Spaces:
Running
Running
File size: 44,900 Bytes
07a8313 |
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 |
import cv2
import torch
import numpy as np
from skimage import morphology
import albumentations as A
import torch.nn.functional as F
import torch.nn as nn
from albumentations.pytorch import ToTensorV2
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
import matplotlib.path as mplPath
import matplotlib.patches as patches
from ultralyticsplus import YOLO
def image_morpho(mask_prediction):
selem2 = morphology.disk(2)
closed = morphology.closing(mask_prediction, selem2)
return closed
def get_segformer_img(image_in, input_size=[224,224]):
transform_img = A.Compose([
A.Resize(height=input_size[0], width=input_size[1], interpolation=cv2.INTER_NEAREST),
A.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225], max_pixel_value=255.0, p=1.0),
ToTensorV2(p=1.0),
])
image_in = cv2.resize(image_in, (1920, 1080))
image_tr = transform_img(image=image_in)['image']
image_tr = image_tr.unsqueeze(0)
image_tr = image_tr.cpu()
return image_tr, image_in
def load_segformer(path_model):
model = torch.load(path_model, map_location=torch.device('cpu'))
model = model.cpu()
model.eval()
return model
def load_yolo(PATH_model):
model = YOLO(PATH_model)
model.overrides['conf'] = 0.25 # NMS confidence threshold
model.overrides['iou'] = 0.45 # NMS IoU threshold
model.overrides['agnostic_nms'] = False # NMS class-agnostic
model.overrides['max_det'] = 1000 # maximum number of detections per image
return model
def find_extreme_y_values(arr, values=[0, 6]):
"""
Optimized function to find the lowest and highest y-values (row indices) in a 2D array where 0 or 6 appears.
Parameters:
- arr: The input 2D NumPy array.
- values: The values to search for (default is [0, 6]).
Returns:
A tuple (lowest_y, highest_y) representing the lowest and highest y-values. If values are not found, returns None.
"""
mask = np.isin(arr, values)
rows_with_values = np.any(mask, axis=1)
y_indices = np.nonzero(rows_with_values)[0] # Directly finding non-zero (True) indices
if y_indices.size == 0:
return None, None # Early return if values not found
return y_indices[0], y_indices[-1]
def find_nearest_pairs(arr1, arr2):
# Convert lists to numpy arrays for vectorized operations
arr1_np = np.array(arr1)
arr2_np = np.array(arr2)
# Determine which array is shorter
if len(arr1_np) < len(arr2_np):
base_array, compare_array = arr1_np, arr2_np
else:
base_array, compare_array = arr2_np, arr1_np
paired_base = []
paired_compare = []
# Mask to keep track of paired elements
paired_mask = np.zeros(len(compare_array), dtype=bool)
for item in base_array:
# Calculate distances from the current item to all items in the compare_array
distances = np.linalg.norm(compare_array - item, axis=1)
nearest_index = np.argmin(distances)
paired_base.append(item)
paired_compare.append(compare_array[nearest_index])
# Mark the paired element to exclude it from further pairing
paired_mask[nearest_index] = True
# Check if all elements from the compare_array have been paired
if paired_mask.all():
break
paired_base = np.array(paired_base)
paired_compare = compare_array[paired_mask]
return (paired_base, paired_compare) if len(arr1_np) < len(arr2_np) else (paired_compare, paired_base)
def filter_crossings(image, edges_dict):
filtered_edges = {}
for key, values in edges_dict.items():
merged = [values[0]]
for start, end in values[1:]:
if start - merged[-1][1] < 50:
key_up = max([0, key-10])
key_down = min([image.shape[0]-1, key+10])
if key_up == 0:
key_up = key+20
if key_down == image.shape[0]-1:
key_down = key-20
edges_to_test_slope1 = robust_edges(image, [key_up], values=[0, 6], min_width=19)
edges_to_test_slope2 = robust_edges(image, [key_down], values=[0, 6], min_width=19)
values1, edges_to_test_slope1 = find_nearest_pairs(values, edges_to_test_slope1)
values2, edges_to_test_slope2 = find_nearest_pairs(values, edges_to_test_slope2)
differences_y = []
for i, value in enumerate(values1):
if start in value:
idx = list(value).index(start)
try:
differences_y.append(abs(start-edges_to_test_slope1[i][idx]))
except:
pass
if merged[-1][1] in value:
idx = list(value).index(merged[-1][1])
try:
differences_y.append(abs(merged[-1][1]-edges_to_test_slope1[i][idx]))
except:
pass
for i, value in enumerate(values2):
if start in value:
idx = list(value).index(start)
try:
differences_y.append(abs(start-edges_to_test_slope2[i][idx]))
except:
pass
if merged[-1][1] in value:
idx = list(value).index(merged[-1][1])
try:
differences_y.append(abs(merged[-1][1]-edges_to_test_slope2[i][idx]))
except:
pass
if any(element > 30 for element in differences_y):
merged[-1] = (merged[-1][0], end)
else:
merged.append((start, end))
else:
merged.append((start, end))
filtered_edges[key] = merged
return filtered_edges
def robust_edges(image, y_levels, values=[0, 6], min_width=19):
for y in y_levels:
row = image[y, :]
mask = np.isin(row, values).astype(int)
padded_mask = np.pad(mask, (1, 1), 'constant', constant_values=0)
diff = np.diff(padded_mask)
starts = np.where(diff == 1)[0]
ends = np.where(diff == -1)[0] - 1
# Filter sequences based on the minimum width criteria
filtered_edges = [(start, end) for start, end in zip(starts, ends) if end - start + 1 >= min_width]
filtered_edges = [(start, end) for start, end in filtered_edges if 0 not in (start, end) and 1919 not in (start, end)]
return filtered_edges
def find_edges(image, y_levels, values=[0, 6], min_width=19):
"""
Find start and end positions of continuous sequences of specified values at given y-levels in a 2D array,
filtering for sequences that meet or exceed a specified minimum width.
Parameters:
- arr: 2D NumPy array to search within.
- y_levels: List of y-levels (row indices) to examine.
- values: Values to search for (default is [0, 6]).
- min_width: Minimum width of sequences to be included in the results.
Returns:
A dict with y-levels as keys and lists of (start, end) tuples for each sequence found in that row that meets the width criteria.
"""
edges_dict = {}
for y in y_levels:
row = image[y, :]
mask = np.isin(row, values).astype(int)
padded_mask = np.pad(mask, (1, 1), 'constant', constant_values=0)
diff = np.diff(padded_mask)
starts = np.where(diff == 1)[0]
ends = np.where(diff == -1)[0] - 1
# Filter sequences based on the minimum width criteria
filtered_edges = [(start, end) for start, end in zip(starts, ends) if end - start + 1 >= min_width]
filtered_edges = [(start, end) for start, end in filtered_edges if 0 not in (start, end) and 1919 not in (start, end)]
edges_with_guard_rails = []
for edge in filtered_edges:
cutout_left = image[y,edge[0]-50:edge[0]][::-1]
cutout_right = image[y,edge[1]:edge[1]+50]
not_ones = np.where(cutout_left != 1)[0]
if len(not_ones) > 0 and not_ones[0] > 0:
last_one_index = not_ones[0] - 1
edge = (edge[0] - last_one_index,) + edge[1:]
else:
last_one_index = None if len(not_ones) == 0 else not_ones[-1] - 1
not_ones = np.where(cutout_right != 1)[0]
if len(not_ones) > 0 and not_ones[0] > 0:
last_one_index = not_ones[0] - 1
edge = (edge[0], edge[1] - last_one_index) + edge[2:]
else:
last_one_index = None if len(not_ones) == 0 else not_ones[-1] - 1
edges_with_guard_rails.append(edge)
edges_dict[y] = edges_with_guard_rails
edges_dict = {k: v for k, v in edges_dict.items() if v}
edges_dict = filter_crossings(image, edges_dict)
return edges_dict
def find_rails(arr, y_levels, values=[9, 10], min_width=5):
edges_all = []
for y in y_levels:
row = arr[y, :]
mask = np.isin(row, values).astype(int)
padded_mask = np.pad(mask, (1, 1), 'constant', constant_values=0)
diff = np.diff(padded_mask)
starts = np.where(diff == 1)[0]
ends = np.where(diff == -1)[0] - 1
# Filter sequences based on the minimum width criteria
filtered_edges = [(start, end) for start, end in zip(starts, ends) if end - start + 1 >= min_width]
filtered_edges = [(start, end) for start, end in filtered_edges if 0 not in (start, end) and 1919 not in (start, end)]
edges_all = filtered_edges
return edges_all
def mark_edges(arr, edges_dict, mark_value=30):
"""
Marks a 5x5 zone around the edges found in the array with a specific value.
Parameters:
- arr: The original 2D NumPy array.
- edges_dict: A dictionary with y-levels as keys and lists of (start, end) tuples for edges.
- mark_value: The value used to mark the edges.
Returns:
The modified array with marked zones.
"""
marked_arr = np.copy(arr) # Create a copy of the array to avoid modifying the original
offset = 2 # To mark a 5x5 area, we go 2 pixels in each direction from the center
for y, edges in edges_dict.items():
for start, end in edges:
# Mark a 5x5 zone around the start and end positions
for dy in range(-offset, offset + 1):
for dx in range(-offset, offset + 1):
# Check array bounds before marking
if 0 <= y + dy < marked_arr.shape[0] and 0 <= start + dx < marked_arr.shape[1]:
marked_arr[y + dy, start + dx] = mark_value
if 0 <= y + dy < marked_arr.shape[0] and 0 <= end + dx < marked_arr.shape[1]:
marked_arr[y + dy, end + dx] = mark_value
return marked_arr
def find_rail_sides(img, edges_dict):
left_border = []
right_border = []
for y,xs in edges_dict.items():
rails = find_rails(img, [y], values=[9,10], min_width=5)
left_border_actual = [min(xs)[0],y]
right_border_actual = [max(xs)[1],y]
for zone in rails:
if abs(zone[1]-left_border_actual[0]) < y*0.04: # dynamic treshold
left_border_actual[0] = zone[0]
if abs(zone[0]-right_border_actual[0]) < y*0.04:
right_border_actual[0] = zone[1]
left_border.append(left_border_actual)
right_border.append(right_border_actual)
# removing detected uncontioussness
left_border, flags_l, _ = robust_rail_sides(left_border) # filter outliers
right_border, flags_r, _ = robust_rail_sides(right_border)
return left_border, right_border, flags_l, flags_r
def robust_rail_sides(border, threshold=7):
border = np.array(border)
if border.size > 0:
# delete borders found on the bottom side of the image
border = border[border[:, 1] != 1079]
steps_x = np.diff(border[:, 0])
median_step = np.median(np.abs(steps_x))
threshold_step = np.abs(threshold*np.abs(median_step))
treshold_overcommings = abs(steps_x) > abs(threshold_step)
flags = []
if True not in treshold_overcommings:
return border, flags, []
else:
overcommings_indices = [i for i, element in enumerate(treshold_overcommings) if element == True]
if overcommings_indices and np.all(np.diff(overcommings_indices) == 1):
overcommings_indices = [overcommings_indices[0]]
filtered_border = border
previously_deleted = []
for i in overcommings_indices:
for item in previously_deleted:
if item[0] < i:
i -= item[1]
first_part = filtered_border[:i+1]
second_part = filtered_border[i+1:]
if len(second_part)<2:
filtered_border = first_part
previously_deleted.append([i,len(second_part)])
elif len(first_part)<2:
filtered_border = second_part
previously_deleted.append([i,len(first_part)])
else:
first_b, _, deleted_first = robust_rail_sides(first_part)
second_b, _, _ = robust_rail_sides(second_part)
filtered_border = np.concatenate((first_b,second_b), axis=0)
if deleted_first:
for deleted_item in deleted_first:
if deleted_item[0]<=i:
i -= deleted_item[1]
flags.append(i)
return filtered_border, flags, previously_deleted
else:
return border, [], []
def find_dist_from_edges(id_map, image, edges_dict, left_border, right_border, real_life_width_mm, real_life_target_mm, mark_value=30):
"""
Mark regions representing a real-life distance (e.g., 2 meters) to the left and right from the furthest edges.
Parameters:
- arr: 2D NumPy array representing the id_map.
- edges_dict: Dictionary with y-levels as keys and lists of (start, end) tuples for edges.
- real_life_width_mm: The real-world width in millimeters that the average sequence width represents.
- real_life_target_mm: The real-world distance in millimeters to mark from the edges.
Returns:
- A NumPy array with the marked regions.
"""
# Calculate the rail widths
diffs_widths = {k: sum(e-s for s, e in v) / len(v) for k, v in edges_dict.items() if v}
diffs_width = {k: max(e-s for s, e in v) for k, v in edges_dict.items() if v}
# Pixel to mm scale factor
scale_factors = {k: real_life_width_mm / v for k, v in diffs_width.items()}
# Converting the real-life target distance to pixels
target_distances_px = {k: int(real_life_target_mm / v) for k, v in scale_factors.items()}
# Mark the regions representing the target distance to the left and right from the furthest edges
end_points_left = {}
region_levels_left = []
for point in left_border:
min_edge = point[0]
# Ensure we stay within the image bounds
#left_mark_start = max(0, min_edge - int(target_distances_px[point[1]]))
left_mark_start = min_edge - int(target_distances_px[point[1]])
end_points_left[point[1]] = left_mark_start
# Left region points
if left_mark_start < min_edge:
y_values = np.arange(left_mark_start, min_edge)
x_values = np.full_like(y_values, point[1])
region_line = np.column_stack((x_values, y_values))
region_levels_left.append(region_line)
end_points_right = {}
region_levels_right = []
for point in right_border:
max_edge = point[0]
# Ensure we stay within the image bounds
right_mark_end = min(id_map.shape[1], max_edge + int(target_distances_px[point[1]]))
if right_mark_end != id_map.shape[1]:
end_points_right[point[1]] = right_mark_end
# Right region points
if max_edge < right_mark_end:
y_values = np.arange(max_edge, right_mark_end)
x_values = np.full_like(y_values, point[1])
region_line = np.column_stack((x_values, y_values))
region_levels_right.append(region_line)
return id_map, end_points_left, end_points_right, region_levels_left, region_levels_right
def bresenham_line(x0, y0, x1, y1):
"""
Generate the coordinates of a line from (x0, y0) to (x1, y1) using Bresenham's algorithm.
"""
line = []
dx = abs(x1 - x0)
dy = -abs(y1 - y0)
sx = 1 if x0 < x1 else -1
sy = 1 if y0 < y1 else -1
err = dx + dy # error value e_xy
while True:
line.append((x0, y0)) # Add the current point to the line
if x0 == x1 and y0 == y1:
break
e2 = 2 * err
if e2 >= dy: # e_xy+e_x > 0
err += dy
x0 += sx
if e2 <= dx: # e_xy+e_y < 0
err += dx
y0 += sy
return line
def interpolate_end_points(end_points_dict, flags):
line_arr = []
ys = list(end_points_dict.keys())
xs = list(end_points_dict.values())
if flags and len(flags) == 1:
pass
elif flags and np.all(np.diff(flags) == 1):
flags = [flags[0]]
for i in range(0, len(ys) - 1):
if i in flags:
continue
y1, y2 = ys[i], ys[i + 1]
x1, x2 = xs[i], xs[i + 1]
line = np.array(bresenham_line(x1, y1, x2, y2))
if np.any(line[:, 0] < 0):
line = line[line[:, 0] > 0]
line_arr = line_arr + list(line)
return line_arr
def extrapolate_line(pixels, image, min_y=None, extr_pixels=10):
"""
Extrapolate a line based on the last segment using linear regression.
Parameters:
- pixels: List of (x, y) tuples representing line pixel coordinates.
- image: 2D numpy array representing the image.
- min_y: Minimum y-value to extrapolate to (optional).
Returns:
- A list of new extrapolated (x, y) pixel coordinates.
"""
if len(pixels) < extr_pixels:
print("Not enough pixels to perform extrapolation.")
return []
recent_pixels = np.array(pixels[-extr_pixels:])
X = recent_pixels[:, 0].reshape(-1, 1) # Reshape for sklearn
y = recent_pixels[:, 1]
model = LinearRegression()
model.fit(X, y)
slope = model.coef_[0]
intercept = model.intercept_
extrapolate = lambda x: slope * x + intercept
# Calculate direction based on last two pixels
dx, dy = 0, 0 # Default values
x_diffs = []
y_diffs = []
for i in range(1,extr_pixels-1):
x_diffs.append(pixels[-i][0] - pixels[-(i+1)][0])
y_diffs.append(pixels[-i][1] - pixels[-(i+1)][1])
x_diff = x_diffs[np.argmax(np.abs(x_diffs))]
y_diff = y_diffs[np.argmax(np.abs(y_diffs))]
if abs(int(x_diff)) >= abs(int(y_diff)):
dx = 1 if x_diff >= 0 else -1
else:
dy = 1 if y_diff >= 0 else -1
last_pixel = pixels[-1]
new_pixels = []
x, y = last_pixel
min_y = min_y if min_y is not None else image.shape[0] - 1
while 0 <= x < image.shape[1] and min_y <= y < image.shape[0]:
if dx != 0: # Horizontal or diagonal movement
x += dx
y = int(extrapolate(x))
elif dy != 0: # Vertical movement
y += dy
# For vertical lines, approximate x based on the last known value
x = int(x)
if 0 <= y < image.shape[0] and 0 <= x < image.shape[1]:
new_pixels.append((x, y))
else:
break
return new_pixels
def extrapolate_borders(dist_marked_id_map, border_l, border_r, lowest_y):
#border_extrapolation_l1 = extrapolate_line(border_l, dist_marked_id_map, lowest_y)
border_extrapolation_l2 = extrapolate_line(border_l[::-1], dist_marked_id_map, lowest_y)
#border_extrapolation_r1 = extrapolate_line(border_r, dist_marked_id_map, lowest_y)
border_extrapolation_r2 = extrapolate_line(border_r[::-1], dist_marked_id_map, lowest_y)
#border_l = border_extrapolation_l2[::-1] + border_l + border_extrapolation_l1
#border_r = border_extrapolation_r2[::-1] + border_r + border_extrapolation_r1
border_l = border_extrapolation_l2[::-1] + border_l
border_r = border_extrapolation_r2[::-1] + border_r
return border_l, border_r
def find_zone_border(id_map, image, edges, irl_width_mm=1435, irl_target_mm=1000, lowest_y = 0):
left_border, right_border, flags_l, flags_r = find_rail_sides(id_map, edges)
dist_marked_id_map, end_points_left, end_points_right, left_region, right_region = find_dist_from_edges(id_map, image, edges, left_border, right_border, irl_width_mm, irl_target_mm)
border_l = interpolate_end_points(end_points_left, flags_l)
border_r = interpolate_end_points(end_points_right, flags_r)
border_l, border_r = extrapolate_borders(dist_marked_id_map, border_l, border_r, lowest_y)
return [border_l, border_r],[left_region, right_region]
def get_clues(segmentation_mask, number_of_clues):
lowest, highest = find_extreme_y_values(segmentation_mask)
if lowest is not None and highest is not None:
clue_step = int((highest - lowest) / number_of_clues+1)
clues = []
for i in range(number_of_clues):
clues.append(highest - (i*clue_step))
clues.append(lowest+int(0.5*clue_step))
return clues
else:
return []
def border_handler(id_map, image, edges, target_distances):
lowest, _ = find_extreme_y_values(id_map)
borders = []
regions = []
for target in target_distances:
borders_regions = find_zone_border(id_map, image, edges, irl_target_mm=target, lowest_y = lowest)
borders.append(borders_regions[0])
regions.append(borders_regions[1])
return borders, id_map, regions
def segment(input_image, model_seg, image_size):
image_norm, image = get_segformer_img(input_image, image_size)
outputs = model_seg(image_norm)
logits = outputs.logits
upsampled_logits = nn.functional.interpolate(
logits,
size=image_norm.shape[-2:],
mode="bilinear",
align_corners=False
)
output = upsampled_logits.float()
confidence_scores = F.softmax(output, dim=1).cpu().detach().numpy().squeeze()
id_map = np.argmax(confidence_scores, axis=0).astype(np.uint8)
id_map = image_morpho(id_map)
id_map = cv2.resize(id_map, [1920,1080], interpolation=cv2.INTER_NEAREST)
return id_map, image
def detect(model_det, image):
results = model_det.predict(image)
return results, model_det, image
def manage_detections(results, model):
bbox = results[0].boxes.xywh.tolist()
cls = results[0].boxes.cls.tolist()
accepted_stationary = np.array([24,25,28,36])
accepted_moving = np.array([0,1,2,3,7,15,16,17,18,19])
boxes_moving = {}
boxes_stationary = {}
if len(bbox) > 0:
for xywh, clss in zip(bbox, cls):
if clss in accepted_moving:
if clss in boxes_moving.keys() and len(boxes_moving[clss]) > 0:
boxes_moving[clss].append(xywh)
else:
boxes_moving[clss] = [xywh]
if clss in accepted_stationary:
if clss in boxes_stationary.keys() and len(boxes_stationary[clss]) > 0:
boxes_stationary[clss].append(xywh)
else:
boxes_stationary[clss] = [xywh]
return boxes_moving, boxes_stationary
def compute_detection_borders(borders, output_dims=[1080,1920]):
det_height = output_dims[0]-1
det_width = output_dims[1]-1
for i,border in enumerate(borders):
border_l = np.array(border[0])
if list(border_l):
pass
else:
border_l=np.array([[0,0],[0,0]])
endpoints_l = [border_l[0],border_l[-1]]
border_r = np.array(border[1])
if list(border_r):
pass
else:
border_r=np.array([[0,0],[0,0]])
endpoints_r = [border_r[0],border_r[-1]]
if np.array_equal(np.array([[0,0],[0,0]]), endpoints_l):
endpoints_l = [[0,endpoints_r[0][1]],[0,endpoints_r[1][1]]]
if np.array_equal(np.array([[0,0],[0,0]]), endpoints_r):
endpoints_r = [[det_width,endpoints_l[0][1]],[det_width,endpoints_l[1][1]]]
interpolated_top = bresenham_line(endpoints_l[1][0],endpoints_l[1][1],endpoints_r[1][0],endpoints_r[1][1])
zero_range = [0,1,2,3]
height_range = [det_height,det_height-1,det_height-2,det_height-3]
width_range = [det_width,det_width-1,det_width-2,det_width-3]
if (endpoints_l[0][0] in zero_range and endpoints_r[0][1] in height_range):
y_values = np.arange(endpoints_l[0][1], det_height)
x_values = np.full_like(y_values, 0)
bottom1 = np.column_stack((x_values, y_values))
x_values = np.arange(0, endpoints_r[0][0])
y_values = np.full_like(x_values, det_height)
bottom2 = np.column_stack((x_values, y_values))
interpolated_bottom = np.vstack((bottom1, bottom2))
elif (endpoints_l[0][1] in height_range and endpoints_r[0][0] in width_range):
y_values = np.arange(endpoints_r[0][1], det_height)
x_values = np.full_like(y_values, det_width)
bottom1 = np.column_stack((x_values, y_values))
x_values = np.arange(endpoints_l[0][0], det_width)
y_values = np.full_like(x_values, det_height)
bottom2 = np.column_stack((x_values, y_values))
interpolated_bottom = np.vstack((bottom1, bottom2))
elif endpoints_l[0][0] in zero_range and endpoints_r[0][0] in width_range:
y_values = np.arange(endpoints_l[0][1], det_height)
x_values = np.full_like(y_values, 0)
bottom1 = np.column_stack((x_values, y_values))
y_values = np.arange(endpoints_r[0][1], det_height)
x_values = np.full_like(y_values, det_width)
bottom2 = np.column_stack((x_values, y_values))
bottom3_mid = bresenham_line(bottom1[-1][0],bottom1[-1][1],bottom2[-1][0],bottom2[-1][1])
interpolated_bottom = np.vstack((bottom1, bottom2, bottom3_mid))
else:
interpolated_bottom = bresenham_line(endpoints_l[0][0],endpoints_l[0][1],endpoints_r[0][0],endpoints_r[0][1])
borders[i].append(interpolated_bottom)
borders[i].append(interpolated_top)
return borders
def get_bounding_box_points(cx, cy, w, h):
top_left = (cx - w / 2, cy - h / 2)
top_right = (cx + w / 2, cy - h / 2)
bottom_right = (cx + w / 2, cy + h / 2)
bottom_left = (cx - w / 2, cy + h / 2)
corners = [top_left, top_right, bottom_right, bottom_left]
def interpolate(point1, point2, fraction):
"""Interpolate between two points at a given fraction of the distance."""
return (point1[0] + fraction * (point2[0] - point1[0]),
point1[1] + fraction * (point2[1] - point1[1]))
points = []
for i in range(4):
next_i = (i + 1) % 4
points.append(corners[i])
points.append(interpolate(corners[i], corners[next_i], 1 / 3))
points.append(interpolate(corners[i], corners[next_i], 2 / 3))
return points
def classify_detections(boxes_moving, boxes_stationary, borders, img_dims, output_dims=[1080,1920]):
img_h, img_w, _ = img_dims
img_h_scaletofullHD = output_dims[1]/img_w
img_w_scaletofullHD = output_dims[0]/img_h
colors = ["yellow","orange","red","green","blue"]
borders = compute_detection_borders(borders,output_dims)
boxes_info = []
if boxes_moving or boxes_stationary:
if boxes_moving:
for item, coords in boxes_moving.items():
for coord in coords:
x = coord[0]*img_w_scaletofullHD
y = coord[1]*img_h_scaletofullHD
w = coord[2]*img_w_scaletofullHD
h = coord[3]*img_h_scaletofullHD
points_to_test = get_bounding_box_points(x, y, w, h)
complete_border = []
criticality = -1
color = None
for i,border in enumerate(reversed(borders)):
border_nonempty = [np.array(arr) for arr in border if np.array(arr).size > 0]
complete_border = np.vstack((border_nonempty))
instance_border_path = mplPath.Path(np.array(complete_border))
is_inside_borders = False
for point in points_to_test:
is_inside = instance_border_path.contains_point(point)
if is_inside:
is_inside_borders = True
if is_inside_borders:
criticality = i
color = colors[i]
if criticality == -1:
color = colors[3]
boxes_info.append([item, criticality, color, [x, y], [w, h], 1])
if boxes_stationary:
for item, coords in boxes_stationary.items():
for coord in coords:
x = coord[0]*img_w_scaletofullHD
y = coord[1]*img_h_scaletofullHD
w = coord[2]*img_w_scaletofullHD
h = coord[3]*img_h_scaletofullHD
points_to_test = get_bounding_box_points(x, y, w, h)
complete_border = []
criticality = -1
color = None
is_inside_borders = 0
for i,border in enumerate(reversed(borders), start=len(borders) - 1):
border_nonempty = [np.array(arr) for arr in border if np.array(arr).size > 0]
complete_border = np.vstack(border_nonempty)
instance_border_path = mplPath.Path(np.array(complete_border))
is_inside_borders = False
for point in points_to_test:
is_inside = instance_border_path.contains_point(point)
if is_inside:
is_inside_borders = True
if is_inside_borders:
criticality = i
color = colors[4]
if criticality == -1:
color = colors[3]
boxes_info.append([item, criticality, color, [x, y], [w, h], 0])
return boxes_info
else:
print("No accepted detections in this image.")
return []
def draw_classification(classification, id_map):
if classification:
for box in classification:
x,y = box[3]
mark_value = 30
x_start = int(max(x - 2, 0))
x_end = int(min(x + 3, id_map.shape[1]))
y_start = int(max(y - 2, 0))
y_end = int(min(y + 3, id_map.shape[0]))
id_map[y_start:y_end, x_start:x_end] = mark_value
else:
return
def get_result(classification, id_map, names, borders, image, regions):
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image = cv2.resize(image, (id_map.shape[1], id_map.shape[0]), interpolation = cv2.INTER_LINEAR)
fig = plt.figure(figsize=(16, 9), dpi=100)
plt.imshow(image, cmap='gray')
if classification:
for box in classification:
boxes = True
cx,cy = box[3]
name = names[box[0]]
if boxes:
w,h = box[4]
x = cx - w / 2
y = cy - h / 2
rect = patches.Rectangle((x, y), w, h, linewidth=2, edgecolor=box[2], facecolor='none')
ax = plt.gca()
ax.add_patch(rect)
plt.text(x, y-17, name, color='black', fontsize=10, ha='center', va='center', fontweight='bold', bbox=dict(facecolor=box[2], edgecolor='none', alpha=1))
else:
plt.imshow(id_map, cmap='gray')
plt.text(cx, cy+10, name, color=box[2], fontsize=10, ha='center', va='center', fontweight='bold')
for region in regions:
for side in region:
for line in side:
line = np.array(line)
plt.plot(line[:,1], line[:,0] ,'-', color='lightgrey', marker=None, linewidth=0.5)
plt.ylim(0, 1080)
plt.xlim(0, 1920)
plt.gca().invert_yaxis()
colors = ['yellow','orange','red']
borders.reverse()
for i,border in enumerate(borders):
for side in border:
side = np.array(side)
if side.size > 0:
plt.plot(side[:,0],side[:,1] ,'-', color=colors[i], marker=None, linewidth=0.6) #color=colors[i]
plt.ylim(0, 1080)
plt.xlim(0, 1920)
plt.gca().invert_yaxis()
plt.tight_layout()
canvas = FigureCanvas(fig)
canvas.draw()
width, height = fig.get_size_inches() * fig.get_dpi()
image = np.frombuffer(canvas.tostring_rgb(), dtype='uint8').reshape(int(height), int(width), 3)
plt.close(fig) # Close the figure to free memory
return image
def run(input_image, model_seg, model_det, image_size, target_distances, num_ys = 10):
segmentation_mask, image = segment(input_image, model_seg, image_size)
# Border search
clues = get_clues(segmentation_mask, num_ys)
edges = find_edges(segmentation_mask, clues, min_width=0)
borders, id_map, regions = border_handler(segmentation_mask, image, edges, target_distances)
# Detection
results, model, image = detect(model_det, input_image)
boxes_moving, boxes_stationary = manage_detections(results, model)
classification = classify_detections(boxes_moving, boxes_stationary, borders, image.shape, output_dims=segmentation_mask.shape)
output_image = get_result(classification, id_map, model.names, borders, image, regions)
cropped_image = output_image[22:output_image.shape[0] - 40, 74:output_image.shape[1] - 33]
return cropped_image
if __name__ == "__main__":
image_size = [1024,1024]
target_distances = [650,1000,2000]
num_ys = 10
PATH_model_seg = 'SegFormer_B3_1024_finetuned.pth'
PATH_model_det = 'yolov8s.pt'
input_image = cv2.imread('rs00006.jpg') #TO CO VLOZI UZIVATEL
model_seg = load_segformer(PATH_model_seg)
model_det = load_yolo(PATH_model_det)
image = run(input_image, model_seg, model_det, image_size, target_distances, num_ys=num_ys)
|