File size: 4,058 Bytes
39b4c8c |
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 |
import numpy as np
from scipy import signal
import sys
import cv2
from pyVHR.utils.HDI import hdi, hdi2
class SkinDetect():
def __init__(self, strength=0.2):
self.description = 'Skin Detection Module'
self.strength = strength
self.stats_computed = False
def compute_stats(self, face):
assert (self.strength > 0 and self.strength < 1), "'strength' parameter must have values in [0,1]"
faceColor = cv2.cvtColor(face, cv2.COLOR_RGB2HSV)
h = faceColor[:,:,0].reshape(-1,1)
s = faceColor[:,:,1].reshape(-1,1)
v = faceColor[:,:,2].reshape(-1,1)
alpha = self.strength #the highest, the stronger the masking
hpd_h, x_h, y_h, modes_h = hdi2(np.squeeze(h), alpha=alpha)
min_s, max_s = hdi(np.squeeze(s), alpha=alpha)
min_v, max_v = hdi(np.squeeze(v), alpha=alpha)
if len(hpd_h) > 1:
self.multiple_modes = True
if len(hpd_h) > 2:
print('WARNING!! Found more than 2 HDIs in Hue Channel empirical Distribution... Considering only 2')
from scipy.spatial.distance import pdist, squareform
m = np.array(modes_h).reshape(-1,1)
d = squareform(pdist(m))
maxij = np.where(d==d.max())[0]
i = maxij[0]
j = maxij[1]
else:
i = 0
j = 1
min_h1 = hpd_h[i][0]
max_h1 = hpd_h[i][1]
min_h2 = hpd_h[j][0]
max_h2 = hpd_h[j][1]
self.lower1 = np.array([min_h1, min_s, min_v], dtype = "uint8")
self.upper1 = np.array([max_h1, max_s, max_v], dtype = "uint8")
self.lower2 = np.array([min_h2, min_s, min_v], dtype = "uint8")
self.upper2 = np.array([max_h2, max_s, max_v], dtype = "uint8")
elif len(hpd_h) == 1:
self.multiple_modes = False
min_h = hpd_h[0][0]
max_h = hpd_h[0][1]
self.lower = np.array([min_h, min_s, min_v], dtype = "uint8")
self.upper = np.array([max_h, max_s, max_v], dtype = "uint8")
self.stats_computed = True
def get_skin(self, face, filt_kern_size=7, verbose=False, plot=False):
if not self.stats_computed:
raise ValueError("ERROR! You must compute stats at least one time")
faceColor = cv2.cvtColor(face, cv2.COLOR_RGB2HSV)
if self.multiple_modes:
if verbose:
print('\nLower1: ' + str(self.lower1))
print('Upper1: ' + str(self.upper1))
print('\nLower2: ' + str(self.lower2))
print('Upper2: ' + str(self.upper2) + '\n')
skinMask1 = cv2.inRange(faceColor, self.lower1, self.upper1)
skinMask2 = cv2.inRange(faceColor, self.lower2, self.upper2)
skinMask = np.logical_or(skinMask1, skinMask2).astype(np.uint8)*255
else:
if verbose:
print('\nLower: ' + str(lower))
print('Upper: ' + str(upper) + '\n')
skinMask = cv2.inRange(faceColor, self.lower, self.upper)
if filt_kern_size > 0:
skinMask = signal.medfilt2d(skinMask, kernel_size=filt_kern_size)
skinFace = cv2.bitwise_and(face, face, mask=skinMask)
if plot:
h = faceColor[:,:,0].reshape(-1,1)
s = faceColor[:,:,1].reshape(-1,1)
v = faceColor[:,:,2].reshape(-1,1)
import matplotlib.pyplot as plt
plt.figure()
plt.subplot(2,2,1)
plt.hist(h, 20)
plt.title('Hue')
plt.subplot(2,2,2)
plt.hist(s, 20)
plt.title('Saturation')
plt.subplot(2,2,3)
plt.hist(v, 20)
plt.title('Value')
plt.subplot(2,2,4)
plt.imshow(skinFace)
plt.title('Masked Face')
plt.show()
return skinFace
|