def from_stream():
fps = FPS().start()
cam = WebcamVideoStream().start()
max_frames = 50
i = 0
while True:
frame = cam.read()
if i > max_frames:
fps.stop()
print(fps.elapsed())
print(fps.fps())
break
i += 1
testcone(frame, stream=True)
fps.update()
cv2.imshow('', frame)
cv2.waitKey(1)
class VideoStream:
def __init__(self,
src=0,
usePiCamera=False,
resolution=(320, 240),
framerate=32,
image_flip=0,
**kwargs):
self.image_flip = 1
# check to see if the picamera module should be used
if usePiCamera:
from pi_camera import PiVideoStream
# initialize picamera stream and let the camera sensor to warmup
self.stream = PiVideoStream(resolution=resolution,
framerate=framerate,
**kwargs)
# otherwise, we are using OpenCV so initialize the webcam stream
else:
self.stream = WebcamVideoStream(src=src,
resolution=resolution,
framerate=framerate)
def start(self):
# start the threaded video stream
return self.stream.start()
def read(self):
# return the current frame
return self.flip_if_needed(self.stream.read())
def read_jpg(self):
frame = self.flip_if_needed(self.stream.read())
ret, jpeg = cv2.imencode('.jpg', frame)
return jpeg.tobytes()
def flip_if_needed(self, frame):
if self.image_flip:
return np.fliplr(frame)
return frame
def stop(self):
# stop the thread and release any resources
self.stream.stop()
def main():
# Create file queue
file_queue = queue.Queue()
# Initialize video stream
vs = WebcamVideoStream(src=0).start()
# Initialize image writer
ir = ImageWriter(queue=file_queue).start()
name = input("Please enter name of individual: ")
while True:
# Read most recent video frame
frame = vs.read()
# Copy frame to display and annotate
frame_copy = frame.copy()
# Overlay name to frame
cv2.putText(frame_copy, name, (25, 25), cv2.FONT_HERSHEY_SIMPLEX, 0.8,
(0, 255, 0), 2, cv2.LINE_AA)
# Display current frame
cv2.imshow("Video", frame_copy)
# Send frame to image writer thread
key = cv2.waitKey(1) & 0xFF
# Enter new name
if key == ord("n"):
name = input("Please enter name of individual: ")
# Capture picture and write to disk
if key == ord("p"):
file_queue.put((name, frame))
# Quit application
if key == ord("q"):
cv2.destroyAllWindows()
break
# Stop the image writing thread
ir.stop()
# Stop the video stream thread
vs.stop()
class VideoStream:
def __init__(self,
src=0,
usePiCamera=False,
resolution=(320, 240),
framerate=32,
rotation=0):
# check to see if the picamera module should be used
if usePiCamera:
# only import the picamera packages unless we are
# explicity told to do so -- this helps remove the
# requirement of `picamera[array]` from desktops or
# laptops that still want to use the `imutils` package
from picam import PiVideoStream
# initialize the picamera stream and allow the camera
# sensor to warmup
self.stream = PiVideoStream(resolution=resolution,
framerate=framerate,
rotation=rotation)
# otherwise, we are using OpenCV so initialize the webcam
# stream
else:
self.stream = WebcamVideoStream(src=src)
def start(self):
# start the threaded video stream
return self.stream.start()
def update(self):
# grab the next frame from the stream
self.stream.update()
def read(self):
# return the current frame
return self.stream.read()
def stop(self):
# stop the thread and release any resources
self.stream.stop()
def test_cam(args):
"""Function to predict for a camera image stream
"""
if torch.cuda.is_available() and not args.no_cuda:
device = torch.device("cuda")
else:
device = torch.device("cpu")
download_model_if_doesnt_exist(args.model_name)
model_path = os.path.join("models", args.model_name)
print("-> Loading model from ", model_path)
encoder_path = os.path.join(model_path, "encoder.pth")
depth_decoder_path = os.path.join(model_path, "depth.pth")
# LOADING PRETRAINED MODEL
print(" Loading pretrained encoder")
encoder = networks.ResnetEncoder(18, False)
loaded_dict_enc = torch.load(encoder_path, map_location=device)
# Extract the height and width of image that this model was trained with
feed_height = loaded_dict_enc['height']
feed_width = loaded_dict_enc['width']
filtered_dict_enc = {
k: v
for k, v in loaded_dict_enc.items() if k in encoder.state_dict()
}
encoder.load_state_dict(filtered_dict_enc)
encoder.to(device)
encoder.eval()
print(" Loading pretrained decoder")
depth_decoder = networks.DepthDecoder(num_ch_enc=encoder.num_ch_enc,
scales=range(4))
loaded_dict = torch.load(depth_decoder_path, map_location=device)
depth_decoder.load_state_dict(loaded_dict)
depth_decoder.to(device)
depth_decoder.eval()
print("-> Loading complete, initializing the camera")
# Initialize camera to capture image stream
# Change the value to 0 when using default camera
video_stream = WebcamVideoStream(src=args.webcam).start()
if not args.no_display:
# Object to display images
image_display = DisplayImage(not args.no_process)
# Flag that records when 'q' is pressed to break out of inference loop below
quit_inference = False
def on_release(key):
if key == keyboard.KeyCode.from_char('q'):
nonlocal quit_inference
quit_inference = True
return False
keyboard.Listener(on_release=on_release).start()
# Number of frames to capture to calculate fps
num_frames = 5
curr_time = np.zeros(num_frames)
with torch.no_grad():
while True:
if quit_inference:
if args.no_display:
print('-> Done')
break
# Capture frame-by-frame
frame = video_stream.read()
# Calculate the fps
curr_time[1:] = curr_time[:-1]
curr_time[0] = time.time()
fps = num_frames / (curr_time[0] - curr_time[len(curr_time) - 1])
# Our operations on the frame come here
input_image = pil.fromarray(frame).convert('RGB')
original_width, original_height = input_image.size
input_image = input_image.resize((feed_width, feed_height),
pil.LANCZOS)
input_image = transforms.ToTensor()(input_image).unsqueeze(0)
# PREDICTION
input_image = input_image.to(device)
features = encoder(input_image)
outputs = depth_decoder(features)
disp = outputs[("disp", 0)]
disp_resized = torch.nn.functional.interpolate(
disp, (original_height, original_width), mode="nearest")
# Get the predict depth
scaled_disp, pred_depth = disp_to_depth(disp_resized, 0.1, 100)
pred_depth_np = pred_depth.squeeze().cpu().detach().numpy()
# Initialize a 3x4 depth map
depth_map = np.zeros([3, 4])
for i in range(len(depth_map)):
for j in range(len(depth_map[0])):
# Cut and store the average value of depth information of 640x480 into 3x4 grid
depth_map[i][j] = get_avg_depth(pred_depth_np, 160 * i,
160 * j, 160 * i + 160,
160 * j + 160)
# Giving a simple decision logic
if depth_map[0, 1] <= 1 or depth_map[1, 1] <= 1 or depth_map[
0, 2] <= 1 or depth_map[1, 2] <= 1:
if depth_map[1, 1] <= 1 and depth_map[1, 2] <= 1:
print("Dangerous!!! AHEAD")
else:
if depth_map[0, 1] <= 1 or depth_map[1, 1] <= 1:
print("Dangerous!!! LEFT")
if depth_map[0, 2] <= 1 or depth_map[1, 2] <= 1:
print("Dangerous!!! RIGHT")
elif np.sum(depth_map[0:2, 2:3]) <= 7 or np.sum(
depth_map[0:2, 2:3]) <= 7:
if np.sum(depth_map[0:2, 0:1]) <= 7:
print("Careful!! LEFT")
if np.sum(depth_map[0:2, 2:3]) <= 7:
print("Careful!! RIGHT")
else:
print("Clear")
if not args.no_display:
# DISPLAY
# Generate color-mapped depth image
disp_resized_np = disp_resized.squeeze().cpu().detach().numpy()
image_display.display(frame,
disp_resized_np,
fps,
original_width,
original_height,
blended=not args.no_blend)
else:
print(f"FPS: {fps}")
# if quit_inference:
# if args.no_display:
# print('-> Done')
# break
# When everything is done, stop camera stream
video_stream.stop()
webcam_thread = WebcamVideoStream(src=1).start()
fps = FPS().start()
#logger.info("Starting inference thread")
INFERENCE_API = ""
inference_thread = InferenceHelper(INFERENCE_API).start()
#logger.info("Starting drawing thread")
drawing_thread = DrawingHelper(webcam_thread.VID_WIDTH,
webcam_thread.VID_HEIGHT).start()
# Run until quit.
# define a new function that we're going to run on a seperate thread:
frame = webcam_thread.read()
frame_name = "./output/frame{}.jpg".format(fps.current_frame_number())
similarity_list = [frame]
"""Need to be tested"""
sim_bucket = []
sim_thresh_list = [frame]
sim_threshold = threshold(sim_thresh_list)
print("threshold: ", sim_threshold)
# Everytime the inference thread processes a frame it returns the json_resp
# This then adds the frame and json_resp to the drawing thread queue.
if inference_thread.queue.empty():
inference_thread.enqueue({
'name': frame_name,
'frame': frame,
class HandTracker:
def __init__(self):
self.webcam = WebcamVideoStream()
self.webcam.start()
self.detection = Detection()
#hand gesture status
self.is_okay = False
self.is_vhand = False
self.is_phand = False
self.is_palm = False
self.is_fist = False
#get location of hand when tracking
self.x_axis = 0.0
self.y_axis = 0.0
#get tracker location
self.yprime_axis = 0
self.xprime_axis = 0
#get middle of frame
self.x_center = (self.webcam.camera().get(
cv2.CAP_PROP_FRAME_WIDTH)) / 2
self.y_center = (self.webcam.camera().get(
cv2.CAP_PROP_FRAME_HEIGHT)) / 2
def _start_up(self):
while self.is_okay == False:
# get image from webcam
image = self.webcam.read()
print("WAITING FOR OK SIGN")
# look for the OK sign to start up
self.is_okay = self.detection.is_item_detected_in_image(
'data/ok_cascade_48x30.xml', image)
if self.is_okay:
# recognized OK Sign
print("OK GESTURE Detected ")
self.is_okay = False
# move to modes stage
self._modes()
if cv2.waitKey(1) == 27:
self._shut_down()
break
def _modes(self):
# Look to recognize a gesture
while True:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
#self.is_phand = self.detection.is_item_detected_in_image('data/face.xml', image )
self.is_fist = self.detection.is_item_detected_in_image(
'data/fist.xml', image)
#self.is_vhand =self.detection.is_item_detected_in_image('data/.xml', image )
#self.is_palm =self.detection.is_item_detected_in_image('data/palm.xml', image )
#check which hand gesture detected
#Fist hand gesture
if self.is_fist:
self.is_fist = False
print("Fist detected, See if it moved")
self.x_axis = self.detection.x_axis
self.y_axis = self.detection.y_axis
self._keepCenter(self.x_axis, self.y_axis)
#Phand gesture
if self.is_phand:
self.is_phand = False
print("Phand detected, See if it moved")
self.x_axis = self.detection.x_axis
self._moveFocus(self.x_axis)
#Vhand gesture
if self.is_vhand:
self.is_vhand = False
print("Vhand detected, See if it moved")
self.x_axis = self.detection.x_axis
self._changeLight(self.x_axis)
#Palm gesture
if self.is_palm:
self.is_palm = False
print("Palm detected, See if it moved")
self.x_axis = self.detection.x_axis
self._powerLight(self)
#Escape from program
if cv2.waitKey(1) == 27:
self._shut_down()
break
return
def _powerButton(self):
#turn on/off L.E.D
return
def _keepCenter(self, x_axis, y_axis):
flag = True
x = 25
while x != 0:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
self.is_fist = self.detection.is_item_detected_in_image(
'data/face.xml', image)
#check which hand gesture detected
if self.is_fist:
x = 25
self.is_fist = False
#Get new position x and y positions
self.xprime_axis = self.detection.x_axis
self.yprime_axis = self.detection.y_axis
#If the new position is different from the intial position take the absoute value to find the difference
if ((self.xprime_axis != self.x_axis
or self.yprime_axis != self.y_axis) and (flag == True)):
dx = self.xprime_axis - self.x_axis
dx = abs(dx)
dy = self.yprime_axis - self.y_axis
dy = abs(dy)
print("x_axis: ", self.x_axis, " y_axis: ", self.y_axis)
print("Face MOVED xprime_axis: ", self.xprime_axis,
"FACE moved yprime_axis: ", self.yprime_axis)
print("dx: ", dx, "dy: ", dy)
#if above threshold for movement
if (dx >= 20 or dy >= 20):
#Hand Ready to be Tracked
flag = False
#If the new position is not equal to the center of the screen continue
if ((self.xprime_axis != self.x_center
or self.yprime_axis != self.y_center)
and (flag == False)):
print("Face centering Going on")
# Calculate how far away from the center
dx = self.xprime_axis - self.x_center
dx = abs(dx)
dy = self.yprime_axis - self.y_center
dy = abs(dy)
print("Face MOVED CENTERX_axis: ", self.x_center,
"FACE moved CENTERY_axis: ", self.y_center)
print("Face MOVED xprime_axis: ", self.xprime_axis,
"FACE moved yprime_axis: ", self.yprime_axis)
#if above threshold for movement
if (dx >= 20 or dy >= 20):
print("Movement of Motors")
self._moveMotors(self.xprime_axis, self.yprime_axis,
dx, dy)
else:
print("No Gesture Detected")
x = x - 1
if cv2.waitKey(1) == 27:
self._shut_down()
break
print("___________****TIME OUT*****__________")
self._start_up()
return
def _moveFocus(self, x_axis):
flag = True
x = 25
while x != 0:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
self.is_phand = self.detection.is_item_detected_in_image(
'data/fist.xml', image)
#check which hand gesture detected
if self.is_phand:
x = 25
self.is_phand = False
#Get the new x position
self.xprime_axis = self.detection.x_axis
#If the new x position is different from the intial position take the absoute value to find the difference
if ((self.xprime_axis != self.x_axis) and (flag == True)):
dx = self.xprime_axis - self.x_axis
dx = abs(dx)
print("x_axis: ", self.x_axis)
print("Phand MOVED xprime_axis: ", self.xprime_axis)
print("dx: ", dx)
#if above threshold for movement
if (dx >= 20):
#Hand Ready to be Tracked
flag = False
#If the new position is not equal to the center of the screen continue
if ((self.xprime_axis != self.x_center) and (flag == False)):
print("Phand centering Going on")
# Calculate how far away from the center
dx = self.xprime_axis - self.x_center
dx = abs(dx)
print("Phand MOVED CENTERX_axis: ", self.x_center)
print("Phand MOVED xprime_axis: ", self.xprime_axis)
#if above threshold for movement
if (dx >= 20):
print("Adjust Focus")
self._moveMotors(self.xprime_axis, -1, dx, -1)
else:
print("No Gesture Detected")
x = x - 1
if cv2.waitKey(1) == 27:
self._shut_down()
break
print("___________****TIME OUT*****__________")
self._start_up()
return
def _changeLight(self, x_axis):
x = 25
while x != 0:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
self.is_vhand = self.detection.is_item_detected_in_image(
'data/fist.xml', image)
#check which hand gesture detected
if self.is_vhand:
x = 25
self.is_vhand = False
#Get new x position
self.xprime_axis = self.detection.x_axis
#If the new position is different from the intial position take the absoute value to find the difference
if ((self.xprime_axis != self.x_axis) and (flag == True)):
dx = self.xprime_axis - self.x_axis
dx = abs(dx)
print("x_axis: ", self.x_axis)
print("Vhand MOVED xprime_axis: ", self.xprime_axis)
print("dx: ", dx)
#if above threshold for movement
if (dx >= 20):
#Hand Ready to be Tracked
flag = False
#If the new position is not equal to the center of the screen continue
if ((self.xprime_axis != self.x_center) and (flag == False)):
print("Vhand centering Going on")
#Calculate how far away from the center
dx = self.xprime_axis - self.x_center
dx = abs(dx)
print("Vhand MOVED CENTERX_axis: ", self.x_center)
print("Vhand MOVED xprime_axis: ", self.xprime_axis)
#if above threshold for movement
if (dx >= 20):
print("Adjust Light Intensity")
self._moveMotors(self.xprime_axis, -1, dx, -1)
else:
print("No Gesture Detected")
x = x - 1
if cv2.waitKey(1) == 27:
self._shut_down()
break
print("___________****TIME OUT*****__________")
self._start_up()
return
def _powerLight(self):
print(" LIGHTS ON/OFF ")
return
def _moveMotors(xpos, ypos, dx, dy):
xcounter = 0
ycounter = 0
#If the new position is to the left of the center
if (xpos < x_center):
#Increase the motor
print("")
if (xcounter < dx):
#MOTOR INCREASE FUNCTION
print("")
increase()
xcounter = xcounter + 1
#If the new position is to the right of the center
elif (xpos > x_center):
#Decrease the motor
print("")
if (xcounter < dx):
#MOTOR DECREASE FUNCTION
print("")
decrease()
xcounter = xcounter + 1
#If the new position is above the center
if ((ypos < y_center) and (ypos != -1)):
#Increase the MOTOR
if (ycounter < dy):
print("")
#MOTOR INCREASE FUNCTION
increase()
ycounter = ycounter + 1
#If the new position is below the centering
elif ((ypos > y_center) and (ypos != -1)):
print("")
#Decrease the Motor
if (ycounter < dy):
#MOTOR DECREASE FUNCTION
print("")
decrease()
ycounter = ycounter + 1
return
#stops webcam and return camera
def _shut_down(self):
self.webcam.stop()
self.webcam.stream.release()
def main(self):
# setup and run OpenGL
return
class HandTracker:
def __init__(self):
self.webcam = WebcamVideoStream()
self.webcam.start()
self.detection = Detection()
#self.servoYaw = Servo(205,409,0)
#self.servoPitch = Servo(205,409,4)
#self.servoYaw.neutralPos()
#self.servoPitch.neutralPos()
self.system = System()
self.system.servoSetAllNeutral()
#lights instances
#hand gesture status
self.is_okay = False
self.is_vhand = False
self.is_phand = False
self.is_palm = False
self.is_fist = False
#get location of hand when tracking
self.x_axis = 0.0
self.y_axis = 0.0
#get tracker location
self.yprime_axis = 0
self.xprime_axis = 0
#get middle of frame
self.x_center = (self.webcam.camera().get(cv2.CAP_PROP_FRAME_WIDTH))/2;
self.y_center = (self.webcam.camera().get(cv2.CAP_PROP_FRAME_HEIGHT))/2;
def _start_up(self):
while self.is_okay == False:
# get image from webcam
image = self.webcam.read()
print("WAITING FOR OK SIGN")
# look for the OK sign to start up
self.is_okay = self.detection.is_item_detected_in_image('data/ok_cascade_48x30.xml', image.copy(), 4 )
if self.is_okay:
# recognized OK Sign
print("OK GESTURE Detected ")
self.system.turnOnGreen()
#self._delay(700)
#self.system.turnOffGreen()
self.is_okay = False
# move to modes stage
self._modes()
else:
self.system.blinkRed()
if cv2.waitKey(1) == 27 :
self._shut_down()
break
def _modes (self):
self.system.turnOffGreen()
# Look to recognize a gesture
while True:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
self.is_phand = self.detection.is_item_detected_in_image('data/phand_cascade.xml', image, 30 )
#phand_cascade
self.is_fist =self.detection.is_item_detected_in_image('data/fist.xml', image, 4 )
#fist.xml
self.is_vhand =self.detection.is_item_detected_in_image('data/vhand_cascade.xml', image,4 )
#vhand_cascade
self.is_palm =self.detection.is_item_detected_in_image('data/goodpalm.xml', image ,20)
#goodpalm
#check which hand gesture detected
#Fist hand gesture
if self.is_fist:
self.system.blinkGreen()
self.is_fist = False
print("Fist detected, See if it moved" )
self.x_axis= self.detection.x_axis
self.y_axis=self.detection.y_axis
self._keepCenter(self.x_axis,self.y_axis)
#Phand gesture
if self.is_phand:
self.system.blinkGreen()
self.is_phand = False
print("Phand detected, See if it moved" )
self.x_axis= self.detection.x_axis
self._moveFocus(self.x_axis)
#Vhand gesture
if self.is_vhand:
self.system.blinkGreen()
self.is_vhand = False
print("Vhand detected, See if it moved" )
self.x_axis= self.detection.x_axis
self._powerLight()
#Palm gesture
if self.is_palm:
self.system.blinkGreen()
self.is_palm = False
print("Palm detected, See if it moved" )
self.x_axis= self.detection.x_axis
self._changeLight(self.x_axis)
#Escape from program
if cv2.waitKey(1) == 27 :
self._modes()
break
return
def _powerButton(self):
#turn on/off L.E.D
return
def _keepCenter(self , x_axis ,y_axis):
self.system.turnOnLaser()
flag = True
x = 50
while x != 0:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
self.is_fist = self.detection.is_item_detected_in_image('data/fist.xml', image, 5)
#check which hand gesture detected
if self.is_fist:
self.system.blinkGreen()
x = 50
self.is_fist = False
#Get new position x and y positions
self.xprime_axis= self.detection.x_axis
self.yprime_axis=self.detection.y_axis
#If the new position is different from the intial position take the absoute value to find the difference
if ((self.xprime_axis != self.x_axis or self.yprime_axis != self.y_axis) and (flag == True)):
dx = self.xprime_axis - self.x_axis
dx = abs(dx)
dy = self.yprime_axis - self.y_axis
dy = abs(dy)
print("x_axis: " , self.x_axis , " y_axis: " , self.y_axis)
print("Fist MOVED xprime_axis: " , self.xprime_axis , "Fist moved yprime_axis: " , self.yprime_axis)
print("dx: ", dx , "dy: ", dy)
#if above threshold for movement
if( dx >= 15 or dy >= 15 ):
#Hand Ready to be Tracked
flag = False
#If the new position is not equal to the center of the screen continue
if ((self.xprime_axis != self.x_center or self.yprime_axis != self.y_center) and (flag == False)):
print("Fist centering Going on")
# Calculate how far away from the center
dx = self.xprime_axis - self.x_center
dx = abs(dx)
dy = self.yprime_axis - self.y_center
dy = abs(dy)
print("Fist MOVED CENTERX_axis: " , self.x_center , "Fist moved CENTERY_axis: " , self.y_center)
print("Fist MOVED xprime_axis: " , self.xprime_axis , "Fist moved yprime_axis: " , self.yprime_axis)
#if above threshold for movement
if( dx >= 15 or dy >= 15 ):
print("Movement of Motors")
self._moveMotors(self.xprime_axis,self.yprime_axis,dx,dy)
else:
print("No Gesture Detected")
self.system.blinkRed()
x = x-1
if cv2.waitKey(1) == 27 :
self._shut_down()
break
print("___________****TIME OUT*****__________")
self.system.blinkRed()
self.system.blinkRed()
self.system.turnOffLaser()
self._start_up()
return
def _moveFocus(self , x_axis):
flag = True
x = 50
while x != 0:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
self.is_phand = self.detection.is_item_detected_in_image('data/phand_cascade.xml', image, 5 )
#check which hand gesture detected
if self.is_phand:
x = 50
self.is_phand = False
#Get the new x position
self.xprime_axis= self.detection.x_axis
#If the new x position is different from the intial position take the absoute value to find the difference
if ((self.xprime_axis != self.x_axis) and (flag == True)):
dx = self.xprime_axis - self.x_axis
dx = abs(dx)
print("x_axis: " , self.x_axis)
print("Phand MOVED xprime_axis: " , self.xprime_axis)
print("dx: ", dx)
#if above threshold for movement
if( dx >= 15):
#Hand Ready to be Tracked
flag = False
#If the new position is not equal to the center of the screen continue
if ((self.xprime_axis != self.x_center) and (flag == False)):
print("Phand centering Going on")
# Calculate how far away from the center
dx = self.xprime_axis - self.x_center
dx = abs(dx)
print("Phand MOVED CENTERX_axis: " , self.x_center)
print("Phand MOVED xprime_axis: " , self.xprime_axis)
#if above threshold for movement
if( dx >= 15):
print("Adjust Focus")
self._moveSpotSize(self.xprime_axis,dx)
else:
print("No Gesture Detected")
x = x-1
if cv2.waitKey(1) == 27 :
self._shut_down()
break
print("___________****TIME OUT*****__________")
self.system.blinkRed()
self._start_up()
return
def _changeLight(self , x_axis):
flag = True
x = 50
while x != 0:
# get image from webcam
image = self.webcam.read()
#different classifier for different modes
self.is_palm = self.detection.is_item_detected_in_image('data/goodpalm.xml', image, 12 )
#check which hand gesture detected
if self.is_palm:
self.system.blinkGreen()
x = 50
self.is_vhand = False
#Get new x position
self.xprime_axis= self.detection.x_axis
#If the new position is different from the intial position take the absoute value to find the difference
if ((self.xprime_axis != self.x_axis) and (flag == True)):
dx = self.xprime_axis - self.x_axis
dx = abs(dx)
print("x_axis: " , self.x_axis)
print("Palm MOVED xprime_axis: " , self.xprime_axis)
print("dx: ", dx)
#if above threshold for movement
if( dx >= 15):
#Hand Ready to be Tracked
flag = False
#If the new position is not equal to the center of the screen continue
if ((self.xprime_axis != self.x_center) and (flag == False)):
print("PALM centering Going on")
#Calculate how far away from the center
dx = self.xprime_axis - self.x_center
dx = abs(dx)
print("PALM MOVED CENTERX_axis: " , self.x_center)
print("Palm MOVED xprime_axis: " , self.xprime_axis)
#if above threshold for movement
if( dx >= 15):
print("Adjust Light Intensity")
self._changeBrightness(self.xprime_axis, dx)
else:
print("No Gesture Detected")
x = x-1
if cv2.waitKey(1) == 27 :
self._shut_down()
break
print("___________****TIME OUT*****__________")
self.system.blinkRed()
self._start_up()
return
def _powerLight(self):
if(self.system.lampLight.curpos != 0):
print(" LIGHT OFF")
self.system.lampLight.setMin()
else:
print(" LIGHT ON")
self.system.lampLight.setMax()
return
def _moveMotors(self,xpos,ypos, dx, dy):
xcounter = 0
ycounter = 0
#If the new position is to the left of the center
if(xpos < self.x_center):
#Increase the motor
#print("")
if( xcounter < dx):
#MOTOR INCREASE FUNCTION
# print("")
self.system.servoYaw.increaseRate(5)
#time.sleep(0.005)
xcounter = xcounter+1
#If the new position is to the right of the center
elif( xpos > self.x_center):
#Decrease the motor
#print("")
if( xcounter < dx):
#MOTOR DECREASE FUNCTION
#print("")
self.system.servoYaw.decreaseRate(5)
#time.sleep(0.005)
xcounter = xcounter+1
#If the new position is above the center
if((ypos < self.y_center) and (ypos != -1)):
#Increase the MOTOR
if( ycounter < dy):
#print("")
#MOTOR INCREASE FUNCTION
self.system.servoPitch.increaseRate(5)
#time.sleep(0.005)
ycounter = ycounter+1
#If the new position is below the centering
elif((ypos > self.y_center) and (ypos != -1)):
#print("")
#Decrease the Motor
if( ycounter < dy):
#MOTOR DECREASE FUNCTION
#print("")
self.system.servoPitch.decreaseRate(5)
#time.sleep(0.005)
ycounter = ycounter+1
return
def _moveSpotSize(self,xpos, dx):
xcounter = 0
#If the new position is to the left of the center
if(xpos < self.x_center):
#Increase the motor
#print("")
if( xcounter < dx):
#MOTOR INCREASE FUNCTION
# print("")
self.system.lampSpot.decreaseRate(5)
xcounter = xcounter+1
#If the new position is to the right of the center
elif( xpos > self.x_center):
#Decrease the motor
#print("")
if( xcounter < dx):
#MOTOR DECREASE FUNCTION
#print("")
self.system.lampSpot.increaseRate(5)
xcounter = xcounter+1
return
def _changeBrightness(self,xpos, dx):
xcounter = 0
#If the new position is to the left of the center
if(xpos < self.x_center):
#Increase the motor
#print("")
if( xcounter < dx):
#MOTOR INCREASE FUNCTION
print("Change Brightness: INCREASE")
self.system.lampLight.increaseRate(150)
print("CURRENT POS: " + str(self.system.lampLight.curpos))
xcounter = xcounter+1
#If the new position is to the right of the center
elif( xpos > self.x_center):
#Decrease the motor
#print("")
if( xcounter < dx):
#MOTOR DECREASE FUNCTION
print("Change Brighness: DECREASE")
self.system.lampLight.decreaseRate(150)
print("CURRENT POS DECREASE: " + str(self.system.lampLight.curpos))
xcounter = xcounter+1
return
def _delay(self, count):
while(count > 0):
count-=1
return
#stops webcam and return camera
def _shut_down (self):
self.webcam.stop()
self.webcam.stream.release()
def main(self):
# setup and run OpenGL
return
# Software
frame = np.zeros((480, 640, 3), np.uint8)
vs.frame = frame
fps = FPS().start()
print('Software execution...')
# cv function
def sw_houghlines(sw_frame):
gray = cv2.cvtColor(sw_frame, cv2.COLOR_BGR2GRAY)
blur = cv2.blur(gray, (3, 3))
edges = cv2.Canny(blur, 20, 30, apertureSize=3)
lines = cv2.HoughLines(edges, 1, np.pi / 90, 80)
segments = cv2.HoughLinesP(edges, 1, np.pi / 90, 80, 30, 5)
return [lines, segments]
# DIP loop pipeline
while fps._numFrames < iterations:
sw_frame = vs.read()
[lines, segments] = sw_houghlines(sw_frame)
fps.update()
fps.stop()
print("[INFO] elapsed time: {:.2f}".format(fps.elapsed()))
print("[INFO] approx. FPS: {:.2f}".format(fps.fps()))
#cv2.destroyAllWindows()
vs.stop()
"-l",
"--limite",
type=int,
default=50,
help="Porcentagem de branco minima encontrada para ativar a saída")
args = vars(ap.parse_args())
#Classe para ler os frames da camera/video com multithread
camera = WebcamVideoStream(src=args['source']).start()
#Classe para contar o FPS do código
fps = FPS().start()
while True:
#Le o frame disponivel da camera através da classe WebcamVideoStream
frame = camera.read()
# frame = imutils.resize(frame, width=680)
#Converte para escala de cinza, aplica um filtro gaussiano e binariza
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
blur = cv2.GaussianBlur(gray, (args['gaussian'], args['gaussian']), 0)
T, bin = cv2.threshold(blur, args['threshold'], 255, cv2.THRESH_BINARY)
#Soma a quantidade de pontos brancos na imagem
soma = bin.sum()
#Calcula a porcentagem de pontos brancos na imagem
area = bin.shape[0] * bin.shape[1] * 255
percent = soma * 100 / area
#Verifica se a porcentagem de branco esta maior que o limite
if percent > args['limite']:
cap_params['threshold'] = threshold
print(cap_params, args)
sender = ImageSender(connect_to='tcp://192.168.1.86:5555')
start_time = datetime.datetime.now()
num_frames = 0
fps = 0
index = 0
manager = CommandManager()
try:
while True:
frame = video_capture.read()
frame = cv2.flip(frame, 1)
image = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
index += 1
out = sender.send_image(image)
#elapsed_time = (datetime.datetime.now() - start_time).total_seconds()
num_frames += 1
#fps = num_frames / elapsed_time
if args.display > 0:
if (out is not None):
loc = out
for box in loc:
box = list(map(int, box))
def test_cam(args):
"""Function to predict for a camera image stream
"""
ctypes.CDLL("../TRT_object_detection/lib/libflattenconcat.so")
COCO_LABELS = coco.COCO_CLASSES_LIST
# initialize
TRT_LOGGER = trt.Logger(trt.Logger.INFO)
trt.init_libnvinfer_plugins(TRT_LOGGER, '')
runtime = trt.Runtime(TRT_LOGGER)
# compile model into TensorRT
if not os.path.isfile(model.TRTbin):
dynamic_graph = model.add_plugin(gs.DynamicGraph(model.path))
uff_model = uff.from_tensorflow(dynamic_graph.as_graph_def(),
model.output_name,
output_filename='tmp.uff')
with trt.Builder(TRT_LOGGER) as builder, builder.create_network(
) as network, trt.UffParser() as parser:
builder.max_workspace_size = 1 << 28
builder.max_batch_size = 1
builder.fp16_mode = True
parser.register_input('Input', model.dims)
parser.register_output('MarkOutput_0')
parser.parse('tmp.uff', network)
engine = builder.build_cuda_engine(network)
buf = engine.serialize()
with open(model.TRTbin, 'wb') as f:
f.write(buf)
# create engine
with open(model.TRTbin, 'rb') as f:
buf = f.read()
engine = runtime.deserialize_cuda_engine(buf)
# create buffer
host_inputs = []
cuda_inputs = []
host_outputs = []
cuda_outputs = []
bindings = []
stream = cuda.Stream()
for binding in engine:
size = trt.volume(
engine.get_binding_shape(binding)) * engine.max_batch_size
host_mem = cuda.pagelocked_empty(size, np.float32)
cuda_mem = cuda.mem_alloc(host_mem.nbytes)
bindings.append(int(cuda_mem))
if engine.binding_is_input(binding):
host_inputs.append(host_mem)
cuda_inputs.append(cuda_mem)
else:
host_outputs.append(host_mem)
cuda_outputs.append(cuda_mem)
context = engine.create_execution_context()
image_queue = LifoQueue()
depth_result_queue = LifoQueue()
#object_result_queue = LifoQueue()
cuda_lock = Lock()
# Initialize and start threads for object detection and depth inference
#object_detection_thread = ObstacleDetectionThread(image_queue, object_result_queue)
depth_inference_thread = DepthInferenceThread(image_queue,
depth_result_queue,
cuda_lock, args)
# Initialize camera to capture image stream
# Change the value to 0 when using default camera
video_stream = WebcamVideoStream(src=args.webcam).start()
if not args.no_display:
print("Trying to initinalize DisplayImage()")
# Object to display images
image_display = DisplayImage(not args.no_process)
print("Finished initializing DisplayImage()")
# Flag that records when 'q' is pressed to break out of inference loop below
quit_inference = False
def on_release(key):
if key == keyboard.KeyCode.from_char('q'):
nonlocal quit_inference
quit_inference = True
return False
keyboard.Listener(on_release=on_release).start()
print("Finished starting keyboard listener")
#object_detection_thread.start()
depth_inference_thread.start()
print("Started depth_inference_thread")
#finished = True
disp_resized = None
danger_level = None
original_width = 640
original_height = 480
# Number of frames to capture to calculate fps
num_frames = 5
curr_time = np.zeros(num_frames)
with torch.no_grad():
print("Starting inference loop")
while True:
if quit_inference:
if args.no_display:
print('-> Done')
break
# Capture and send frame to obstacle detection and depth inference thread to be process
frame = video_stream.read()
copy_frame = frame
# Capture and send frame to obstacle detection and depth inference thread to be process
#if finished:
print("Sent image to depth thread")
image_queue.put(copy_frame)
# finished = False
#else:
# print("Still doing last frame")
image = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
image = cv2.resize(image, (model.dims[2], model.dims[1]))
image = (2.0 / 255.0) * image - 1.0
image = image.transpose((2, 0, 1))
np.copyto(host_inputs[0], image.ravel())
start_time = time.time()
print("Right before copying inputs, acquiring lock")
try:
cuda_lock.acquire()
print("Object acquired lock")
cuda.memcpy_htod_async(cuda_inputs[0], host_inputs[0], stream)
print("Right before execute")
context.execute_async(bindings=bindings,
stream_handle=stream.handle)
print("Finished execute")
cuda.memcpy_dtoh_async(host_outputs[1], cuda_outputs[1],
stream)
print("Finished copying outputs")
cuda.memcpy_dtoh_async(host_outputs[0], cuda_outputs[0],
stream)
print("Finished copying outputs 2")
stream.synchronize()
print("Synchronized stream")
cuda_lock.release()
print("Object released lock")
print("execute times " + str(time.time() - start_time))
except:
print("Object couldn't acquire lock, skipping")
continue
output = host_outputs[0]
height, width, channels = frame.shape
for i in range(int(len(output) / model.layout)):
prefix = i * model.layout
index = int(output[prefix + 0])
label = int(output[prefix + 1])
conf = output[prefix + 2]
xmin = int(output[prefix + 3] * width)
ymin = int(output[prefix + 4] * height)
xmax = int(output[prefix + 5] * width)
ymax = int(output[prefix + 6] * height)
if conf > 0.7:
print("Detected {} with confidence {}".format(
COCO_LABELS[label], "{0:.0%}".format(conf)))
cv2.rectangle(frame, (xmin, ymin), (xmax, ymax),
(0, 0, 255), 3)
cv2.putText(frame, COCO_LABELS[label],
(xmin + 10, ymin + 10),
cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255),
2, cv2.LINE_AA)
# Calculate the fps
curr_time[1:] = curr_time[:-1]
curr_time[0] = time.time()
fps = num_frames / (curr_time[0] - curr_time[len(curr_time) - 1])
print("Requesting depth thread to send data back")
# Receive results from threads
#frame = None
print("Requesting obstacle thread to send data back")
#detections, frame = object_result_queue.get()
try:
disp_resized, danger_level = depth_result_queue.get()
#finished = True
except:
print("Didn't get frame from depth thread -- still working")
#print(f"Detections: {detections}")
print(danger_level)
original_width = 640
original_height = 480
if not args.no_display and disp_resized is not None:
print("About to use image_display")
# DISPLAY
# Generate color-mapped depth image
image_display.display(frame,
disp_resized,
fps,
original_width,
original_height,
blended=not args.no_blend)
#if frame is not None:
#cv2.imshow("Object detection", frame)
#else:
# continue
#cv2.waitKey(1)
else:
print(f"FPS: {fps}")
# When everything is done, stop camera stream
video_stream.stop()
depth_inference_thread.join()