forked from thomasballinger/gazer
-
Notifications
You must be signed in to change notification settings - Fork 0
/
demo.py
executable file
·192 lines (169 loc) · 7.56 KB
/
demo.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
#!/usr/bin/env python
"""Reports where on screen eyes are pointing"""
import os
import time
import cv
import cv2
import numpy
import cvnumpyconvert
from mouse import setMousePosition
class EyeTracker(object):
def __init__(self):
self.storage = cv.CreateMemStorage(0)
self.last_face_position = None
self.face_cascade = cv.Load(os.path.expanduser('haarcascade_frontalface_default.xml'))
#self.face_cascade = cv.Load('haarcascade_frontalface_alt.xml')
#self.eye_cascade = cv.Load(os.path.expanduser('~/Downloads/parojos-22x15.xml'))
self.eye_cascade = cv.Load(os.path.expanduser('parojosG-45x11.xml'))
self.detect_times = []
self.eye_pair_history = []
self.xamount_histories = [[], []]
self.yamount_histories = [[], []]
self.xpos_history = []
self.ypos_history = []
def detect(self, image):
f = self.find_face(image)
if f.size > 0:
eyes = self.find_eyes(image, f)
if eyes:
rolling_eyes = self.rolling_eye_pair(eyes, samples=5)
# find black portions of outside thirds of image
pupils = self.find_pupils(image, rolling_eyes)
self.fps(False)
def rolling_eye_pair(self, eye_pair, samples=3):
self.eye_pair_history = self.eye_pair_history[-(samples-1):]
self.eye_pair_history.append(eye_pair)
ave_eye_pair = [0,0,0,0]
for i in range(4):
ave_eye_pair[i] = sum([es[i] for es in self.eye_pair_history])/len(self.eye_pair_history)
rect(image, ave_eye_pair, (100,110,30))
return [ave_eye_pair, '?']
def fps(self, display=True):
t = time.time()
self.detect_times.append(t)
if len(self.detect_times) > 3:
if display:
print 'FPS: %.1d' % (len(self.detect_times) / (t - self.detect_times[0]))
if len(self.detect_times) > 100:
self.detect_times.pop(0)
def find_pupils(self, image, eyes):
h, w, d = image.shape
left = cv.CreateImage((eyes[0][2]/3, eyes[0][3]*2/3,), 8, 3)
right = cv.CreateImage((eyes[0][2]/3, eyes[0][3]*2/3,), 8, 3)
left_gray = cv.CreateImage((eyes[0][2]/3, eyes[0][3]*2/3,), 8, 1)
right_gray = cv.CreateImage((eyes[0][2]/3, eyes[0][3]*2/3,), 8, 1)
left_region = cv.GetSubRect(cv.fromarray(image), (eyes[0][0], eyes[0][1]+eyes[0][3]/6,
eyes[0][2]/3, eyes[0][3]*2/3))
right_region = cv.GetSubRect(cv.fromarray(image), (eyes[0][0]+eyes[0][2]*2/3, eyes[0][1]+eyes[0][3]/6,
eyes[0][2]/3, eyes[0][3]*2/3))
cv.Copy(left_region, left)
cv.Copy(right_region, right)
cv.CvtColor(left, left_gray, cv.CV_BGR2GRAY)
cv.CvtColor(right, right_gray, cv.CV_BGR2GRAY)
l = numpy.squeeze(cvnumpyconvert.cv2array(left_gray))
r = numpy.squeeze(cvnumpyconvert.cv2array(right_gray))
if False:
from matplotlib.pylab import imshow, show
imshow(l)
show()
imshow(r)
show()
#cv.ShowImage('left', left_gray)
#cv.ShowImage('right', right_gray)
for i, arr, side in [(0, l, 'left'), (1, r, 'right')]:
squared = arr**2
ave = numpy.average(squared**2)
rightness = numpy.array([range(arr.shape[1]) for _ in range(arr.shape[0])])
downness = numpy.array([[x for _ in range(arr.shape[1])] for x in range(arr.shape[0])])
xamount = numpy.sum(squared*(rightness**2)) / numpy.sum(rightness**2 * ave)
yamount = numpy.sum(squared*(downness **2)) / numpy.sum(downness **2 * ave)
self.xamount_histories[i].append(xamount)
self.yamount_histories[i].append(yamount)
#xmin = min(self.xamount_histories[i])
#xmax = max(self.xamount_histories[i])
sorted_history = list(self.xamount_histories[i])
sorted_history.sort()
try:
xmin = min(sorted_history[len(sorted_history)/6:len(sorted_history)*5/6])
xmax = max(sorted_history[len(sorted_history)/6:len(sorted_history)*5/6])
except ValueError:
print 'jump starting,', len(sorted_history)
xmin = sorted_history[0]
xmax = sorted_history[0]
sorted_history = list(self.yamount_histories[i])
sorted_history.sort()
try:
ymin = min(sorted_history[len(sorted_history)/6:len(sorted_history)*5/6])
ymax = max(sorted_history[len(sorted_history)/6:len(sorted_history)*5/6])
except ValueError:
print 'jump starting,', len(sorted_history)
ymin = sorted_history[0]
ymax = sorted_history[0]
screen_width = 1440
screen_height = 900
xpos = (xamount - xmin) / (xmax - xmin) * screen_width
ypos = (yamount - ymin) / (ymax - ymin) * screen_height
self.xpos_history.append(xpos)
self.ypos_history.append(ypos)
smooth = 20
xsmoothed = sum(self.xpos_history[-smooth:]) / len(self.xpos_history[-smooth:])
ysmoothed = sum(self.ypos_history[-smooth:]) / len(self.ypos_history[-smooth:])
setMousePosition(xsmoothed, ysmoothed)
#raw_input('break to allow for ctrl-c')
#print r
#raw_input()
def find_eyes(self, image, f):
h, w, d = image.shape
small = cv.CreateImage((f[2], f[3]*2/3,), 8, 3)
src_region = cv.GetSubRect(cv.fromarray(image), (f[0], f[1],
f[2], f[3]*2/3))
cv.Copy(src_region, small)
grayscale = cv.CreateImage((f[2], f[3]*2/3), 8, 1)
cv.CvtColor(small, grayscale, cv.CV_BGR2GRAY)
#eyecascade = cv2.CascadeClassifier('parojosG-45x11.xml')
#eye_pairs = eyecascade.detectMultiScale(src_region, 10, 10)
eye_pairs = cv.HaarDetectObjects(grayscale, self.eye_cascade, self.storage, 1.2, 2, 0, (10, 10))
for eye_pair in eye_pairs:
eye_pair = (eye_pair[0][0]+f[0], eye_pair[0][1]+f[1], eye_pair[0][2], eye_pair[0][3])
rect(image, eye_pair, (255,0,255))
return eye_pair
def find_face(self, image):
h, w, d = image.shape
grayscale = cv.CreateImage((w, h), 8, 1)
cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')
faces = cascade.detectMultiScale(image)
#faces = cascade.detectMultiScale(image, 1.2, 2, 0, 300, 250)
#faces = cv.HaarDetectObjects(grayscale, self.face_cascade, self.storage, 1.2, 2, 0, (300, 250))
if faces.size > 0:
print 'face detected!'
for f in faces:
rect(image, f, (0, 255, 0))
self.frames_since_face = 0
self.last_face_position = f
return f
elif self.last_face_position:
print 'can\'t find face, using old postion'
self.frames_since_face += 1
f = self.last_face_position
rect(image, f, (0, 100, 200))
return f
else:
print 'no face'
def rect(image, result, color=(0,0,255)):
f = result
cv2.rectangle(image, (f[0], f[1]),
(f[0]+f[2], f[1]+f[3]),
cv.RGB(*color), 3, 8, 0)
if __name__ == '__main__':
cv2.namedWindow('a_window', cv2.WINDOW_AUTOSIZE)
cap = cv2.VideoCapture(0)
et = EyeTracker()
while True:
ret, image = cap.read()
if image is not None:
et.detect(image)
cv2.imshow('a_window', image)
cv2.waitKey(100)
cv2.destroyAllWindows()
raw_input()