Spaces:
Sleeping
Sleeping
File size: 16,500 Bytes
9bf4bd7 |
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 |
# Copyright (c) OpenMMLab. All rights reserved.
import math
import operator
from functools import reduce
from typing import List, Optional, Sequence, Tuple, Union
import numpy as np
import pyclipper
import shapely
from mmengine.utils import is_list_of
from shapely.geometry import MultiPolygon, Polygon
from mmocr.utils import bbox2poly, valid_boundary
from mmocr.utils.check_argument import is_2dlist
from mmocr.utils.typing_utils import ArrayLike
def rescale_polygon(polygon: ArrayLike,
scale_factor: Tuple[int, int],
mode: str = 'mul') -> np.ndarray:
"""Rescale a polygon according to scale_factor.
The behavior is different depending on the mode. When mode is 'mul', the
coordinates will be multiplied by scale_factor, which is usually used in
preprocessing transforms such as :func:`Resize`.
The coordinates will be divided by scale_factor if mode is 'div'. It can be
used in postprocessors to recover the polygon in the original
image size.
Args:
polygon (ArrayLike): A polygon. In any form can be converted
to an 1-D numpy array. E.g. list[float], np.ndarray,
or torch.Tensor. Polygon is written in
[x1, y1, x2, y2, ...].
scale_factor (tuple(int, int)): (w_scale, h_scale).
model (str): Rescale mode. Can be 'mul' or 'div'. Defaults to 'mul'.
Returns:
np.ndarray: Rescaled polygon.
"""
assert len(polygon) % 2 == 0
assert mode in ['mul', 'div']
polygon = np.array(polygon, dtype=np.float32)
poly_shape = polygon.shape
reshape_polygon = polygon.reshape(-1, 2)
scale_factor = np.array(scale_factor, dtype=float)
if mode == 'div':
scale_factor = 1 / scale_factor
polygon = (reshape_polygon * scale_factor[None]).reshape(poly_shape)
return polygon
def rescale_polygons(polygons: Union[ArrayLike, Sequence[ArrayLike]],
scale_factor: Tuple[int, int],
mode: str = 'mul'
) -> Union[ArrayLike, Sequence[np.ndarray]]:
"""Rescale polygons according to scale_factor.
The behavior is different depending on the mode. When mode is 'mul', the
coordinates will be multiplied by scale_factor, which is usually used in
preprocessing transforms such as :func:`Resize`.
The coordinates will be divided by scale_factor if mode is 'div'. It can be
used in postprocessors to recover the polygon in the original
image size.
Args:
polygons (list[ArrayLike] or ArrayLike): A list of polygons, each
written in [x1, y1, x2, y2, ...] and in any form can be converted
to an 1-D numpy array. E.g. list[list[float]],
list[np.ndarray], or list[torch.Tensor].
scale_factor (tuple(int, int)): (w_scale, h_scale).
model (str): Rescale mode. Can be 'mul' or 'div'. Defaults to 'mul'.
Returns:
list[np.ndarray] or np.ndarray: Rescaled polygons. The type of the
return value depends on the type of the input polygons.
"""
results = []
for polygon in polygons:
results.append(rescale_polygon(polygon, scale_factor, mode))
if isinstance(polygons, np.ndarray):
results = np.array(results)
return results
def poly2bbox(polygon: ArrayLike) -> np.array:
"""Converting a polygon to a bounding box.
Args:
polygon (ArrayLike): A polygon. In any form can be converted
to an 1-D numpy array. E.g. list[float], np.ndarray,
or torch.Tensor. Polygon is written in
[x1, y1, x2, y2, ...].
Returns:
np.array: The converted bounding box [x1, y1, x2, y2]
"""
assert len(polygon) % 2 == 0
polygon = np.array(polygon, dtype=np.float32)
x = polygon[::2]
y = polygon[1::2]
return np.array([min(x), min(y), max(x), max(y)])
def poly2shapely(polygon: ArrayLike) -> Polygon:
"""Convert a polygon to shapely.geometry.Polygon.
Args:
polygon (ArrayLike): A set of points of 2k shape.
Returns:
polygon (Polygon): A polygon object.
"""
polygon = np.array(polygon, dtype=np.float32)
assert polygon.size % 2 == 0 and polygon.size >= 6
polygon = polygon.reshape([-1, 2])
return Polygon(polygon)
def polys2shapely(polygons: Sequence[ArrayLike]) -> Sequence[Polygon]:
"""Convert a nested list of boundaries to a list of Polygons.
Args:
polygons (list): The point coordinates of the instance boundary.
Returns:
list: Converted shapely.Polygon.
"""
return [poly2shapely(polygon) for polygon in polygons]
def shapely2poly(polygon: Polygon) -> np.array:
"""Convert a nested list of boundaries to a list of Polygons.
Args:
polygon (Polygon): A polygon represented by shapely.Polygon.
Returns:
np.array: Converted numpy array
"""
return np.array(polygon.exterior.coords).reshape(-1, )
def crop_polygon(polygon: ArrayLike,
crop_box: np.ndarray) -> Union[np.ndarray, None]:
"""Crop polygon to be within a box region.
Args:
polygon (ndarray): polygon in shape (N, ).
crop_box (ndarray): target box region in shape (4, ).
Returns:
np.array or None: Cropped polygon. If the polygon is not within the
crop box, return None.
"""
poly = poly_make_valid(poly2shapely(polygon))
crop_poly = poly_make_valid(poly2shapely(bbox2poly(crop_box)))
area, poly_cropped = poly_intersection(poly, crop_poly, return_poly=True)
if area == 0 or area is None or not isinstance(
poly_cropped, shapely.geometry.polygon.Polygon):
return None
else:
poly_cropped = poly_make_valid(poly_cropped)
poly_cropped = np.array(poly_cropped.boundary.xy, dtype=np.float32)
poly_cropped = poly_cropped.T
# reverse poly_cropped to have clockwise order
poly_cropped = poly_cropped[::-1, :].reshape(-1)
return poly_cropped
def poly_make_valid(poly: Polygon) -> Polygon:
"""Convert a potentially invalid polygon to a valid one by eliminating
self-crossing or self-touching parts. Note that if the input is a line, the
returned polygon could be an empty one.
Args:
poly (Polygon): A polygon needed to be converted.
Returns:
Polygon: A valid polygon, which might be empty.
"""
assert isinstance(poly, Polygon)
fixed_poly = poly if poly.is_valid else poly.buffer(0)
# Sometimes the fixed_poly is still a MultiPolygon,
# so we need to find the convex hull of the MultiPolygon, which should
# always be a Polygon (but could be empty).
if not isinstance(fixed_poly, Polygon):
fixed_poly = fixed_poly.convex_hull
return fixed_poly
def poly_intersection(poly_a: Polygon,
poly_b: Polygon,
invalid_ret: Optional[Union[float, int]] = None,
return_poly: bool = False
) -> Tuple[float, Optional[Polygon]]:
"""Calculate the intersection area between two polygons.
Args:
poly_a (Polygon): Polygon a.
poly_b (Polygon): Polygon b.
invalid_ret (float or int, optional): The return value when the
invalid polygon exists. If it is not specified, the function
allows the computation to proceed with invalid polygons by
cleaning the their self-touching or self-crossing parts.
Defaults to None.
return_poly (bool): Whether to return the polygon of the intersection
Defaults to False.
Returns:
float or tuple(float, Polygon): Returns the intersection area or
a tuple ``(area, Optional[poly_obj])``, where the `area` is the
intersection area between two polygons and `poly_obj` is The Polygon
object of the intersection area, which will be `None` if the input is
invalid. `poly_obj` will be returned only if `return_poly` is `True`.
"""
assert isinstance(poly_a, Polygon)
assert isinstance(poly_b, Polygon)
assert invalid_ret is None or isinstance(invalid_ret, (float, int))
if invalid_ret is None:
poly_a = poly_make_valid(poly_a)
poly_b = poly_make_valid(poly_b)
poly_obj = None
area = invalid_ret
if poly_a.is_valid and poly_b.is_valid:
if poly_a.intersects(poly_b):
poly_obj = poly_a.intersection(poly_b)
area = poly_obj.area
else:
poly_obj = Polygon()
area = 0.0
return (area, poly_obj) if return_poly else area
def poly_union(
poly_a: Polygon,
poly_b: Polygon,
invalid_ret: Optional[Union[float, int]] = None,
return_poly: bool = False
) -> Tuple[float, Optional[Union[Polygon, MultiPolygon]]]:
"""Calculate the union area between two polygons.
Args:
poly_a (Polygon): Polygon a.
poly_b (Polygon): Polygon b.
invalid_ret (float or int, optional): The return value when the
invalid polygon exists. If it is not specified, the function
allows the computation to proceed with invalid polygons by
cleaning the their self-touching or self-crossing parts.
Defaults to False.
return_poly (bool): Whether to return the polygon of the union.
Defaults to False.
Returns:
tuple: Returns a tuple ``(area, Optional[poly_obj])``, where
the `area` is the union between two polygons and `poly_obj` is the
Polygon or MultiPolygon object of the union of the inputs. The type
of object depends on whether they intersect or not. Set as `None`
if the input is invalid. `poly_obj` will be returned only if
`return_poly` is `True`.
"""
assert isinstance(poly_a, Polygon)
assert isinstance(poly_b, Polygon)
assert invalid_ret is None or isinstance(invalid_ret, (float, int))
if invalid_ret is None:
poly_a = poly_make_valid(poly_a)
poly_b = poly_make_valid(poly_b)
poly_obj = None
area = invalid_ret
if poly_a.is_valid and poly_b.is_valid:
poly_obj = poly_a.union(poly_b)
area = poly_obj.area
return (area, poly_obj) if return_poly else area
def poly_iou(poly_a: Polygon,
poly_b: Polygon,
zero_division: float = 0.) -> float:
"""Calculate the IOU between two polygons.
Args:
poly_a (Polygon): Polygon a.
poly_b (Polygon): Polygon b.
zero_division (float): The return value when invalid polygon exists.
Returns:
float: The IoU between two polygons.
"""
assert isinstance(poly_a, Polygon)
assert isinstance(poly_b, Polygon)
area_inters = poly_intersection(poly_a, poly_b)
area_union = poly_union(poly_a, poly_b)
return area_inters / area_union if area_union != 0 else zero_division
def is_poly_inside_rect(poly: ArrayLike, rect: np.ndarray) -> bool:
"""Check if the polygon is inside the target region.
Args:
poly (ArrayLike): Polygon in shape (N, ).
rect (ndarray): Target region [x1, y1, x2, y2].
Returns:
bool: Whether the polygon is inside the cropping region.
"""
poly = poly2shapely(poly)
rect = poly2shapely(bbox2poly(rect))
return rect.contains(poly)
def offset_polygon(poly: ArrayLike, distance: float) -> ArrayLike:
"""Offset (expand/shrink) the polygon by the target distance. It's a
wrapper around pyclipper based on Vatti clipping algorithm.
Warning:
Polygon coordinates will be casted to int type in PyClipper. Mind the
potential precision loss caused by the casting.
Args:
poly (ArrayLike): A polygon. In any form can be converted
to an 1-D numpy array. E.g. list[float], np.ndarray,
or torch.Tensor. Polygon is written in
[x1, y1, x2, y2, ...].
distance (float): The offset distance. Positive value means expanding,
negative value means shrinking.
Returns:
np.array: 1-D Offsetted polygon ndarray in float32 type. If the
result polygon is invalid or has been split into several parts,
return an empty array.
"""
poly = np.array(poly).reshape(-1, 2)
pco = pyclipper.PyclipperOffset()
pco.AddPath(poly, pyclipper.JT_ROUND, pyclipper.ET_CLOSEDPOLYGON)
# Returned result will be in type of int32, convert it back to float32
# following MMOCR's convention
result = np.array(pco.Execute(distance), dtype=object)
if len(result) > 0 and isinstance(result[0], list):
# The processed polygon has been split into several parts
result = np.array([])
result = result.astype(np.float32)
# Always use the first polygon since only one polygon is expected
# But when the resulting polygon is invalid, return the empty array
# as it is
return result if len(result) == 0 else result[0].flatten()
def boundary_iou(src: List,
target: List,
zero_division: Union[int, float] = 0) -> float:
"""Calculate the IOU between two boundaries.
Args:
src (list): Source boundary.
target (list): Target boundary.
zero_division (int or float): The return value when invalid
boundary exists.
Returns:
float: The iou between two boundaries.
"""
assert valid_boundary(src, False)
assert valid_boundary(target, False)
src_poly = poly2shapely(src)
target_poly = poly2shapely(target)
return poly_iou(src_poly, target_poly, zero_division=zero_division)
def sort_points(points):
# TODO Add typehints & test & docstring
"""Sort arbitrary points in clockwise order in Cartesian coordinate, you
may need to reverse the output sequence if you are using OpenCV's image
coordinate.
Reference:
https://github.com/novioleo/Savior/blob/master/Utils/GeometryUtils.py.
Warning: This function can only sort convex polygons.
Args:
points (list[ndarray] or ndarray or list[list]): A list of unsorted
boundary points.
Returns:
list[ndarray]: A list of points sorted in clockwise order.
"""
assert is_list_of(points, np.ndarray) or isinstance(points, np.ndarray) \
or is_2dlist(points)
center_point = tuple(
map(operator.truediv,
reduce(lambda x, y: map(operator.add, x, y), points),
[len(points)] * 2))
return sorted(
points,
key=lambda coord: (180 + math.degrees(
math.atan2(*tuple(map(operator.sub, coord, center_point))))) % 360)
def sort_vertex(points_x, points_y):
# TODO Add typehints & test
"""Sort box vertices in clockwise order from left-top first.
Args:
points_x (list[float]): x of four vertices.
points_y (list[float]): y of four vertices.
Returns:
tuple[list[float], list[float]]: Sorted x and y of four vertices.
- sorted_points_x (list[float]): x of sorted four vertices.
- sorted_points_y (list[float]): y of sorted four vertices.
"""
assert is_list_of(points_x, (float, int))
assert is_list_of(points_y, (float, int))
assert len(points_x) == 4
assert len(points_y) == 4
vertices = np.stack((points_x, points_y), axis=-1).astype(np.float32)
vertices = _sort_vertex(vertices)
sorted_points_x = list(vertices[:, 0])
sorted_points_y = list(vertices[:, 1])
return sorted_points_x, sorted_points_y
def _sort_vertex(vertices):
# TODO Add typehints & docstring & test
assert vertices.ndim == 2
assert vertices.shape[-1] == 2
N = vertices.shape[0]
if N == 0:
return vertices
center = np.mean(vertices, axis=0)
directions = vertices - center
angles = np.arctan2(directions[:, 1], directions[:, 0])
sort_idx = np.argsort(angles)
vertices = vertices[sort_idx]
left_top = np.min(vertices, axis=0)
dists = np.linalg.norm(left_top - vertices, axis=-1, ord=2)
lefttop_idx = np.argmin(dists)
indexes = (np.arange(N, dtype=np.int_) + lefttop_idx) % N
return vertices[indexes]
def sort_vertex8(points):
# TODO Add typehints & docstring & test
"""Sort vertex with 8 points [x1 y1 x2 y2 x3 y3 x4 y4]"""
assert len(points) == 8
vertices = _sort_vertex(np.array(points, dtype=np.float32).reshape(-1, 2))
sorted_box = list(vertices.flatten())
return sorted_box
|