for hours — we
simply stop early and leave the remaining duplicates in place (correctness is
preserved; only downstream speed suffers slightly).
"""
start_time = time.time()
delete_blocks = []
list_of_lists_of_sets = []
for arrays_list in [x[arrays_type]["arrays"].values() for x in sample["train"][1:]] + [
x[arrays_type]["arrays"].values() for x in sample["test"]
]:
list_of_lists_of_sets.append([])
for array in arrays_list:
list_of_lists_of_sets[-1].append({get_dict_hash(params_dict) for params_dict in array["params"]})
bailed = False
for initial_array in sample["train"][0][arrays_type]["arrays"].values():
if time.time() - start_time > max_time:
bailed = True
break
if len(initial_array["params"]) > 1:
for j, params_dict1 in enumerate(initial_array["params"][::-1][:-1]):
hash1 = get_dict_hash(params_dict1)
delete = True
for params_dict1 in initial_array["params"][::-1][j + 1 :]:
hash2 = get_dict_hash(params_dict1)
for lists_of_sets in list_of_lists_of_sets:
found = False
for hash_set in lists_of_sets:
if hash1 in hash_set and hash2 in hash_set:
found = True
break
if not found:
delete = False
break
if delete:
delete_blocks.append(hash1)
break
if bailed:
print(f"filter_{arrays_type} budget exceeded; skipping remaining dedup")
for arrays_list in [x[arrays_type]["arrays"].values() for x in sample["train"]] + [
x[arrays_type]["arrays"].values() for x in sample["test"]
]:
for array in arrays_list:
params_list = array["params"]
j = 0
while j < len(params_list):
if get_dict_hash(params_list[j]) in delete_blocks:
del params_list[j]
else:
j += 1
return
def extract_target_blocks(sample, color_params=None):
target_blocks_cache = []
params = ["initial", "block_with_side_colors", "min_max_blocks", "max_area_covered", "cut_parts"]
for n in range(len(sample["train"])):
target_image = np.uint8(sample["train"][n]["output"])
target_blocks_cache.append(get_color_scheme(target_image, params=color_params))
target_blocks_cache[-1].update(generate_blocks(target_image, target_blocks_cache[-1], params=params))
final_arrays = list(
set.intersection(
*[set(target_blocks_cache[n]["blocks"]["arrays"].keys()) for n in range(len(target_blocks_cache))]
)
)
for i, key in enumerate(final_arrays):
for n in range(len(sample["train"])):
params_list = [[{"type": "target", "k": i}]]
add_block(
sample["train"][n]["blocks"], target_blocks_cache[0]["blocks"]["arrays"][key]["array"], params_list
)
for n in range(len(sample["test"])):
params_list = [[{"type": "target", "k": i}]]
add_block(sample["test"][n]["blocks"], target_blocks_cache[0]["blocks"]["arrays"][key]["array"], params_list)
def preprocess_sample(
sample,
params=None,
color_params=None,
process_whole_ds=False,
max_blocks=200000,
max_masks=200000,
max_time=600,
):
""" make the whole preprocessing for particular sample
Extra knobs (all optional, defaults preserve historical behaviour):
- max_blocks / max_masks: per-image caps on the abstraction search space.
- max_time: per-image wall-time budget (seconds) for block/mask generation.
"""
for n, image in enumerate(sample["train"]):
original_image = np.uint8(image["input"])
target_image = np.uint8(sample["train"][n]["output"])
sample["train"][n].update(get_color_scheme(original_image, target_image=target_image, params=color_params))
for n, image in enumerate(sample["test"]):
original_image = np.uint8(image["input"])
sample["test"][n].update(get_color_scheme(original_image, params=color_params))
filter_colors(sample)
for n, image in enumerate(sample["train"]):
original_image = np.uint8(image["input"])
target_image = np.uint8(sample["train"][n]["output"])
sample["train"][n].update(
generate_blocks(
original_image,
sample["train"][n],
max_time=max_time,
max_blocks=max_blocks,
max_masks=max_masks,
target_image=target_image,
params=params,
)
)
for n, image in enumerate(sample["test"]):
original_image = np.uint8(image["input"])
sample["test"][n].update(
generate_blocks(
original_image,
sample["test"][n],
max_time=max_time,
max_blocks=max_blocks,
max_masks=max_masks,
params=params,
)
)
if params is not None and "target" in params:
extract_target_blocks(sample, color_params)
filter_blocks(sample)
for n, image in enumerate(sample["train"]):
original_image = np.uint8(image["input"])
target_image = np.uint8(sample["train"][n]["output"])
sample["train"][n].update(
generate_masks(
original_image,
sample["train"][n],
max_time=max_time,
max_blocks=max_blocks,
max_masks=max_masks,
target_image=target_image,
params=params,
)
)
for n, image in enumerate(sample["test"]):
original_image = np.uint8(image["input"])
sample["test"][n].update(
generate_masks(
original_image,
sample["test"][n],
max_time=max_time,
max_blocks=max_blocks,
max_masks=max_masks,
params=params,
)
)
return sample
|