Spaces:
Build error
Build error
File size: 5,929 Bytes
d61b9c7 |
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 |
import cv2
import numpy as np
from PIL import Image, ImageOps
'''
PIL resize (W,H)
Torch resize is (H,W)
'''
class Shrink:
def __init__(self):
self.tps = cv2.createThinPlateSplineShapeTransformer()
self.translateXAbs = TranslateXAbs()
self.translateYAbs = TranslateYAbs()
def __call__(self, img, mag=-1, prob=1.):
if np.random.uniform(0,1) > prob:
return img
W, H = img.size
img = np.array(img)
srcpt = list()
dstpt = list()
W_33 = 0.33 * W
W_50 = 0.50 * W
W_66 = 0.66 * W
H_50 = 0.50 * H
P = 0
#frac = 0.4
b = [.2, .3, .4]
if mag<0 or mag>=len(b):
index = 0
else:
index = mag
frac = b[index]
# left-most
srcpt.append([P, P])
srcpt.append([P, H-P])
x = np.random.uniform(frac-.1, frac)*W_33
y = np.random.uniform(frac-.1, frac)*H_50
dstpt.append([P+x, P+y])
dstpt.append([P+x, H-P-y])
# 2nd left-most
srcpt.append([P+W_33, P])
srcpt.append([P+W_33, H-P])
dstpt.append([P+W_33, P+y])
dstpt.append([P+W_33, H-P-y])
# 3rd left-most
srcpt.append([P+W_66, P])
srcpt.append([P+W_66, H-P])
dstpt.append([P+W_66, P+y])
dstpt.append([P+W_66, H-P-y])
# right-most
srcpt.append([W-P, P])
srcpt.append([W-P, H-P])
dstpt.append([W-P-x, P+y])
dstpt.append([W-P-x, H-P-y])
N = len(dstpt)
matches = [cv2.DMatch(i, i, 0) for i in range(N)]
dst_shape = np.array(dstpt).reshape((-1, N, 2))
src_shape = np.array(srcpt).reshape((-1, N, 2))
self.tps.estimateTransformation(dst_shape, src_shape, matches)
img = self.tps.warpImage(img)
img = Image.fromarray(img)
if np.random.uniform(0, 1) < 0.5:
img = self.translateXAbs(img, val=x)
else:
img = self.translateYAbs(img, val=y)
return img
class Rotate:
def __init__(self, square_side=224):
self.side = square_side
def __call__(self, img, iscurve=False, mag=-1, prob=1.):
if np.random.uniform(0,1) > prob:
return img
W, H = img.size
if H!=self.side or W!=self.side:
img = img.resize((self.side, self.side), Image.BICUBIC)
b = [20., 40, 60]
if mag<0 or mag>=len(b):
index = 1
else:
index = mag
rotate_angle = b[index]
angle = np.random.uniform(rotate_angle-20, rotate_angle)
if np.random.uniform(0, 1) < 0.5:
angle = -angle
#angle = np.random.normal(loc=0., scale=rotate_angle)
#angle = min(angle, 2*rotate_angle)
#angle = max(angle, -2*rotate_angle)
expand = False if iscurve else True
img = img.rotate(angle=angle, resample=Image.BICUBIC, expand=expand)
img = img.resize((W, H), Image.BICUBIC)
return img
class Perspective:
def __init__(self):
pass
def __call__(self, img, mag=-1, prob=1.):
if np.random.uniform(0,1) > prob:
return img
W, H = img.size
# upper-left, upper-right, lower-left, lower-right
src = np.float32([[0, 0], [W, 0], [0, H], [W, H]])
#low = 0.3
b = [.1, .2, .3]
if mag<0 or mag>=len(b):
index = 2
else:
index = mag
low = b[index]
high = 1 - low
if np.random.uniform(0, 1) > 0.5:
toprightY = np.random.uniform(low, low+.1)*H
bottomrightY = np.random.uniform(high-.1, high)*H
dest = np.float32([[0, 0], [W, toprightY], [0, H], [W, bottomrightY]])
else:
topleftY = np.random.uniform(low, low+.1)*H
bottomleftY = np.random.uniform(high-.1, high)*H
dest = np.float32([[0, topleftY], [W, 0], [0, bottomleftY], [W, H]])
M = cv2.getPerspectiveTransform(src, dest)
img = np.array(img)
img = cv2.warpPerspective(img, M, (W, H) )
img = Image.fromarray(img)
return img
class TranslateX:
def __init__(self):
pass
def __call__(self, img, mag=-1, prob=1.):
if np.random.uniform(0,1) > prob:
return img
b = [.03, .06, .09]
if mag<0 or mag>=len(b):
index = 2
else:
index = mag
v = b[index]
v = np.random.uniform(v-0.03, v)
v = v * img.size[0]
if np.random.uniform(0,1) > 0.5:
v = -v
return img.transform(img.size, Image.AFFINE, (1, 0, v, 0, 1, 0))
class TranslateY:
def __init__(self):
pass
def __call__(self, img, mag=-1, prob=1.):
if np.random.uniform(0,1) > prob:
return img
b = [.07, .14, .21]
if mag<0 or mag>=len(b):
index = 2
else:
index = mag
v = b[index]
v = np.random.uniform(v-0.07, v)
v = v * img.size[1]
if np.random.uniform(0,1) > 0.5:
v = -v
return img.transform(img.size, Image.AFFINE, (1, 0, 0, 0, 1, v))
class TranslateXAbs:
def __init__(self):
pass
def __call__(self, img, val=0, prob=1.):
if np.random.uniform(0,1) > prob:
return img
v = np.random.uniform(0, val)
if np.random.uniform(0,1) > 0.5:
v = -v
return img.transform(img.size, Image.AFFINE, (1, 0, v, 0, 1, 0))
class TranslateYAbs:
def __init__(self):
pass
def __call__(self, img, val=0, prob=1.):
if np.random.uniform(0,1) > prob:
return img
v = np.random.uniform(0, val)
if np.random.uniform(0,1) > 0.5:
v = -v
return img.transform(img.size, Image.AFFINE, (1, 0, 0, 0, 1, v))
|