forked from bartoszptak/PupilTracker
/
main.py
252 lines (201 loc) · 9.67 KB
/
main.py
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
from imutils import face_utils
import numpy as np
import dlib
import cv2
import ctypes
from keras.models import load_model
import os
data_path = 'data'
detector = dlib.get_frontal_face_detector()
predictor = dlib.shape_predictor(os.path.join(data_path, 'shape_predictor_68_face_landmarks.dat'))
model = load_model(os.path.join(data_path, 'model.h5'))
model.load_weights(os.path.join(data_path, 'weights.h5'))
user32 = ctypes.windll.user32
screenX, screenY = user32.GetSystemMetrics(0), user32.GetSystemMetrics(1)
window_name = "Pupils Tracker"
cv2.namedWindow(window_name, cv2.WND_PROP_FULLSCREEN)
cv2.setWindowProperty(window_name, cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)
background = np.zeros((screenY, screenX), np.uint8)
background = np.bitwise_not(background)
background = cv2.cvtColor(background, cv2.COLOR_GRAY2BGR)
bg = background.copy()
cv2.imshow(window_name, bg)
cap = cv2.VideoCapture(1)
cursor = False
calibration = 8
pupilPositions = []
sightFocus = [0, 0, 0, 0, 0, 0]
bgz = cv2.imread(os.path.join(data_path, 'bg.png'))
#Increase brightness of image
def increase_brightness(img, value):
hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HLS)
h, l, s = cv2.split(hsv)
l[:] = int(255 * value)
final_hsl = cv2.merge([h, l, s])
img2 = cv2.cvtColor(final_hsl, cv2.COLOR_HLS2BGR)
return img2
#Calculate distance between two eye positions
def distance(A, B):
dXA = A[3] - A[1]
dYA = A[2] - A[0]
dXB = B[3] - B[1]
dYB = B[2] - B[0]
a = abs(dXB - dXA)
b = abs(dYB - dYA)
return (a ** 2 + b ** 2) ** (1 / 2)
#Left eye recognition
def searching(image):
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
rects = detector(gray, 1)
for (i, rect) in enumerate(rects):
shape = predictor(gray, rect)
shape = face_utils.shape_to_np(shape)
clone = image
leftXs, leftYs = [], []
for x, y in shape[36:42]:
leftXs.append(x)
leftYs.append(y)
left = clone[min(leftYs) - 5:max(leftYs) + 5, min(leftXs) - 5:max(leftXs) + 5]
return cv2.resize(left, (100, 100))
#Pupil and cornea position prediction
def prediction(eye_img):
eye_img = cv2.resize(eye_img, (100, 100))
eye_img_gray = cv2.cvtColor(eye_img, cv2.COLOR_RGB2GRAY)
eye_img_gray = cv2.equalizeHist(eye_img_gray)
eye_img_normalized = eye_img_gray * 1. / 255
Z1 = np.expand_dims(eye_img_normalized, axis=0)
Z = Z1[..., np.newaxis]
result = model.predict(Z)
result *= 100
return [int(result[0][0]), int(result[0][1]), int(result[0][2]), int(result[0][3])]
#Calculate average of two numbers
def averaging(a, b):
return int((a + b) / 2)
#Enable cursor control with look (high resolution camera needed)
def enableCursor(K):
xleftRangeBorder = averaging(pupilPositions[1][2], pupilPositions[4][2]) \
- averaging(pupilPositions[1][0], pupilPositions[4][0])
xRightRangeBorder = averaging(pupilPositions[3][2], pupilPositions[6][2]) \
- averaging(pupilPositions[3][0], pupilPositions[6][0])
yUpperRangeBorder = averaging(pupilPositions[1][1], pupilPositions[3][1]) \
- averaging(pupilPositions[1][3], pupilPositions[3][3])
yLowerRangeBorder = averaging(pupilPositions[4][1], pupilPositions[6][1]) \
- averaging(pupilPositions[4][3], pupilPositions[6][3])
xRange = xleftRangeBorder - xRightRangeBorder
yRange = yLowerRangeBorder - yUpperRangeBorder
xRealTime = K[2] - K[0] - xRightRangeBorder
yRealTime = K[1] - K[3] - yUpperRangeBorder
xScale = 1 - xRealTime / xRange
yScale = yRealTime / yRange
pX = int(screenX * xScale)
pY = int(screenY * yScale)
si = 20
cv2.line(bg, (pX - si, pY), (pX + si, pY), (0, 0, 255), 3)
cv2.line(bg, (pX, pY - si), (pX, pY + si), (0, 0, 255), 3)
while True:
ret, frame = cap.read()
image = cv2.flip(frame, +1)
key = cv2.waitKey(1)
cv2.rectangle(bg, (0, 0), (int(screenX / 3), int(screenY / 2)), (0, 0, 10), -1) # 00
cv2.rectangle(bg, (int(screenX / 3), 0), (int(screenX / 3 * 2), int(screenY / 2)), (42, 0, 10), -1) # 10
cv2.rectangle(bg, (int(screenX / 3 * 2), 0), (screenX, int(screenY / 2)), (84, 0, 10), -1) # 20
cv2.rectangle(bg, (0, int(screenY / 2)), (int(screenX / 3), screenY), (126, 0, 10), -1) # 01
cv2.rectangle(bg, (int(screenX / 3), int(screenY / 2)), (int(screenX / 3 * 2), screenY), (168, 0, 10), -1) # 11
cv2.rectangle(bg, (int(screenX / 3 * 2), int(screenY / 2)), (screenX, screenY), (255, 0, 10), -1) # 21
if calibration == 8:
cv2.putText(bg, "Cisnij spacje tej!", (400, 500), cv2.FONT_HERSHEY_SIMPLEX, 4, (255, 255, 255), 2,
cv2.LINE_AA)
elif calibration == 7:
cv2.circle(bg, (int(screenX / 24), int(screenY / 16)), 10, (0, 0, 255), -1)
elif calibration == 6:
cv2.circle(bg, (int(screenX / 2), int(screenY / 16)), 10, (0, 0, 255), -1)
elif calibration == 5:
cv2.circle(bg, (int(screenX / 24 * 23), int(screenY / 16)), 10, (0, 0, 255), -1)
elif calibration == 4:
cv2.circle(bg, (int(screenX / 24), int(screenY / 16 * 15)), 10, (0, 0, 255), -1)
elif calibration == 3:
cv2.circle(bg, (int(screenX / 2), int(screenY / 16 * 15)), 10, (0, 0, 255), -1)
elif calibration == 2:
cv2.circle(bg, (int(screenX / 24 * 23), int(screenY / 16 * 15)), 10, (0, 0, 255), -1)
elif calibration == 1:
cv2.circle(bg, (int(screenX / 3), int(screenY / 2)), 10, (0, 0, 255), -1)
elif calibration == 0:
cv2.circle(bg, (int(screenX / 3 * 2), int(screenY / 2)), 10, (0, 0, 255), -1)
else:
bg = bgz.copy()
left = searching(image)
if left is not None:
K = prediction(left)
distancesFromRegions = \
[distance(K, pupilPositions[1]),
distance(K, pupilPositions[2]),
distance(K, pupilPositions[3]),
distance(K, pupilPositions[4]),
distance(K, pupilPositions[5]),
distance(K, pupilPositions[6])]
closestRegionIndex = distancesFromRegions.index(min(distancesFromRegions))
sightFocus[closestRegionIndex] += 1
sightFocusSum = sum(sightFocus)
reg0 = bg[0:int(screenY / 2), 0:int(screenX / 3)]
reg1 = bg[0:int(screenY / 2), int(screenX / 3):int(screenX / 3 * 2)]
reg2 = bg[0:int(screenY / 2), int(screenX / 3 * 2):int(screenX)]
reg3 = bg[int(screenY / 2):int(screenY), 0:int(screenX / 3)]
reg4 = bg[int(screenY / 2):int(screenY), int(screenX / 3):int(screenX / 3 * 2)]
reg5 = bg[int(screenY / 2):int(screenY), int(screenX / 3 * 2):int(screenX)]
reg0 = cv2.convertScaleAbs(reg0, alpha=sightFocus[0] / sightFocusSum)
reg1 = cv2.convertScaleAbs(reg1, alpha=sightFocus[1] / sightFocusSum)
reg2 = cv2.convertScaleAbs(reg2, alpha=sightFocus[2] / sightFocusSum)
reg3 = cv2.convertScaleAbs(reg3, alpha=sightFocus[3] / sightFocusSum)
reg4 = cv2.convertScaleAbs(reg4, alpha=sightFocus[4] / sightFocusSum)
reg5 = cv2.convertScaleAbs(reg5, alpha=sightFocus[5] / sightFocusSum)
bg[0:int(screenY / 2), 0:int(screenX / 3)] = reg0
bg[0:int(screenY / 2), int(screenX / 3):int(screenX / 3 * 2)] = reg1
bg[0:int(screenY / 2), int(screenX / 3 * 2):int(screenX)] = reg2
bg[int(screenY / 2):int(screenY), 0:int(screenX / 3)] = reg3
bg[int(screenY / 2):int(screenY), int(screenX / 3):int(screenX / 3 * 2)] = reg4
bg[int(screenY / 2):int(screenY), int(screenX / 3 * 2):int(screenX)] = reg5
cv2.putText(bg,
str(sightFocus[0]),
(int(screenX / 12), int(screenY / 8)),
cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 0, 255), 2, cv2.LINE_AA)
cv2.putText(bg,
str(sightFocus[1]),
(int(screenX / 2), int(screenY / 8)),
cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 0, 255), 2, cv2.LINE_AA)
cv2.putText(bg,
str(sightFocus[2]),
(int(screenX / 12 * 11), int(screenY / 8)),
cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 0, 255), 2, cv2.LINE_AA)
cv2.putText(bg,
str(sightFocus[3]),
(int(screenX / 12), int(screenY / 8 * 7)),
cv2.FONT_HERSHEY_SIMPLEX,
1, (0, 0, 255), 2, cv2.LINE_AA)
cv2.putText(bg,
str(sightFocus[4]),
(int(screenX / 2), int(screenY / 8 * 7)),
cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 0, 255), 2, cv2.LINE_AA)
cv2.putText(bg,
str(sightFocus[5]),
(int(screenX / 12 * 11), int(screenY / 8 * 7)),
cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 0, 255), 2, cv2.LINE_AA)
cv2.circle(left, (K[0], K[1]), 2, (0, 0, 255), -2)
cv2.circle(left, (K[2], K[3]), 2, (0, 255, 255), -2)
if cursor:
enableCursor(K)
left = cv2.resize(left, None, None, 2, 2)
bg[int(bg.shape[0] / 2 - left.shape[0] / 2):int(bg.shape[0] / 2 + left.shape[0] / 2),
int(bg.shape[1] / 2 - left.shape[1] / 2):int(bg.shape[1] / 2 + left.shape[1] / 2)] = left
if key == 32 and calibration >= 0:
calibration -= 1
pupilPositions.append(prediction(searching(image)))
elif key == 27:
break
cv2.imshow(window_name, bg)
cap.release()
cv2.destroyAllWindows()