File size: 8,747 Bytes
91ef820 |
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 |
# -*- coding: utf-8 -*-
"""
Generic Image Transform utillities.
"""
import cv2
import random, math
import numpy as np
from collections import Iterable
import torch.nn.functional as F
from torch.autograd import Variable
class ResizePad:
"""
Resize and pad an image to given size.
"""
def __init__(self, size):
if not isinstance(size, (int, Iterable)):
raise TypeError('Got inappropriate size arg: {}'.format(size))
self.h, self.w = size
def __call__(self, img):
h, w = img.shape[:2]
scale = min(self.h / h, self.w / w)
resized_h = int(np.round(h * scale))
resized_w = int(np.round(w * scale))
pad_h = int(np.floor(self.h - resized_h) / 2)
pad_w = int(np.floor(self.w - resized_w) / 2)
resized_img = cv2.resize(img, (resized_w, resized_h))
# if img.ndim > 2:
if img.ndim > 2:
new_img = np.zeros(
(self.h, self.w, img.shape[-1]), dtype=resized_img.dtype)
else:
resized_img = np.expand_dims(resized_img, -1)
new_img = np.zeros((self.h, self.w, 1), dtype=resized_img.dtype)
new_img[pad_h: pad_h + resized_h,
pad_w: pad_w + resized_w, ...] = resized_img
return new_img
class CropResize:
"""Remove padding and resize image to its original size."""
def __call__(self, img, size):
if not isinstance(size, (int, Iterable)):
raise TypeError('Got inappropriate size arg: {}'.format(size))
im_h, im_w = img.data.shape[:2]
input_h, input_w = size
scale = max(input_h / im_h, input_w / im_w)
# scale = torch.Tensor([[input_h / im_h, input_w / im_w]]).max()
resized_h = int(np.round(im_h * scale))
# resized_h = torch.round(im_h * scale)
resized_w = int(np.round(im_w * scale))
# resized_w = torch.round(im_w * scale)
crop_h = int(np.floor(resized_h - input_h) / 2)
# crop_h = torch.floor(resized_h - input_h) // 2
crop_w = int(np.floor(resized_w - input_w) / 2)
# crop_w = torch.floor(resized_w - input_w) // 2
# resized_img = cv2.resize(img, (resized_w, resized_h))
resized_img = F.upsample(
img.unsqueeze(0).unsqueeze(0), size=(resized_h, resized_w),
mode='bilinear')
resized_img = resized_img.squeeze().unsqueeze(0)
return resized_img[0, crop_h: crop_h + input_h,
crop_w: crop_w + input_w]
class ResizeImage:
"""Resize the largest of the sides of the image to a given size"""
def __init__(self, size):
if not isinstance(size, (int, Iterable)):
raise TypeError('Got inappropriate size arg: {}'.format(size))
self.size = size
def __call__(self, img):
im_h, im_w = img.shape[-2:]
scale = min(self.size / im_h, self.size / im_w)
resized_h = int(np.round(im_h * scale))
resized_w = int(np.round(im_w * scale))
out = F.upsample(
Variable(img).unsqueeze(0), size=(resized_h, resized_w),
mode='bilinear').squeeze().data
return out
class ResizeAnnotation:
"""Resize the largest of the sides of the annotation to a given size"""
def __init__(self, size):
if not isinstance(size, (int, Iterable)):
raise TypeError('Got inappropriate size arg: {}'.format(size))
self.size = size
def __call__(self, img):
im_h, im_w = img.shape[-2:]
scale = min(self.size / im_h, self.size / im_w)
resized_h = int(np.round(im_h * scale))
resized_w = int(np.round(im_w * scale))
out = F.upsample(
Variable(img).unsqueeze(0).unsqueeze(0),
size=(resized_h, resized_w),
mode='bilinear').squeeze().data
return out
class ToNumpy:
"""Transform an torch.*Tensor to an numpy ndarray."""
def __call__(self, x):
return x.numpy()
def letterbox(img, mask, height, color=(123.7, 116.3, 103.5)): # resize a rectangular image to a padded square
shape = img.shape[:2] # shape = [height, width]
ratio = float(height) / max(shape) # ratio = old / new
new_shape = (round(shape[1] * ratio), round(shape[0] * ratio))
dw = (height - new_shape[0]) / 2 # width padding
dh = (height - new_shape[1]) / 2 # height padding
top, bottom = round(dh - 0.1), round(dh + 0.1)
left, right = round(dw - 0.1), round(dw + 0.1)
img = cv2.resize(img, new_shape, interpolation=cv2.INTER_AREA) # resized, no border
img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color) # padded square
if mask is not None:
mask = cv2.resize(mask, new_shape, interpolation=cv2.INTER_NEAREST) # resized, no border
# print(top, bottom, left, right)
# input()
mask = cv2.copyMakeBorder(mask, top, bottom, left, right, cv2.BORDER_CONSTANT, value=1) # padded square
# print(mask)
return img, mask, ratio, dw, dh
def random_affine(img, mask, targets, degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-2, 2),
borderValue=(123.7, 116.3, 103.5), all_bbox=None):
border = 0 # width of added border (optional)
height = max(img.shape[0], img.shape[1]) + border * 2
# Rotation and Scale
R = np.eye(3)
a = random.random() * (degrees[1] - degrees[0]) + degrees[0]
# a += random.choice([-180, -90, 0, 90]) # 90deg rotations added to small rotations
s = random.random() * (scale[1] - scale[0]) + scale[0]
R[:2] = cv2.getRotationMatrix2D(angle=a, center=(img.shape[1] / 2, img.shape[0] / 2), scale=s)
# Translation
T = np.eye(3)
T[0, 2] = (random.random() * 2 - 1) * translate[0] * img.shape[0] + border # x translation (pixels)
T[1, 2] = (random.random() * 2 - 1) * translate[1] * img.shape[1] + border # y translation (pixels)
# Shear
S = np.eye(3)
S[0, 1] = math.tan((random.random() * (shear[1] - shear[0]) + shear[0]) * math.pi / 180) # x shear (deg)
S[1, 0] = math.tan((random.random() * (shear[1] - shear[0]) + shear[0]) * math.pi / 180) # y shear (deg)
M = S @ T @ R # Combined rotation matrix. ORDER IS IMPORTANT HERE!!
imw = cv2.warpPerspective(img, M, dsize=(height, height), flags=cv2.INTER_LINEAR,
borderValue=borderValue) # BGR order borderValue
if mask is not None:
maskw = cv2.warpPerspective(mask, M, dsize=(height, height), flags=cv2.INTER_NEAREST,
borderValue=1) # BGR order borderValue
else:
maskw = None
# Return warped points also
if type(targets)==type([1]):
targetlist=[]
for bbox in targets:
targetlist.append(wrap_points(bbox, M, height, a))
return imw, maskw, targetlist, M
elif all_bbox is not None:
targets = wrap_points(targets, M, height, a)
for ii in range(all_bbox.shape[0]):
all_bbox[ii,:] = wrap_points(all_bbox[ii,:], M, height, a)
return imw, maskw, targets, all_bbox, M
elif targets is not None: ## previous main
targets = wrap_points(targets, M, height, a)
return imw, maskw, targets, M
else:
return imw
def wrap_points(targets, M, height, a):
# n = targets.shape[0]
# points = targets[:, 1:5].copy()
points = targets.copy()
# area0 = (points[:, 2] - points[:, 0]) * (points[:, 3] - points[:, 1])
area0 = (points[2] - points[0]) * (points[3] - points[1])
# warp points
xy = np.ones((4, 3))
xy[:, :2] = points[[0, 1, 2, 3, 0, 3, 2, 1]].reshape(4, 2) # x1y1, x2y2, x1y2, x2y1
xy = (xy @ M.T)[:, :2].reshape(1, 8)
# create new boxes
x = xy[:, [0, 2, 4, 6]]
y = xy[:, [1, 3, 5, 7]]
xy = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, 1).T
# apply angle-based reduction
radians = a * math.pi / 180
reduction = max(abs(math.sin(radians)), abs(math.cos(radians))) ** 0.5
x = (xy[:, 2] + xy[:, 0]) / 2
y = (xy[:, 3] + xy[:, 1]) / 2
w = (xy[:, 2] - xy[:, 0]) * reduction
h = (xy[:, 3] - xy[:, 1]) * reduction
xy = np.concatenate((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).reshape(4, 1).T
# reject warped points outside of image
np.clip(xy, 0, height, out=xy)
w = xy[:, 2] - xy[:, 0]
h = xy[:, 3] - xy[:, 1]
area = w * h
ar = np.maximum(w / (h + 1e-16), h / (w + 1e-16))
i = (w > 4) & (h > 4) & (area / (area0 + 1e-16) > 0.1) & (ar < 10)
## print(targets, xy)
## [ 56 36 108 210] [[ 47.80464857 15.6096533 106.30993434 196.71267693]]
# targets = targets[i]
# targets[:, 1:5] = xy[i]
targets = xy[0]
return targets |