def __init__(self, parent, expected_memory):
self.log = logs.logger.add_module("Threading")
QThread.__init__(self, parent)
self.cam = Camera(expected_memory)
self.cmd_fifo = queue.Queue()
self.available_space = self.cam.get_available_space()
self.filename = ""
class EtrCalibration(ConfiguredClient):
"""A simple class to convey data from multiplexer (UGM_UPDATE_MESSAGE)
to ugm_engine using udp. That level of comminication is needed, as
pyqt won`t work with multithreading..."""
@log_crash
def __init__(self, addresses):
super(EtrCalibration, self).__init__(addresses=addresses, type=peers.ETR_CALIBRATION)
HOST = self.get_param("rcv_ip")
PORT = int(self.get_param("rcv_port"))
self.socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.socket.bind((HOST, PORT))
self.socket.listen(1)
self.conn, addr = self.socket.accept()
self.ready()
self.logger.info("Start initializin etr amplifier...")
self.init()
self.initCamera()
def __del__(self):
self.socket.close()
def init():
self._treshold = 200
self.invMatrix = np.eye(3)
def initCamera(self):
self.cam = Camera()
self.cam.setTreshold(self._treshold)
def run(self):
try:
while True:
l_data = self.conn.recv(1024)
msg = variables_pb2.Variable()
msg.ParseFromString(l_data)
# ~ msg.key, msg.value # jedno z nich to timestamp i start_calibration
l_msg = None # tu bedzie parsowanie wiadomosci o starcie i koncu kalibracji
if l_msg is not None:
pass
# tu pewnie będzie czytanie obrazu z kamery i po otrzymaniu info o stopie kalibracji
# pewnie bedzie odpalenie send_resultsc
finally:
self.socket.close()
def _send_results(self):
r = variables_pb2.Sample()
r.timestamp = time.time()
for i in range(8):
r.channels.append(random.random())
self.conn.send_message(message=r.SerializeToString(), type=types.ETR_CALIBRATION_RESULTS, flush=True)
class EtrCalibration(ConfiguredClient):
"""A simple class to convey data from multiplexer (UGM_UPDATE_MESSAGE)
to ugm_engine using udp. That level of comminication is needed, as
pyqt won`t work with multithreading..."""
@log_crash
def __init__(self, addresses):
super(EtrCalibration, self).__init__(addresses=addresses, type=peers.ETR_CALIBRATION)
HOST = self.get_param('rcv_ip')
PORT = int(self.get_param('rcv_port'))
self.socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.socket.bind((HOST,PORT))
self.socket.listen(1)
self.conn, addr = self.socket.accept()
self.ready()
self.logger.info("Start initializin etr amplifier...")
self.init()
self.initCamera()
def __del__(self):
self.socket.close()
def init():
self._treshold = 200
self.invMatrix = np.eye(3)
def initCamera(self):
self.cam = Camera()
self.cam.setTreshold(self._treshold)
def run(self):
try:
while True:
l_data = self.conn.recv(1024)
msg = variables_pb2.Variable()
msg.ParseFromString(l_data)
#~ msg.key, msg.value # jedno z nich to timestamp i start_calibration
l_msg = None #tu bedzie parsowanie wiadomosci o starcie i koncu kalibracji
if l_msg is not None:
pass
#tu pewnie będzie czytanie obrazu z kamery i po otrzymaniu info o stopie kalibracji
#pewnie bedzie odpalenie send_resultsc
finally:
self.socket.close()
def _send_results(self):
r = variables_pb2.Sample()
r.timestamp = time.time()
for i in range(8):
r.channels.append(random.random())
self.conn.send_message(message = r.SerializeToString(),
type = types.ETR_CALIBRATION_RESULTS, flush=True)
def cam_update_img():
global cam
if not cam:
cam = Camera(800, 600)
try:
cam.start_once(get_yolo_result)
# _thread.start_new_thread(hello, (1234, ))
return jsonify({
'success': True,
})
except:
print("Error: start thread error ")
return jsonify({
'success': False,
})
def __init__(self, mrz_port="/dev/ttyACM0"):
with timer.time(f'load-total'):
self.card_reader = CardReader()
self.card_reader.set_mrz_port(mrz_port)
with timer.time(f'load-mtcnn'):
self.mtcnn = MTCNN()
self.mtcnn.init(480, 640)
with timer.time(f'load-facenet'):
self.face_embedder = FaceEmbedder()
self.camera = Camera()
#
self.camera_face = None
self.camera_emb = None
self.reader_face = None
self.reader_emb = None
class Main:
def __init__(self):
self.camera = Camera()
with timer.time(f'load'):
self.mtcnn = MTCNN()
self.mtcnn.init(480,640)
def main(self, manual_brake=None):
while not manual_brake:
frame = self.camera.read()
try:
with timer.time(f'inference'):
bboxes = self.mtcnn.findFace(frame) # [3,h,w]
for x1,y1,x2,y2 in bboxes:
cv2.rectangle(frame, (x1,y1), (x2,y2), (255,0,0), 2)
except Exception as e: # face not found
print(e)
continue
manual_brake = self.camera.show_frame(frame)
self.camera.clean()
timer.print_stats()
def main():
'''
Main loop that initializes camera and other
'''
args = parse_args()
config = Config()
#Start Camera async class
cam = Camera(args)
cam.start()
#Start io writing queue
write_queue = WriteQueue()
write_queue.start()
if config.Inference.mode == 'detect':
print('Running detection')
detector = OpenVinoDetectorAsync(config.Inference)
else:
detector = OpenVinoClassifierAsync(config.Inference)
while True:
_,frame = cam.read()
start_time = time.time()
infer_frame = deepcopy(frame)
detections = detector.run(infer_frame)
timestamp = datetime.now(tz=timezone.utc).strftime('%Y-%m-%d-%H-%M-%S-%f')
path = 'tmp/' + timestamp + '.jpg'
if detections:
'''
Need to fix the bounding box locations
for detection in detections:
xmin = detection.position.xmin
ymin = detection.position.ymin
xmax = detection.position.xmax
ymax = detection.position.ymax
cv2.rectangle(frame, (int(xmin), int(ymin)), (int(xmax), int(ymax)), detection.color, 2)
'''
cv2.putText(frame,'HUMAN',(10,400), cv2.FONT_HERSHEY_SIMPLEX, 4,(25,25,255),2,cv2.LINE_AA)
cv2.imshow('frame', frame)
cv2.imwrite(path, frame)
#This has to RTT
upload_frame(path, config)
else:
#add to upload queue
write_queue.enqueue(path, frame)
end_time = time.time()
print("[LOGS] ---PIPELINE TIME--- **{}**".format(end_time-start_time))
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cam.stop()
write_queue.stop()
cv2.destroyAllWindows()
def start_cam_service():
global cam
if not cam:
cam = Camera(800, 600)
if cam and (not cam.operating):
try:
_thread.start_new_thread(cam.start, (get_yolo_result, ))
# _thread.start_new_thread(hello, (1234, ))
return jsonify({'success': True, 'operating': True})
except:
print("Error: start thread error ")
return jsonify({'success': False, 'error': 'START_THREAD_ERROR'})
else:
return jsonify({'success': True, 'operating': True})
from temperature import Temperature
from light import Light
from cam import Camera
import time
import RPi.GPIO as GPIO
if __name__ == "__main__":
GPIO.setmode(GPIO.BCM)
camera = Camera("pub_monitor_thread")
camera.start()
while True:
# get and pub temperature
temperature = Temperature()
temperature.pub_temperature()
# get and pub light
light = Light(17, 0)
light.pub_light()
time.sleep(1)
from m2 import M2
from wow import WoW
from cam import Camera
from objects import GameObject
# TODO: RENDER in argv
# RENDER = False
RENDER = True
GODI_PATH = 'Y:\\dbc\\GameObjectDisplayInfo.dbc'
CMD_PATH = 'Y:\\dbc\\CreatureModelData.dbc'
MODELS_PREFIX = 'Y:\\model.mpq'
w = WoW(godi_path=GODI_PATH, cmd_path=CMD_PATH)
cam = Camera(w)
it = []
it += list(w.gen_game_objects())
it += list(w.gen_players())
print(' Snapshoting screen...')
with mss.mss() as sct:
monitor = sct.monitors[1]
img = sct.grab(monitor)
img = np.asarray(img)[:, :, [2, 1, 0]]
print(' Snapshoted!', img.shape)
plt.close('all')
fig = plt.figure()
ax = fig.add_subplot(111)
class Main:
def __init__(self, mrz_port="/dev/ttyACM0"):
with timer.time(f'load-total'):
self.card_reader = CardReader()
self.card_reader.set_mrz_port(mrz_port)
with timer.time(f'load-mtcnn'):
self.mtcnn = MTCNN()
self.mtcnn.init(480, 640)
with timer.time(f'load-facenet'):
self.face_embedder = FaceEmbedder()
self.camera = Camera()
#
self.camera_face = None
self.camera_emb = None
self.reader_face = None
self.reader_emb = None
def prep_im(self, im):
im = cv2.resize(im, (160, 160))
return torch.tensor(im).permute(2, 1, 0).to('cuda')
async def camera_loop(self):
while True:
with timer.time(f'camera-total'):
frame = self.camera.read()
print('[camera] frame:', frame.shape)
with timer.time(f'camera-mtcnn'):
bboxes = self.mtcnn.findFace(frame)
print('[camera] bboxes:', bboxes)
for x1, y1, x2, y2 in bboxes:
self.camera_face = frame[y1:y2, x1:x2]
print('[camera] face:', self.camera_face.shape)
with timer.time(f'camera-facenet'):
im = self.prep_im(self.camera_face)
self.camera_emb = self.face_embedder.face2embedding(im)
await asyncio.sleep(0)
async def reader_loop(self):
while True:
print('[reader] starting read...')
chip_img_future = self.card_reader.read_image_bytes_async()
t0 = time()
while not chip_img_future.isDone():
await asyncio.sleep(0)
byte_array = chip_img_future.get()
if len(byte_array) > 0:
print('[reader] byte arr len:', len(byte_array))
chip_img = self.card_reader.bytes2np(byte_array) # [h,w]
timer.D['reader-read'].append(time() - t0)
print('[reader] chip img:', chip_img.shape)
chip_img = np.concatenate(
[chip_img[:, :, None] for _ in range(3)],
axis=2) # [h,w,3]
with timer.time(f'reader-mtcnn'):
bboxes = self.mtcnn.findFace(chip_img)
print('[reader] bboxes:', bboxes)
for y1, x1, y2, x2 in bboxes:
self.reader_face = chip_img[y1:y2, x1:x2]
print('[reader] chip face:', self.reader_face.shape)
with timer.time(f'reader-facenet'):
im = self.prep_im(self.reader_face)
self.reader_emb = self.face_embedder.face2embedding(im)
timer.D['reader-total'].append(time() - t0)
self.compare_embeddings_and_exit()
def compare_embeddings_and_exit(self):
if self.reader_face is None:
print("[ERROR] Couldn't read face from reader...")
if self.camera_face is None:
print("[ERROR] Couldn't read face from camera...")
cv2.imwrite('reader_face.png', self.reader_face)
cv2.imwrite('camera_face.png', self.camera_face)
score = float((self.reader_emb - self.camera_emb).norm())
print('score:', score)
timer.print_stats()
sys.exit()
def main(self):
asyncio.gather(self.camera_loop(), self.reader_loop())
asyncio.get_event_loop().run_forever()
def initCamera(self):
self.cam = Camera()
self.cam.setTreshold(self._treshold)
cam.rot +=1
mv = Vector()
if kbd.a:
mv.y -= 1
if kbd.d:
mv.y += 1
if kbd.w:
mv.x += 1
if kbd.s:
mv.x -= 1
if kbd.up:
cam.rays +=1
if kbd.down:
cam.rays -=1
cam.move(mv)
cam.colides(lines)
lines.append(Line(Vector(300,100),Vector(300,200),True,wallImg))
cam = Camera()
kbd = Keyboard()
physloop = simplegui.create_timer(1000/60,pysloop)
frame = simplegui.create_frame("Points", Camera.WIDTH , Camera.HEIGHT,0)
frame.set_draw_handler(draw)
frame.set_canvas_background("White")
frame.set_keydown_handler(kbd.keyDown)
frame.set_keyup_handler(kbd.keyUp)
# pos the frame animation
frame.set_mouseclick_handler(mouse)
physloop.start()
frame.start()
import pandas as pd
import shapely.geometry as sg
import pymem
from show import patchify_points, patchify_polys
from m2 import M2
from wow import WoW
from cam import Camera
from objects import GameObject
import constants
GODI_PATH = 'Y:\\dbc\\GameObjectDisplayInfo.dbc'
CMD_PATH = 'Y:\\dbc\\CreatureModelData.dbc'
MODELS_PREFIX = 'Y:\\model.mpq'
w = WoW(godi_path=GODI_PATH, cmd_path=CMD_PATH)
cam = Camera(w)
rows = []
df = globals().get('df', pd.DataFrame())
# Mem Explorer
DEPTH_MAX = 5
BCOUNT_START = 2500
BCOUNT_IN_DEPTH = 2000
seen = set()
def find_at(addr, bcount, depth=()):
if addr in seen:
return
seen.add(addr)
class imageGradientDescent():
# Constants that will be necessary for evaluation of cost function
X_PIXELS = 2160 // 10
Y_PIXELS = 2160 // 10
X_MIDPOINT = X_PIXELS / 2
Y_MIDPOINT = Y_PIXELS / 2
IDEAL_Z_DISP = 0
EPSILON = 95
MAX_STEPS = 100000
def __init__(self):
# Initalize the camera and motorController for use in the gradient descent
self.cam = Camera()
self.controller = MotorController()
# Plot radon transform of Intensity vs projection angle and return sinogram
def plot_radon_angle_x(self, image):
# Open image
img = image
# resize image
scale_percent = 10 # percent of original size
width = int(img.shape[1] * scale_percent / 100)
height = int(img.shape[0] * scale_percent / 100)
dim = (width, height)
resized = cv2.resize(img, dim, interpolation=cv2.INTER_AREA)
# Radon transform resized image
nSteps = width
thetaStart = -10
thetaEnd = 10
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(resized, theta=thetas, circle=True)
# Take absolute max of sinogram across axis and find location where it occurs
absmax_sinogram = np.amax(sinogram, axis=0)
absmax_loc = thetas[np.where(absmax_sinogram == absmax_sinogram.max())]
# Plot original image and absMax of Intensity of radon transform vs projection angle
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(8, 4.5))
ax1.set_title("Original")
ax1.imshow(img, cmap=plt.cm.Greys_r)
ax2.set_title("Radon transform\n(Sinogram)")
ax2.set_xlabel("Projection angle (deg)")
ax2.set_ylabel("Maximum Intensity")
ax2.plot(thetas, absmax_sinogram, '-')
fig.tight_layout()
plt.show()
# Plot radon transform of Intensity vs x-pixels and return sinogram
def plot_radon_pixels_x(self, image):
# Open image
img = image
# resize image
scale_percent = 10 # percent of original size
width = int(img.shape[1] * scale_percent / 100)
height = int(img.shape[0] * scale_percent / 100)
dim = (width, height)
resized = cv2.resize(img, dim, interpolation=cv2.INTER_AREA)
# Radon transform resized image
nSteps = height
thetaStart = 80
thetaEnd = 100
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(resized, theta=thetas, circle=True)
# Take absmax of intensity in dimensions of pixels
absmax_sinogram = np.amax(sinogram, axis=1)
peaks, properties = sp.signal.find_peaks(absmax_sinogram,
prominence=100)
# Plot original image and absMax of Intensity of radon transform vs projection angle
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(8, 4.5))
ax1.set_title("Original")
ax1.imshow(img, cmap=plt.cm.Greys_r)
ax2.set_title("Radon transform\n(Sinogram)")
ax2.set_xlabel("Projection angle (deg)")
ax2.set_ylabel("Maximum Intensity")
ax2.plot(peaks, absmax_sinogram[peaks], 'x')
fig.tight_layout()
plt.show()
# Plot radon transform of Intensity vs projection angle and return sinogram
def plot_radon_angle_y(self, image):
# Open image
img = image
# resize image
scale_percent = 10 # percent of original size
width = int(img.shape[1] * scale_percent / 100)
height = int(img.shape[0] * scale_percent / 100)
dim = (width, height)
resized = cv2.resize(img, dim, interpolation=cv2.INTER_AREA)
# Radon transform resized image
nSteps = height
thetaStart = 80
thetaEnd = 100
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(resized, theta=thetas, circle=True)
# Take absolute max of sinogram across axis and find location where it occurs
absmax_sinogram = np.amax(sinogram, axis=0)
absmax_loc = thetas[np.where(absmax_sinogram == absmax_sinogram.max())]
# Plot original image and absMax of Intensity of radon transform vs projection angle
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(8, 4.5))
ax1.set_title("Original")
ax1.imshow(img, cmap=plt.cm.Greys_r)
ax2.set_title("Radon transform\n(Sinogram)")
ax2.set_xlabel("Projection angle (deg)")
ax2.set_ylabel("Maximum Intensity")
ax2.plot(thetas, absmax_sinogram, '-')
fig.tight_layout()
plt.show()
# Plot radon transform of Intensity vs y-pixels and return sinogram
def plot_radon_y_pixels(self, image):
# Open image
img = image
# resize image
scale_percent = 10 # percent of original size
width = int(img.shape[1] * scale_percent / 100)
height = int(img.shape[0] * scale_percent / 100)
dim = (width, height)
resized = cv2.resize(img, dim, interpolation=cv2.INTER_AREA)
# Radon transform resized image
nSteps = height
thetaStart = 80
thetaEnd = 100
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(resized, theta=thetas, circle=True)
absmax_sinogram = np.amax(sinogram, axis=1)
peaks, properties = sp.signal.find_peaks(absmax_sinogram,
prominence=100)
# Plot original image and absMax of Intensity of radon transform vs projection angle
fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(8, 4.5))
ax1.set_title("Original")
ax1.imshow(img, cmap=plt.cm.Greys_r)
ax2.set_title("Radon transform\n(Sinogram)")
ax2.set_xlabel("Radii (pixels)")
ax2.set_ylabel("Maximum Intensity")
ax2.plot(peaks, absmax_sinogram[peaks], 'x')
fig.tight_layout()
plt.show()
# Returns location of absolute max of Intensity vs Projection angle radon transform
def getRT_angle_x(self, image, scale):
# Create image object
img = image
# resize image
scale_percent = scale # percent of original size
width, height = img.shape
width = (width * scale_percent // 100)
height = (height * scale_percent // 100)
dim = [width, height]
img = np.resize(img, dim)
# Radon transform resized image
nSteps = width
thetaStart = -10
thetaEnd = 10
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(img, theta=thetas, circle=True)
# Take absolute max of sinogram across axis and find location where it occurs
absmax_sinogram = np.amax(sinogram, axis=0)
absmax_loc = thetas[np.where(absmax_sinogram == absmax_sinogram.max())]
return absmax_loc
# Returns tuple of abs max and rel max locations and magnitudes
# Returns [rmax1_mag_x, rmax1_loc_x, amax_mag_x, amax_loc_x, rmax2_mag_x, rmax2_loc_x]
def getRT_pixel_x(self, image, scale):
# Open image
img = image
# resize image
scale_percent = scale # percent of original size
width, height = img.shape
width = (width * scale_percent // 100)
height = (height * scale_percent // 100)
dim = [width, height]
img = np.resize(img, dim)
# Radon transform resized image
nSteps = width
thetaStart = -10
thetaEnd = 10
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(img, theta=thetas, circle=True)
# Take absmax of intensity in dimensions of pixels
absmax_sinogram = np.amax(sinogram, axis=1)
peaks, properties = sp.signal.find_peaks(absmax_sinogram,
height=180,
distance=width / 4)
vals = [peaks, absmax_sinogram[peaks]]
return vals
# Returns tuple of abs max and rel max locations and magnitudes
# Returns [rmax1_mag_y, rmax1_loc_y, amax_mag_y, amax_loc_y, rmax2_mag_y, rmax2_loc_y]
def getRT_angle_y(self, image, scale):
# Open image
img = image
# resize image
scale_percent = scale # percent of original size
width, height = img.shape
width = (width * scale_percent // 100)
height = (height * scale_percent // 100)
dim = [width, height]
img = np.resize(img, dim)
# Radon transform resized image
nSteps = height
thetaStart = 80
thetaEnd = 100
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(img, theta=thetas, circle=True)
# Take absolute max of sinogram across axis and find location where it occurs
absmax_sinogram = np.amax(sinogram, axis=0)
absmax_loc = thetas[np.where(absmax_sinogram == absmax_sinogram.max())]
return absmax_loc
# Ye boi
def getRT_pixel_y(self, image, scale):
# Open image
img = image
# resize image
scale_percent = scale # percent of original size
width, height = img.shape
width = (width * scale_percent // 100)
height = (height * scale_percent // 100)
dim = [width, height]
img = np.resize(img, dim)
# Radon transform resized image
nSteps = height
thetaStart = 80
thetaEnd = 100
thetas = np.linspace(thetaStart, thetaEnd, nSteps)
sinogram = radon(img, theta=thetas, circle=True)
# Take absmax of intensity in dimensions of pixels
absmax_sinogram = np.amax(sinogram, axis=1)
peaks, properties = sp.signal.find_peaks(absmax_sinogram,
height=180,
distance=width / 4)
vals = [peaks, absmax_sinogram[peaks]]
return vals
# Evaluate cost function based on the three sinograms to be generated from image
def calc_cf(self, image, scale):
# Values needed for evaluation of cost function
xparams = self.getRT_pixel_x(image, scale)
yparams = self.getRT_pixel_y(image, scale)
print('x params ' + str(xparams) + ' y params ' + str(yparams))
zdisp_x = 0 - self.getRT_angle_x(image, scale)
zdisp_y = 0 - self.getRT_angle_y(image, scale)
rmax1_loc_x = xparams[0][0]
rmax1_mag_x = xparams[1][0]
rmax2_loc_x = xparams[0][2]
rmax2_mag_x = xparams[1][2]
rmax1_loc_y = yparams[0][0]
rmax1_mag_y = yparams[1][0]
rmax2_loc_y = yparams[0][2]
rmax2_mag_y = yparams[1][2]
amax_loc_x = xparams[0][1]
amax_mag_x = xparams[1][1]
amax_loc_y = yparams[0][1]
amax_mag_y = yparams[1][1]
# Cost function components
s1 = np.square(self.IDEAL_Z_DISP - zdisp_x)
s2 = np.square(self.X_MIDPOINT - amax_loc_x)
s3 = np.square(rmax1_mag_x - rmax2_mag_x)
s4 = np.square((amax_loc_x - rmax1_loc_x) - (rmax2_loc_x - amax_loc_x))
s5 = np.square(self.IDEAL_Z_DISP - zdisp_y)
s6 = np.square(self.Y_MIDPOINT - amax_loc_y)
s7 = np.square(rmax1_mag_y - rmax2_mag_y)
s8 = np.square((amax_loc_y - rmax1_loc_y) - (rmax2_loc_y - amax_loc_y))
#print(str(s1) + ',' + str(s2) + ',' + str(s3) + ',' + str(s4) + ',' + str(s5) + ',' + str(s6) + ',' + str(s7) + ',' + str(s8))
costFunction = np.sqrt(s1 + s2 + s3 + s4 + s5 + s6 + s7 + s8)
print(str(costFunction))
# If the number of peaks is greater than 3, return cost function times 1.5
if len(xparams) > 3:
return int(costFunction) * 2
return int(costFunction)
# Determines the stepSize based on value of cost function
def calc_stepSize(self, cf):
if cf > 0 and cf < 100:
stepSize = 1
return stepSize
elif cf >= 100 and cf <= 200:
stepSize = 5
return stepSize
elif cf > 200 and cf < 300:
stepSize = 10
return stepSize
else:
stepSize = 25
return stepSize
# Determines image resolution based on value of cf
def calc_scaleFactor(self, cf):
if cf > 0 and cf < 100:
scaleFactor = 10
return scaleFactor
elif cf >= 100 and cf <= 200:
scaleFactor = 10
return scaleFactor
else:
scaleFactor = 10
return scaleFactor
#
def gradientDescent(self):
# Create dequeue of motors for descent
motorDeque = collections.deque([1, 2, 3, 4, 5])
# Initiliaze variables to keep track of number of steps and the step size
numSteps = 0
stepSize = 10
scale = 10
# Take a picture
image = self.cam.takePicture().astype('uint8')
# evaluate cost function based on the picture
cf = self.calc_cf(image, scale)
scale = self.calc_scaleFactor(cf)
stepSize = self.calc_stepSize(cf)
if (cf < self.EPSILON):
return
while (cf > self.EPSILON and numSteps < self.MAX_STEPS):
# Start with first motor in dequeue
curMotor = motorDeque.popleft()
# Make sure to place element at end of deque
motorDeque.append(curMotor)
# numSteps++
numSteps += 1
# Choose a direction to try adjusting the motor
direction = 0
# Move appropriate motor
# Rotational Z motor
if curMotor == 1:
self.controller.turnZ(direction, stepSize)
# Rotational X motor
elif curMotor == 2:
self.controller.turnX(direction, stepSize)
# Rotational Y motor
elif curMotor == 3:
self.controller.turnY(direction, stepSize)
# Translational X motor
elif curMotor == 4:
self.controller.moveX(direction, stepSize)
# Translational Z motor
else:
self.controller.moveZ(direction, stepSize)
# Take a picture & evaluate new_cf
image = self.cam.takePicture().astype('uint8')
new_cf = self.calc_cf(image, scale)
scale = self.calc_scaleFactor(new_cf)
stepSize = self.calc_stepSize(new_cf)
# From cf determine proper step size and image scale
# If new_cf is higher, we must go in the other direction
if new_cf > cf:
direction = 1
cf = new_cf
# If cf is lower than epsilon, break the loop
if cf < self.EPSILON:
break
while new_cf <= cf:
# Tweak same motor in correct direction
# Move appropriate motor
# Rotational Z motor
if curMotor == 1:
self.controller.turnZ(direction, stepSize)
# Rotational X motor
elif curMotor == 2:
self.controller.turnX(direction, stepSize)
# Rotational Y motor
elif curMotor == 3:
self.controller.turnY(direction, stepSize)
# Translational X motor
elif curMotor == 4:
self.controller.moveX(direction, stepSize)
# Translational Z motor
else:
self.controller.moveZ(direction, stepSize)
numSteps += 1
cf = new_cf
image = self.cam.takePicture().astype('uint8')
new_cf = self.calc_cf(image, scale)
scale = self.calc_scaleFactor(new_cf)
stepSize = self.calc_stepSize(new_cf)
# If cf lower than epsilon, break the loop
if new_cf < self.EPSILON:
break
def initCamera(self):
self.cam = Camera()
self.cam.setTreshold(self._treshold)
def setupApp(self):
self.current_user['username'] = self.login.username_lineedit.text()
print(self.current_user['uid'])
self.startNetFinder()
self.show()
def startNetFinder(self):
self.thread_finder = QThread()
self.online_finder = NetFinder(self.current_user)
self.online_finder.moveToThread(self.thread_finder)
self.thread_finder.started.connect(self.online_finder.run)
self.online_finder.new_client.connect(self.online_dialog.addOnlineUser)
self.thread_finder.start()
if __name__ == '__main__':
camera = Camera(0)
app = QApplication([])
# start_window = UI_Window(camera)
start_window = NetCrawler()
# apply_stylesheet(app, theme='dark_teal.xml')
# start_window.show()
app.exit(app.exec_())
def __init__(self):
# Initalize the camera and motorController for use in the gradient descent
self.cam = Camera()
self.controller = MotorController()
class Threading(QThread):
sig_live_view = pyqtSignal(object)
sig_photo = pyqtSignal(object)
sig_error = pyqtSignal(object)
sig_critical = pyqtSignal(int)
def __init__(self, parent, expected_memory):
self.log = logs.logger.add_module("Threading")
QThread.__init__(self, parent)
self.cam = Camera(expected_memory)
self.cmd_fifo = queue.Queue()
self.available_space = self.cam.get_available_space()
self.filename = ""
def sendThreadCommand(self, cmd):
if not self.isRunning():
self.start()
self.cmd_fifo.put(cmd)
def start_live(self):
'''
Startet die Vorschau. Daten werden via sig_live_view gesendet
'''
self.log.debug("Start liveview")
self.sendThreadCommand(Action.PREVIEW)
def capture_image(self):
'''
Erstellt Foto. Daten werden via sig_foto gesendet
'''
self.log.debug("Capture image")
self.sendThreadCommand(Action.CAPTURE)
def get_available_space(self):
return self.available_space
def store_last(self, filename):
self.log.debug("Store image")
self.filename = filename
self.sendThreadCommand(Action.STORE)
def dismiss_last(self):
self.log.debug("Delete image")
self.sendThreadCommand(Action.DISMISS)
def stop_thread(self):
self.sendThreadCommand(Action.TERMINATE)
while self.isRunning():
self.yieldCurrentThread()
self.log.info("Thread terminated")
def run(self):
self.log.info("Thread started")
state = const.STATE_LIVE
ts = 0
while True:
old_state = state
if not self.cmd_fifo.empty():
state = self.cmd_fifo.get()
ts = time.time()
if state == Action.PREVIEW:
# kontinuierlich livebild abholen und an GUI senden
self.sig_live_view.emit(self.cam.fetch_preview())
elif state == Action.CAPTURE:
# einmal Foto machen und an GUI senden
# danach warten bis weiter
photo = self.cam.capture_image()
if self.cam.last_image:
self.sig_photo.emit(photo)
else:
self.sig_error.emit(photo)
self.available_space = self.cam.get_available_space()
state = Action.NONE
elif state == Action.STORE:
self.cam.store_last(self.filename)
state = old_state
elif state == Action.DISMISS:
self.cam.dismiss_last()
state = old_state
elif state == Action.TERMINATE:
return
else:
# nichts machen, um CPU zu schonen, Pause
# BUGFIX: wenn die Kamera 30min nicht verwendet wurde, schaltet sie sich automatisch aus.
# wenn man mind. diese Zeit auf dem Auswahlbildschirm steht, stuerzt beim naechsten Kamerazugriff
# die Applikation ab. Loesung: kontinuierlich alle x Sekunden preview von Kamera abholen,
# Daten aber verwerfen
if ((ts + 10) < time.time()):
ts = time.time()
self.cam.fetch_preview()
else:
self.msleep(100)
if self.cam.error_count > const.MAX_ERROR_COUNT:
self.sig_critical.emit(self.cam.error_count)
from evaluate_gpu import get_nearest_neighbors
from evaluate_rerank import get_nearest_neighbors_rerank
import scipy.io
import torch
from flask import Flask, jsonify, request
from flask_cors import CORS
from detect import get_model, get_bbox_of_image, arg_parse
from cam import Camera
import _thread
import os
import numpy as np
app = Flask(__name__)
CORS(app)
cam = Camera(800, 600)
import os
model_pcb = None
model_dense = None
model_yolo = None
result_pcb = scipy.io.loadmat('reid/pytorch_result_PCB.mat')
gf_pcb = torch.FloatTensor(result_pcb['gallery_f'])
gf_pcb = gf_pcb.cuda()
gc_pcb = result_pcb['gallery_cam'][0]
gl_pcb = result_pcb['gallery_label'][0]
gp_pcb = result_pcb['gallery_path'][..., 0]
# g_g_dist = np.dot(gf_pcb, np.transpose(gf_pcb))
def __init__(self):
self.camera = Camera()
with timer.time(f'load'):
self.mtcnn = MTCNN()
self.mtcnn.init(480,640)
from base64 import b64decode
from fastapi import FastAPI
from fastapi.responses import StreamingResponse
from tortoise.contrib.fastapi import register_tortoise
from authentication import router, check_token, credentials_exception
from cam import Camera
camera = Camera()
app = FastAPI()
app.include_router(router)
def stream_video():
while True:
frame = camera.get_frame()
yield (b'--frame\r\n'
b'Content-Type: image/jpeg\r\n\r\n' + frame + b'\r\n')
@app.get('/mstreamj.mjpg')
async def main(t: bytes):
token = b64decode(t).decode('utf-8')
print(token)
if await check_token(token):
return StreamingResponse(
stream_video(),
media_type='multipart/x-mixed-replace; boundary=frame')
raise credentials_exception
