def process(frame, x0, y0, width, height):
global guessGesture, visualize, mod, gestname, path
#Call CNN model loading callback
print "Will load default weight file"
mod = myNN.loadCNN(0)
masked = skinMask(frame, x0, y0, width, height)
cv2.imshow('ROI', masked)
def Main():
global guessGesture, mod, binaryMode, x0, y0, width, height, saveImg, gestname, path
quietMode = False
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.5
fx = 10
fy = 355
fh = 18
while True:
ans = int(raw_input(banner))
if ans == 1:
print "Will load default weight file"
mod = myNN.loadCNN(0)
break
else:
print "Your face is so ugly ,even my code does not want to see u!!"
return 0
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
ret = cap.set(3, 640)
ret = cap.set(4, 480)
while (True):
ret, frame = cap.read()
max_area = 0
frame = cv2.flip(frame, 3)
if ret == True:
if binaryMode == True:
roi = binaryMask(frame, x0, y0, width, height)
if not quietMode:
cv2.imshow('Original', frame)
cv2.imshow('ROI', roi)
key = cv2.waitKey(10) & 0xff
if key == 27:
break
elif key == ord('b'):
binaryMode = not binaryMode
if binaryMode:
print "Binary Threshold filter active"
elif key == ord('g'):
guessGesture = not guessGesture
print "Prediction Mode - {}".format(guessGesture)
cap.release()
cv2.destroyAllWindows()
def Main():
global guessGesture, visualize, mod, binaryMode, bkgrndSubMode, mask, takebkgrndSubMask, x0, y0, width, height, saveImg, gestname, path
quietMode = False
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.5
fx = 10
fy = 350
fh = 18
while True:
ans = int(input(banner))
if ans == 1:
mod = myNN.loadCNN()
break
elif ans == 2:
mod = myNN.loadCNN(True)
myNN.trainModel(mod)
input("Press any key to continue")
break
else:
print("Exiting...")
return 0
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
ret = cap.set(3, 640)
ret = cap.set(4, 480)
framecount = 0
fps = ""
start = time.time()
plot = np.zeros((512, 512, 3), np.uint8)
while (True):
ret, frame = cap.read()
max_area = 0
frame = cv2.flip(frame, 3)
frame = cv2.resize(frame, (640, 480))
if ret == True:
if bkgrndSubMode == True:
roi = bkgrndSubMask(frame, x0, y0, width, height, framecount,
plot)
elif binaryMode == True:
roi = binaryMask(frame, x0, y0, width, height, framecount,
plot)
else:
roi = skinMask(frame, x0, y0, width, height, framecount, plot)
framecount = framecount + 1
end = time.time()
timediff = (end - start)
if (timediff >= 1):
fps = 'FPS:%s' % (framecount)
start = time.time()
framecount = 0
cv2.putText(frame, fps, (10, 20), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'Options:', (fx, fy), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'b - Toggle Binary/SkinMask', (fx, fy + fh), font,
size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'g - Toggle Prediction Mode', (fx, fy + 3 * fh),
font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'ESC - Exit', (fx, fy + 7 * fh), font, size,
(0, 255, 0), 1, 1)
if not quietMode:
cv2.imshow('Original', frame)
cv2.imshow('ROI', roi)
if guessGesture == True:
plot = np.zeros((512, 512, 3), np.uint8)
plot = myNN.update(plot)
cv2.imshow('Gesture Probability', plot)
key = cv2.waitKey(5) & 0xff
if key == 27:
break
elif key == ord('b'):
binaryMode = not binaryMode
bkgrndSubMode = False
if binaryMode:
print("Binary Threshold filter active")
else:
print("SkinMask filter active")
elif key == ord('x'):
takebkgrndSubMask = True
bkgrndSubMode = True
print("BkgrndSubMask filter active")
elif key == ord('g'):
guessGesture = not guessGesture
print("Prediction Mode - {}".format(guessGesture))
elif key == ord('i'):
y0 = y0 - 5
elif key == ord('k'):
y0 = y0 + 5
elif key == ord('j'):
x0 = x0 - 5
elif key == ord('l'):
x0 = x0 + 5
elif key == ord('q'):
quietMode = not quietMode
print("Quiet Mode - {}".format(quietMode))
elif key == ord('s'):
saveImg = not saveImg
if gestname != '':
saveImg = True
else:
print("Enter a gesture group name first, by pressing 'n'")
saveImg = False
elif key == ord('n'):
gestname = input("Enter the gesture folder name: ")
try:
os.makedirs(gestname)
except OSError as e:
# if directory already present
if e.errno != 17:
print('Some issue while creating the directory named -' +
gestname)
path = "./" + gestname + "/"
cap.release()
cv2.destroyAllWindows()
def Main():
global guessGesture, visualize, mod, binaryMode, bkgrndSubMode, mask, takebkgrndSubMask, x0, y0, width, height, saveImg, gestname, path
quietMode = False
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.5
fx = 10
fy = 350
fh = 18
#Call CNN model loading callback
while True:
ans = int(input(banner))
if ans == 2:
mod = myNN.loadCNN(-1)
myNN.trainModel(mod)
input("Press any key to continue")
break
elif ans == 1:
print("Will load default weight file")
mod = myNN.loadCNN(0)
break
elif ans == 3:
if not mod:
w = int(input("Which weight file to load (0 or 1)"))
mod = myNN.loadCNN(w)
else:
print("Will load default weight file")
img = int(input("Image number "))
layer = int(input("Enter which layer to visualize "))
myNN.visualizeLayers(mod, img, layer)
input("Press any key to continue")
continue
else:
print("Get out of here!!!")
return 0
## Grab camera input
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
# set rt size as 640x480
ret = cap.set(3, 640)
ret = cap.set(4, 480)
framecount = 0
fps = ""
start = time.time()
plot = np.zeros((512, 512, 3), np.uint8)
global x
while (True):
ret, frame = cap.read()
max_area = 0
frame = cv2.flip(frame, 3)
frame = cv2.resize(frame, (640, 480))
if ret == True:
if bkgrndSubMode == True:
roi = bkgrndSubMask(frame, x0, y0, width, height, framecount,
plot)
elif binaryMode == True:
roi = binaryMask(frame, x0, y0, width, height, framecount,
plot)
else:
roi = skinMask(frame, x0, y0, width, height, framecount, plot)
framecount = framecount + 1
end = time.time()
timediff = (end - start)
if (timediff >= 1):
#timediff = end - start
fps = 'FPS:%s' % (framecount)
start = time.time()
framecount = 0
cv2.putText(frame, fps, (10, 20), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'Options:', (fx, fy), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'b - Toggle Binary/SkinMask', (fx, fy + fh), font,
size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'x - Toggle Background Sub Mask', (fx, fy + 2 * fh),
font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'g - Toggle Prediction Mode', (fx, fy + 3 * fh),
font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'q - Toggle Quiet Mode', (fx, fy + 4 * fh), font,
size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'n - To enter name of new gesture folder',
(fx, fy + 5 * fh), font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 's - To start capturing new gestures for training',
(fx, fy + 6 * fh), font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'ESC - Exit', (fx, fy + 7 * fh), font, size,
(0, 255, 0), 1, 1)
## If enabled will stop updating the main openCV windows
## Way to reduce some processing power :)
if not quietMode:
cv2.imshow('Original', frame)
cv2.imshow('ROI', roi)
if guessGesture == True:
#plot = np.zeros((512,512,3), np.uint8)
plot = myNN.update(plot, x)
x = (x + 1) % 2
cv2.imshow('Gesture Probability', plot)
#plot = np.zeros((512,512,3), np.uint8)
############## Keyboard inputs ##################
key = cv2.waitKey(5) & 0xff
## Use Esc key to close the program
if key == 27:
break
## Use b key to toggle between binary threshold or skinmask based filters
elif key == ord('b'):
binaryMode = not binaryMode
bkgrndSubMode = False
if binaryMode:
print("Binary Threshold filter active")
else:
print("SkinMask filter active")
## Use g key to start gesture predictions via CNN
elif key == ord('x'):
takebkgrndSubMask = True
bkgrndSubMode = True
print("BkgrndSubMask filter active")
## Use g key to start gesture predictions via CNN
elif key == ord('g'):
guessGesture = not guessGesture
print("Prediction Mode - {}".format(guessGesture))
## This option is not yet complete. So disabled for now
## Use v key to visualize layers
#elif key == ord('v'):
# visualize = True
## Use i,j,k,l to adjust ROI window
elif key == ord('i'):
y0 = y0 - 5
elif key == ord('k'):
y0 = y0 + 5
elif key == ord('j'):
x0 = x0 - 5
elif key == ord('l'):
x0 = x0 + 5
## Quiet mode to hide gesture window
elif key == ord('q'):
quietMode = not quietMode
print("Quiet Mode - {}".format(quietMode))
## Use s key to start/pause/resume taking snapshots
## numOfSamples controls number of snapshots to be taken PER gesture
elif key == ord('s'):
saveImg = not saveImg
if gestname != '':
saveImg = True
else:
print("Enter a gesture group name first, by pressing 'n'")
saveImg = False
## Use n key to enter gesture name
elif key == ord('n'):
gestname = input("Enter the gesture folder name: ")
try:
os.makedirs(gestname)
except OSError as e:
# if directory already present
if e.errno != 17:
print('Some issue while creating the directory named -' +
gestname)
path = "./" + gestname + "/"
#elif key != 255:
# print key
#Realse & destroy
cap.release()
cv2.destroyAllWindows()
cv2.waitKey(0)
def load_model():
global mod, binaryMode, x0, y0, width, height
print "Will load default weight file"
mod = myNN.loadCNN(0)
def Main():
global guessGesture, visualize, mod, binaryMode, bkgrndSubMode, mask, takebkgrndSubMask, x0, y0, width, height, saveImg, gestname, path, flag
quietMode = False
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.5
fx = 10
fy = 350
fh = 18
#Call CNN model loading callback
mod = gCNN.loadCNN()
## Grab camera input
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
# set rt size as 640x480
ret = cap.set(3, 640)
ret = cap.set(4, 480)
framecount = 0
fps = ""
start = time.time()
plot = np.zeros((512, 512, 3), np.uint8)
while (True):
ret, frame = cap.read()
max_area = 0
frame = cv2.flip(frame, 3)
frame = cv2.resize(frame, (640, 480))
if ret == True:
if bkgrndSubMode == True:
roi = bkgrndSubMask(frame, x0, y0, width, height, framecount,
plot)
elif binaryMode == True:
roi = binaryMask(frame, x0, y0, width, height, framecount,
plot)
else:
roi = skinMask(frame, x0, y0, width, height, framecount, plot)
framecount = framecount + 1
end = time.time()
timediff = (end - start)
if (timediff >= 1):
#timediff = end - start
fps = 'FPS:%s' % (framecount)
start = time.time()
framecount = 0
cv2.putText(frame, fps, (10, 20), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'Options:', (fx, fy), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'b - Toggle Binary/SkinMask', (fx, fy + fh), font,
size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'x - Toggle Background Sub Mask', (fx, fy + 2 * fh),
font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'g - Toggle Prediction Mode', (fx, fy + 3 * fh),
font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'q - Toggle Quiet Mode', (fx, fy + 4 * fh), font,
size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'ESC - Exit', (fx, fy + 7 * fh), font, size,
(0, 255, 0), 1, 1)
## If enabled will stop updating the main openCV windows
## Way to reduce some processing power :)
if not quietMode:
cv2.imshow('Original', frame)
cv2.imshow('ROI', roi)
if guessGesture == True:
plot = np.zeros((512, 512, 3), np.uint8)
plot = gCNN.update(plot)
cv2.imshow('Gesture Probability', plot)
if gCNN.jsonarray:
gesture = max(gCNN.jsonarray, key=gCNN.jsonarray.get)
if gesture == 'PUNCH':
flag = 1
elif gesture == 'STOP':
flag = 2
elif gesture == 'NOTHING':
flag = 0
else:
flag = -1
pickle.dump(flag, open('gesture.p', 'wb'))
############## Keyboard inputs ##################
key = cv2.waitKey(5) & 0xff
## Use Esc key to close the program
if key == 27:
break
## Use b key to toggle between binary threshold or skinmask based filters
elif key == ord('b'):
binaryMode = not binaryMode
bkgrndSubMode = False
if binaryMode:
print("Binary Threshold filter active")
else:
print("SkinMask filter active")
## Use x key to use and refresh Background SubMask filter
elif key == ord('x'):
takebkgrndSubMask = True
bkgrndSubMode = True
print("BkgrndSubMask filter active")
## Use g key to start gesture predictions via CNN
elif key == ord('g'):
guessGesture = not guessGesture
print("Prediction Mode - {}".format(guessGesture))
## This option is not yet complete. So disabled for now
## Use v key to visualize layers
#elif key == ord('v'):
# visualize = True
## Use i,j,k,l to adjust ROI window
elif key == ord('i'):
y0 = y0 - 5
elif key == ord('k'):
y0 = y0 + 5
elif key == ord('j'):
x0 = x0 - 5
elif key == ord('l'):
x0 = x0 + 5
## Quiet mode to hide gesture window
elif key == ord('q'):
quietMode = not quietMode
print("Quiet Mode - {}".format(quietMode))
#elif key != 255:
# print key
#Realse & destroy
cap.release()
cv2.destroyAllWindows()
time.sleep(0.01)
return res, lastgesture
if __name__ == "__main__":
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
# set rt size as 640x480
ret = cap.set(3, 640)
ret = cap.set(4, 480)
print "Loading default weight file..."
mod = myNN.loadCNN()
while (True):
ret, frame = cap.read()
if ret == True:
roi, gesture = skinMask(frame, x0, y0, width, height)
cv2.imshow('Original', frame)
cv2.imshow('ROI', roi)
if gesture == -1:
pic = cv2.imread("Nothing.png")
cv2.imshow('gesture', pic)
else:
pic = cv2.imread(myNN.output[gesture] + '.png')
cv2.imshow('gesture', pic)
def main():
global model, binaryMode, x0, y0, x1, y1, width, height, saveImg, gestname, path
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.5
fx = 10
fy = 350
fh = 18
tracker1 = InputTracker()
tracker2 = InputTracker()
model = myNN.loadCNN(0)
## Grab camera input
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
# set rt size as 640x480
ret = cap.set(3,640)
ret = cap.set(4,480)
framecount = 0
fps = ""
start = time.time()
plot = np.zeros((512,512,3), np.uint8)
plot2 = np.zeros((512,512,3), np.uint8)
while(True):
ret, frame = cap.read()
max_area = 0
frame = cv2.flip(frame, 3)
frame = cv2.resize(frame, (640,480))
if ret == True:
timestamp = time.time()
roi, guess, plot = binaryMask(frame, x0, y0, width, height, False, framecount, plot)
roi2, guess2, plot2 = binaryMask(frame, x1, y1, width, height, True, framecount, plot2)
cv2.putText(frame, guess, (x0, y0), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, guess2, (x1, y1), font, 0.7, (0, 255, 0), 2, 1)
action = tracker1.update(guess, timestamp)
action2 = tracker2.update(guess2, timestamp)
if action != "":
cv2.putText(frame, action, (x0 + 100, y0), font, 0.7, (0, 255, 0), 2, 1)
if action == "DOWN":
tetris.action("down")
elif action == "FLIP":
tetris.action("flip")
elif action == "MOVE":
tetris.action("right")
if action2 != "":
cv2.putText(frame, action2, (x1 + 100, y1), font, 0.7, (0, 255, 0), 2, 1)
if action2 == "DOWN":
tetris.action("down")
elif action2 == "FLIP":
tetris.action("flip")
elif action2 == "MOVE":
tetris.action("left")
framecount = framecount + 1
timediff = timestamp - start
if( timediff >= 1):
#timediff = end - start
fps = 'FPS:%s' %(framecount)
start = time.time()
framecount = 0
cv2.putText(frame,fps,(10,20), font, 0.7,(0,255,0),2,1)
cv2.putText(frame,'n - To enter name of new gesture folder',(fx,fy + 5*fh), font, size,(0,255,0),1,1)
cv2.putText(frame,'s - To start capturing new gestures for training',(fx,fy + 6*fh), font, size,(0,255,0),1,1)
cv2.putText(frame,'ESC - Exit',(fx,fy + 7*fh), font, size,(0,255,0),1,1)
cv2.imshow('Original',frame)
cv2.imshow('ROI', roi)
cv2.imshow('ROI2', roi2)
# cv2.imshow('Gesture Probability', plot)
# cv2.imshow('Gesture Probability 2', plot2)
############## Keyboard inputs ##################
key = cv2.waitKey(5) & 0xff
## Use Esc key to close the program
if key == 27:
break
## Use s key to start/pause/resume taking snapshots
## numOfSamples controls number of snapshots to be taken PER gesture
elif key == ord('s'):
if gestname != '':
saveImg = True
else:
print("Enter a gesture group name first, by pressing 'n'")
saveImg = False
## Use n key to enter gesture name
elif key == ord('n'):
gestname = input("Enter the gesture name: ")
# Release & destroy
cap.release()
cv2.destroyAllWindows()
def main():
mod = myNN.loadCNN(-1)
myNN.trainModel(mod)
def Main():
global guessGesture, visualize, mod, binaryMode, x0, y0, width, height, saveImg, gestname, path, img_flag, cap
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.7
fx = 10
fy = 355
fh = 18
#Call CNN model loading callback
while True:
# ans = int(raw_input(banner))
ans = 1
if ans == 2:
mod = myNN.loadCNN(-1)
myNN.trainModel(mod)
raw_input("Press any key to continue")
break
elif ans == 1:
print "Will load default weight file"
mod = myNN.loadCNN(0)
break
elif ans == 3:
if not mod:
w = int(raw_input("Which weight file to load (0 or 1)"))
mod = myNN.loadCNN(w)
else:
print "Will load default weight file"
img = int(raw_input("Image number "))
layer = int(raw_input("Enter which layer to visualize "))
myNN.visualizeLayers(mod, img, layer)
raw_input("Press any key to continue")
continue
else:
print "Get out of here!!!"
return 0
import os
print os.getpid()
## Grab camera input
cv2.namedWindow(
'Original',
cv2.WND_PROP_FULLSCREEN) # Set Camera Size, cv2.WINDOW_NORMAL
cv2.setWindowProperty('Original', cv2.WND_PROP_FULLSCREEN,
cv2.WINDOW_FULLSCREEN)
import signal
# MAC : SIGPROF 27
def signalCameraOnHandler(signum, frame):
print signum
global img_flag
global cap
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
img_flag = "CAM"
# MAC : SIGINFO 29
def signalMapHandler(signum, frame):
print signum
global img_flag
global cap
cap = cv2.VideoCapture('map.gif')
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
img_flag = "MAP"
# MAC : SIGUSR1 30
def signalSmileFaceHandler(signum, frame):
print signum
global img_flag
global cap
cap = cv2.VideoCapture('smile_glow.gif')
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
img_flag = "SMILE"
# MAC : SIGUSR2 31
def signalDefaultFaceHandler(signum, frame):
print signum
global img_flag
global cap
cap = cv2.VideoCapture('normal_glow.gif')
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
img_flag = "NORMAL"
# Set Signal
signal.signal(signal.SIGPROF, signalCameraOnHandler)
signal.signal(signal.SIGINFO, signalMapHandler)
signal.signal(signal.SIGUSR1, signalSmileFaceHandler)
signal.signal(signal.SIGUSR2, signalDefaultFaceHandler)
reset_flag = False
thumbs_up_guess_stack = 0
smile_face_stack = 0
camera_page_stack = 0
map_page_stack = 0
while True:
ret, frame = cap.read()
# for end of gif
if not ret:
if img_flag == "NORMAL":
cap = cv2.VideoCapture('normal_glow.gif')
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
elif img_flag == "SMILE":
cap = cv2.VideoCapture('smile_glow.gif')
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
elif img_flag == "MAP":
cap = cv2.VideoCapture('map.gif')
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
ret, frame = cap.read()
if img_flag == "CAM":
camera_page_stack += 1
frame = cv2.flip(frame, 3)
guess_res = ":)"
if ret:
if binaryMode: # on
roi, guess_res = binaryMask(frame, x0, y0, width, height)
else:
roi, guess_res = skinMask(frame, x0, y0, width, height)
# detact thumbs-up frame
if guess_res.split()[0] == "PEACE" or guess_res.split()[0] == "OK":
camera_page_stack -= 1
thumbs_up_guess_stack += 1
else:
thumbs_up_guess_stack = 0
# print GUESS
cv2.putText(frame,
str(thumbs_up_guess_stack) + " " + guess_res, (fx, fy),
font, 1, (0, 0, 255), 2, 1)
cv2.putText(frame, "Show Thumbs-Up on TONY's hat camera!",
(fx, fy + fh), font, size, (0, 0, 255), 1, 1)
elif img_flag == "SMILE":
smile_face_stack += 1
cap = cv2.VideoCapture('smile_glow.gif')
elif img_flag == "NORMAL":
cap = cv2.VideoCapture('normal_glow.gif')
elif img_flag == "MAP":
map_page_stack += 1
cap = cv2.VideoCapture('map.gif')
# for signal call
if thumbs_up_guess_stack >= 6:
# paging to smile
os.system("kill -30 " + str(os.getpid()))
reset_flag = True
if smile_face_stack >= 30:
# paging to map
os.system("kill -29 " + str(os.getpid()))
reset_flag = True
if camera_page_stack >= 150:
# paging to map
os.system("kill -29 " + str(os.getpid()))
reset_flag = True
# now threshold == 300
if map_page_stack >= 200:
# paging to camera
os.system("kill -27 " + str(os.getpid()))
reset_flag = True
if reset_flag:
thumbs_up_guess_stack = 0
smile_face_stack = 0
camera_page_stack = 0
map_page_stack = 0
reset_flag = False
cv2.imshow('Original', frame)
key = cv2.waitKey(10) & 0xff
if key == 27:
break
#Realse & destroy
cap.release()
cv2.destroyAllWindows()
def Main():
# pdb.set_trace()
global guessGesture, visualize, mod, binaryMode, x0, y0, width, height, saveImg, gestname, path
quietMode = False
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.5
fx = 10
fy = 355
fh = 18
#Call CNN model loading callback
while True:
ans = int(raw_input( banner))
if ans == 2:
# 训练模型
mod = myNN.loadCNN(-1)
myNN.trainModel(mod)
raw_input("Press any key to continue")
break
elif ans == 1:
# 使用已有模型
print "Will load default weight file"
mod = myNN.loadCNN(0)
break
elif ans == 3:
# 打印某张手势图片经过每层的处理
if not mod:
w = int(raw_input("Which weight file to load (0 or 1)"))
mod = myNN.loadCNN(w)
else:
print "Will load default weight file"
img = int(raw_input("Image number "))
layer = int(raw_input("Enter which layer to visualize "))
myNN.visualizeLayers(mod, img, layer)
raw_input("Press any key to continue")
continue
else:
print "Get out of here!!!"
return 0
## Grab camera input
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
# set rt size as 640x480
ret = cap.set(3,640)
ret = cap.set(4,480)
while(True):
ret, frame = cap.read()
max_area = 0
frame = cv2.flip(frame, 3)
if ret == True:
# ! 其间 可以做保存图片的操作
if binaryMode == True:
roi = binaryMask(frame, x0, y0, width, height)
else:
roi = skinMask(frame, x0, y0, width, height)
cv2.putText(frame,'Options:',(fx,fy), font, 0.7,(0,255,0),2,1)
cv2.putText(frame,'b - Toggle Binary/SkinMask',(fx,fy + fh), font, size,(0,255,0),1,1)
cv2.putText(frame,'g - Toggle Prediction Mode',(fx,fy + 2*fh), font, size,(0,255,0),1,1)
cv2.putText(frame,'q - Toggle Quiet Mode',(fx,fy + 3*fh), font, size,(0,255,0),1,1)
cv2.putText(frame,'n - To enter name of new gesture folder',(fx,fy + 4*fh), font, size,(0,255,0),1,1)
cv2.putText(frame,'s - To start capturing new gestures for training',(fx,fy + 5*fh), font, size,(0,255,0),1,1)
cv2.putText(frame,'ESC - Exit',(fx,fy + 6*fh), font, size,(0,255,0),1,1)
## If enabled will stop updating the main openCV windows
## Way to reduce some processing power :)
if not quietMode:
cv2.imshow('Original',frame) # 图像窗口
cv2.imshow('ROI', roi) # 截取手势窗口
# Keyboard inputs
key = cv2.waitKey(10) & 0xff
## Use Esc key to close the program
if key == 27:
break
## Use b key to toggle between binary threshold or skinmask based filters
elif key == ord('b'):
binaryMode = not binaryMode
if binaryMode:
print "Binary Threshold filter active"
else:
print "SkinMask filter active"
## Use g key to start gesture predictions via CNN
elif key == ord('g'):
guessGesture = not guessGesture
print "Prediction Mode - {}".format(guessGesture)
## This option is not yet complete. So disabled for now
## Use v key to visualize layers
# elif key == ord('v'):
# visualize = True
## Use i,j,k,l to adjust ROI window
elif key == ord('i'):
y0 = y0 - 5
elif key == ord('k'):
y0 = y0 + 5
elif key == ord('j'):
x0 = x0 - 5
elif key == ord('l'):
x0 = x0 + 5
## Quiet mode to hide gesture window
elif key == ord('q'):
quietMode = not quietMode
print "Quiet Mode - {}".format(quietMode)
## Use s key to start/pause/resume taking snapshots
## numOfSamples controls number of snapshots to be taken PER gesture
elif key == ord('s'):
saveImg = not saveImg
if gestname != '':
# 保存图片操作 binaryMask 或 skinMask 步骤操作
saveImg = True
else:
print "Enter a gesture group name first, by pressing 'n'"
saveImg = False
## Use n key to enter gesture name
elif key == ord('n'):
# 获取 手势名称 建立文件夹，然后会自动保存图片 到该目录
gestname = raw_input("Enter the gesture folder name: ")
try:
os.makedirs(gestname)
except OSError as e:
# if directory already present
if e.errno != 17:
print 'Some issue while creating the directory named -' + gestname
path = "./"+gestname+"/"
#elif key != 255:
# print key
#Realse & destroy
cap.release()
cv2.destroyAllWindows()
def Main():
global flip, change_bgd, binary_mode, guess_gesture, visualize, mod, bgdSub_mode, mask, change_bgd
global x0, y0, width, height, save_image, gest_name, path
quite_mode = False
while True:
ans = int(input(banner))
if ans == 1:
mod = nn.loadCNN()
break
else:
print('BYE!')
return 0
cap = cv.VideoCapture(0)
cv.namedWindow('Original', cv.WINDOW_NORMAL)
ret = cap.set(3, 640)
ret = cap.set(4, 480)
framecount = 0
fps = ""
start = time.time()
while True:
ret, frame = cap.read()
frame = cv.flip(frame, 3)
frame = cv.resize(frame, (640, 480))
cv.rectangle(frame, (x0, y0), (x0+width, y0+height), (0,255,0), 1)
roi = frame[y0:y0+height, x0:x0+width]
if ret == True:
if bgdSub_mode:
roi = bgdSubMask(roi, framecount)
elif binary_mode:
roi = binaryMask(roi, framecount)
else:
roi = skinMask(roi, framecount)
framecount += 1
end = time.time()
if(end - start >= 1):
fps = 'FPS:%s' %(framecount)
start = time.time()
framecount = 0
font = cv.FONT_HERSHEY_COMPLEX_SMALL
size = 0.7
fx, fy, fh = 10, 350, 18
color = (0, 255, 0)
cv.putText(frame, fps, (10,20), font, size, color, 2, 1)
cv.putText(frame, 'Options:', (fx, fy), font, size, color, 2, 1)
cv.putText(frame, 'b- Toggle Binary/SkinMask', (fx, fy+fh), font, size, color, 2, 1)
cv.putText(frame, 'x- Toggle Background SubMask:', (fx, fy+2*fh), font, size, color, 2, 1)
cv.putText(frame, 'g- Toggle Prediction Mode', (fx, fy+3*fh), font, size, color, 2, 1)
cv.putText(frame, 'q- Toggle Quiet Mode', (fx, fy+4*fh), font, size, color, 2, 1)
cv.putText(frame, 'n- To enter name of new gesture folder', (fx, fy+5*fh), font, size, color, 2, 1)
cv.putText(frame, 's- To start capturing new gestures', (fx, fy+6*fh), font, size, color, 2, 1)
cv.putText(frame, 'Esc - exit', (fx, fy+7*fh), font, size, color, 1, 1)
if flip:
roi = roi[:,-1::-1]
if save_image:
saveROI(roi)
if not quite_mode:
cv.imshow('Original', frame)
cv.imshow('ROI', cv.resize(roi, (400, 400)))
if guess_gesture:
plot = np.zeros((512, 512, 3), np.uint8)
plot = nn.update(plot)
cv.imshow('Gesture Probability', plot)
##Key functions##
key = cv.waitKey(5) & 0xff
if key == 27:
break
elif key == ord('b'):
binary_mode = not binary_mode
bgdSub_mode = False
if binary_mode:
print("Binary Threshold filter active")
else:
print("SkinMask filter active")
elif key == ord('x'):
change_bgd = True
bgdSub_mode = True
print("BgdSubMask filter active")
elif key == ord('g'):
guess_gesture = not guess_gesture
print("Prediction Mode - {}".format(guess_gesture))
elif key == ord('i'):
y0 = y0 - 5
elif key == ord('k'):
y0 = y0 + 5
elif key == ord('j'):
x0 = x0 - 5
elif key == ord('l'):
x0 = x0 + 5
elif key == ord('q'):
quite_mode = not quite_mode
print("Quiet Mode - {}".format(quite_mode))
elif key == ord('s'):
save_image = not save_image
if gest_name != '':
save_image = True
else:
print("Enter a gesture group name first, by pressing 'n'")
save_image = False
elif key == ord('n'):
gest_name = input("Enter the gesture folder name: ")
if not os.path.isdir(os.path.join('./test' ,gest_name)):
os.makedirs(os.path.join('./test' ,gest_name))
path = "./test/"+gest_name+"/"
elif key == ord('f'):
flip = not flip
if flip:
print('Flipping ON...')
else:
print('Flipping OFF...')
cap.release()
cv.destroyAllWindows()
def __init__(self):
self.mod=myNN.loadCNN(0)
self.skinkernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(5,5))
def Main():
global guessGesture, visualize, mod, binaryMode, bkgrndSubMode, mask, takebkgrndSubMask, x0, y0, width, height, saveImg, gestname, path
quietMode = False
font = cv2.FONT_HERSHEY_SIMPLEX
size = 0.5
fx = 10
fy = 350
fh = 18
mod = myNN.loadCNN() #LOAD THE MODEL
cap = cv2.VideoCapture(0)
cv2.namedWindow('Original', cv2.WINDOW_NORMAL)
# set rt size as 640x480
ret = cap.set(3, 640)
ret = cap.set(4, 480)
framecount = 0
fps = ""
start = time.time()
plot = np.zeros((512, 512, 3), np.uint8)
while (True):
ret, frame = cap.read()
max_area = 0
frame = cv2.flip(frame, 3)
frame = cv2.resize(frame, (640, 480))
if ret == True:
if bkgrndSubMode == True:
roi = bkgrndSubMask(frame, x0, y0, width, height, framecount,
plot)
elif binaryMode == True:
roi = binaryMask(frame, x0, y0, width, height, framecount,
plot)
else:
roi = skinMask(frame, x0, y0, width, height, framecount, plot)
framecount = framecount + 1
end = time.time()
timediff = (end - start)
if (timediff >= 1):
#timediff = end - start
fps = 'FPS:%s' % (framecount)
start = time.time()
framecount = 0
cv2.putText(frame, fps, (10, 20), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'Options:', (fx, fy), font, 0.7, (0, 255, 0), 2, 1)
cv2.putText(frame, 'b - Toggle Binary/SkinMask', (fx, fy + fh), font,
size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'x - Toggle Background Sub Mask', (fx, fy + 2 * fh),
font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'g - Toggle Prediction Mode', (fx, fy + 3 * fh),
font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'q - Toggle Quiet Mode', (fx, fy + 4 * fh), font,
size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'n - To enter name of new gesture folder',
(fx, fy + 5 * fh), font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 's - To start capturing new gestures for training',
(fx, fy + 6 * fh), font, size, (0, 255, 0), 1, 1)
cv2.putText(frame, 'ESC - Exit', (fx, fy + 7 * fh), font, size,
(0, 255, 0), 1, 1)
if not quietMode:
cv2.imshow('Original', frame)
cv2.imshow('ROI', roi)
if guessGesture == True:
f = open("./LivePlot/sample.txt", 'w')
myNN.update(f)
key = cv2.waitKey(5) & 0xFF
## Use Esc key to close the program
if key == 27:
break
## Use b key to toggle between binary threshold or skinmask based filters
elif key == ord('b'):
binaryMode = not binaryMode
bkgrndSubMode = False
if binaryMode:
print("Binary Threshold filter active")
else:
print("SkinMask filter active")
## Use x key to use and refresh Background SubMask filter
elif key == ord('x'):
takebkgrndSubMask = True
bkgrndSubMode = True
print("BkgrndSubMask filter active")
## Use g key to start gesture predictions via CNN
elif key == ord('g'):
guessGesture = not guessGesture
print("Prediction Mode - {}".format(guessGesture))
elif key == ord('i'):
y0 = y0 - 5
elif key == ord('k'):
y0 = y0 + 5
elif key == ord('j'):
x0 = x0 - 5
elif key == ord('l'):
x0 = x0 + 5
## Quiet mode to hide gesture window
elif key == ord('q'):
quietMode = not quietMode
print("Quiet Mode - {}".format(quietMode))
## Use s key to start/pause/resume taking snapshots
## numOfSamples controls number of snapshots to be taken PER gesture
elif key == ord('s'):
saveImg = not saveImg
if gestname != '':
saveImg = True
else:
print("Enter a gesture group name first, by pressing 'n'")
saveImg = False
## Use n key to enter gesture name
elif key == ord('n'):
gestname = input("Enter the gesture folder name: ")
try:
os.makedirs(gestname)
except OSError as e:
if e.errno != 17:
print('Some issue while creating the directory named -' +
gestname)
path = "./" + gestname + "/"
cap.release()
cv2.destroyAllWindows()
