def create_img(path, num):
'''
对每一张图片进行翻转，旋转，仿射变换，具体执行哪种操作是随机的，以此来增大样本数量
参数解释，
path：需要加载的图片绝对地址
num：每一张图片生成几张样本图片，如num = 5，则对每一张图片对会再随机生成五张样本图片
'''
filename = os.listdir(path)
for i, name in enumerate(filename):
i += 1
imgpath = path + '\\' + name
img = cv2.imread(imgpath)
h, w, c = img.shape
for j in range(num):
j += 1
a = random.randint(1, 4)
if a == 1:
# 将图片绕x轴翻转
img = cv2.flip(img, 0)
cv2.imwrite(path + '\\' + str(i) + '_' + str(j) + '.jpg', img)
elif a == 2:
# 将图片绕y轴翻转
img = cv2.flip(img, 1)
cv2.imwrite(path + '\\' + str(i) + '_' + str(j) + '.jpg', img)
elif a == 3:
# 将图片绕x，y轴翻转
img = cv2.flip(img, -1)
cv2.imwrite(path + '\\' + str(i) + '_' + str(j) + '.jpg', img)
def data_augment(xb, yb):
if np.random.random() < 0.25:
xb, yb = rotate(xb, yb, 90)
if np.random.random() < 0.25:
xb, yb = rotate(xb, yb, 180)
if np.random.random() < 0.25:
xb, yb = rotate(xb, yb, 270)
if np.random.random() < 0.25:
xb = cv2.flip(xb, 1) # flipcode > 0：沿y轴翻转
yb = cv2.flip(yb, 1)
if np.random.random() < 0.25:
xb = random_gamma_transform(xb, 1.0)
if np.random.random() < 0.25:
xb = blur(xb)
# 双边过滤
if np.random.random() < 0.25:
xb = cv2.bilateralFilter(xb, 9, 75, 75)
# 高斯滤波
if np.random.random() < 0.25:
xb = cv2.GaussianBlur(xb, (5, 5), 1.5)
if np.random.random() < 0.2:
xb = add_noise(xb)
return xb, yb
def take_snapshot(cam, bucket, config):
if cam is not None and cam.isOpened():
_, frame = cam.read()
for transform in config['transformations']:
if transform['action'] == 'flip':
if transform['axis'] == 'x':
frame = cv2.flip(frame, 0)
elif transform['axis'] == 'y':
frame = cv2.flip(frame, 1)
elif transform['axis'] == 'xy':
frame = cv2.flip(frame, -1)
elif transform['action'] == 'rotate':
if transform['direction'] == 'CW90':
frame = cv2.rotate(frame, cv2.ROTATE_90_CLOCKWISE)
elif transform['direction'] == 'CCW90':
frame = cv2.rotate(frame, cv2.ROTATE_90_COUNTERCLOCKWISE)
elif transform['direction'] == '180':
frame = cv2.rotate(frame, cv2.ROTATE_180)
cv2.imwrite('./{0}.png'.format(bucket), frame)
upload_coc('./{0}.png'.format(bucket), bucket)
dt = datetime.now()
if (dt.hour < dt_range[0]) | (dt.hour > dt_range[1]):
ticker = Timer(30.0 * 60.0,
take_snapshot,
args=[cam, bucket, config])
else:
ticker = Timer(config['interval'],
take_snapshot,
args=[cam, bucket, config])
tickers[config['index']] = ticker
ticker.start()
def save_webcam(outPath, fps, mirror=False):
# Capturing video from webcam:
cap = cv.VideoCapture(0)
currentFrame = 0
# Get current width of frame
width = cap.get(cv.CAP_PROP_FRAME_WIDTH) # float
# Get current height of frame
height = cap.get(cv.CAP_PROP_FRAME_HEIGHT) # float
# Define the codec and create VideoWriter object
fourcc = cv.VideoWriter_fourcc(*"XVID")
out = cv.VideoWriter(outPath, fourcc, fps, (int(width), int(height)))
prev_ret, prev_frame = cap.read()
if prev_ret:
prev_frame = cv.flip(prev_frame, 1)
while cap.isOpened():
# Capture frame-by-frame
ret, frame = cap.read()
if ret:
if mirror:
# Mirror the output video frame
frame = cv.flip(frame, 1)
# Saves for video
out.write(frame)
# Calculate the distance of the 2 pictures
diff = np.sum(
(prev_frame.astype("float") - frame.astype("float")) *
(prev_frame.astype("float") - frame.astype("float")))
diff = diff / (prev_frame.shape[0] * frame.shape[1])
if diff > threshold:
print(
"/!\ Attention /!\ Quelqu'un s'est introduit dans la matrice !"
+ diff.__str__())
prev_frame = frame
# Display the resulting frame
cv.imshow('frame', frame)
else:
break
if cv.waitKey(1) & 0xFF == ord('q'): # if 'q' is pressed then quit
break
# To stop duplicate images
currentFrame += 1
# When everything done, release the capture
cap.release()
out.release()
cv.destroyAllWindows()
def _cv2_trim(self) -> bool:
"""
Remove black borders from a cv2 image array.
This method is a f*****g waste of time as most sources are already
properly cropped. We need to use it because of a few shitty WEB sources.
F*****g unbelievable.
:param cv2_image: cv2 image array
"""
logger.info("Trying to remove black borders with cv2")
og_w, og_h = self._cv2.shape[1], self._cv2.shape[0]
logger.debug("Original dimensions: %dx%d", og_w, og_h)
og_quotient = og_w / og_h
first_img = _remove_lateral_cv2(self._cv2)
tmp_img = cv2.transpose(first_img)
tmp_img = cv2.flip(tmp_img, flipCode=1)
if tmp_img is None:
raise exceptions.InvalidRequest("Possible all-black image found")
final = _remove_lateral_cv2(tmp_img)
out = cv2.transpose(final)
final_img = cv2.flip(out, flipCode=0)
if final_img is None:
raise exceptions.InvalidRequest("Possible all-black image found")
new_w, new_h = final_img.shape[1], final_img.shape[0]
logger.debug("New dimensions: %dx%d", new_w, new_h)
new_quotient = new_w / new_h
if abs(new_quotient - og_quotient) > 0.9:
logger.info("Possible bad quotient found: %s -> %s", og_quotient,
new_quotient)
return False
width_percent = (100 / og_w) * new_w
height_percent = (100 / og_h) * new_h
if any(percent <= 65 for percent in (width_percent, height_percent)):
logger.info("Possible bad trim found: %s -> %s", width_percent,
height_percent)
return False
self._cv2 = final_img
return True
def process():
image = request.data
params = request.args
actions = [param for param in params]
actions = json.loads(actions[0])
# convert image from string to uint8
nparr = np.fromstring(image, np.uint8)
# decode image
img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
for action in actions:
if action['action'] == 'flip_H':
img = cv2.flip(img, 0)
continue
elif action['action'] == 'flip_V':
img = cv2.flip(img, 1)
continue
elif action['action'] == 'rotate_R':
img = rotateImage(img, -90)
continue
elif action['action'] == 'rotate_L':
img = rotateImage(img, 90)
continue
elif action['action'] == 'rotate':
img = rotateImage(img, int(action['param']))
continue
elif action['action'] == 'resize':
img = resizeImage(int(action['param']), img)
continue
elif action['action'] == 'thumbnail':
img = resizeImage(10, img)
continue
elif action['action'] == 'grayscale':
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
continue
_, img_encoded = cv2.imencode('.jpeg', img)
return img_encoded.tostring(), 200
def capture_frames(source: Union[str, int], frame_q: Queue, model_class):
try:
cap = cv.VideoCapture(source)
while cap.isOpened():
# Capture frame-by-frame
retval, frame = cap.read()
if retval:
frame_q.put(cv.flip(frame, 1))
yield (b'--frame\r\n' + b'Content-Type: image/jpeg\r\n\r\n' +
cv.imencode('.jpg', cv.flip(frame, 1))[1].tobytes() +
b'\r\n')
cv.waitKey(1)
finally:
cap.release()
def dataset_augmentation(images_array):
dataset = []
for image in images_array:
horizontal_flipped_img = cv2.flip(image, 1)
dataset.append(horizontal_flipped_img)
vertical_flipped_img = cv2.flip(image, 0)
dataset.append(vertical_flipped_img)
angles = [10, 15, 20]
for angle in angles:
height, width = image.shape[:2]
matrix = cv2.getRotationMatrix2D((int(width/2), int(height/2)), angle, 1)
rotated_img = cv2.warpAffine(image, matrix, (width, height))
dataset.append(rotated_img)
return np.array(dataset)
def detectAndDisplay(frame):
frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
frame_gray = cv2.equalizeHist(frame_gray)
# Detectar faces
# Divide a imagem em diversos retângulos e retorna uma lista com eles
faces = face_cascade.detectMultiScale(frame_gray)
for (x, y, w, h) in faces:
center = (x + w // 2, y + h // 2)
# Elipses rosas
frame = cv2.ellipse(frame, center, (w // 2, h // 2), 0, 0, 360,
(255, 0, 255), 4)
faceROI = frame_gray[y:y + h, x:x + w]
# Em cada face, detectar olhos
eyes = eyes_cascade.detectMultiScale(faceROI)
for (x2, y2, w2, h2) in eyes:
eye_center = (x + x2 + w2 // 2, y + y2 + h2 // 2)
radius = int(round((w2 + h2) * 0.25))
# Circulos azuis
frame = cv2.circle(frame, eye_center, radius, (255, 0, 0), 4)
# Frame invertido horizontalmente por questão de estética =)
frame = cv2.flip(frame, 1)
cv2.imshow('Deteccao de Face e Olho', frame)
return frame
def process_image(img):
img = cv2.flip(img, 1)
img = cv2.resize(img, (64, 64))
img = np.array(img, dtype=np.float32)
img = np.reshape(img, (-1, 64, 64, 3))
img = img.astype('float32') / 255.
return img
def __next__(self):
self.count += 1
if cv2.waitKey(1) == ord('q'): # q to quit
self.cap.release()
cv2.destroyAllWindows()
raise StopIteration
# Read frame
if self.pipe == 0: # local camera
ret_val, img0 = self.cap.read()
img0 = cv2.flip(img0, 1) # flip left-right
else: # IP camera
n = 0
while True:
n += 1
self.cap.grab()
if n % 30 == 0: # skip frames
ret_val, img0 = self.cap.retrieve()
if ret_val:
break
# Print
assert ret_val, f'Camera Error {self.pipe}'
img_path = 'webcam.jpg'
print(f'webcam {self.count}: ', end='')
# Padded resize
img = letterbox(img0, new_shape=self.img_size)[0]
# Convert
img = img[:, :, ::-1].transpose(2, 0, 1) # BGR to RGB, to 3x416x416
img = np.ascontiguousarray(img)
return img_path, img, img0, None
def _start_stream(self):
"""Actual stream and it's logic"""
while self._ret:
self._ret, frame = self._cap.read()
frame = cv2.flip(frame, 1)
# Convert image to HSV
frame_HSV = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
# Apply Color Filter
lr, ur = self._apply_color_filter()
# Get Mask
mask = self._create_mask(frame_HSV, lr, ur)
# Draw bounding boxes
self._draw_boxes(mask, frame)
# Draw Center of Area/Mass
com_center = self._draw_com(frame)
# Compute Distance from COM to Center of Frame
self._get_distance(com_center)
# Show Original Frame
cv2.imshow("Original Video", frame)
# Show Masked Frame
cv2.imshow("Masked Video", mask)
# Close stream if 'q' is pressed
if cv2.waitKey(1) & 0xFF == ord('q'):
return 0
def DynamicHist(src,
winname,
section=[0, 256],
removeInZero=False,
displaySize=(480, 640)):
hist = cv2.calcHist([src], [0], None, [section[1] - section[0]], section)
hist = hist[:, 0]
if removeInZero:
hist[0] = 0
binWidth = int(displaySize[1] / hist.shape[0])
k = np.max(hist) / displaySize[0]
hist = hist / k
black = np.zeros((displaySize[0], hist.shape[0] * binWidth), np.uint8)
i = 0
j = 0
for x in np.nditer(hist):
j += binWidth
black[:int(x), i:j] = 200
i += binWidth
p = int(displaySize[0] / 20)
scale = np.arange(0, black.shape[1], 50 * binWidth)
black[-p:, scale] = 0XFF
black = cv2.flip(black, 0)
cv2.imshow(winname, black)
def preprocess(self, image):
# initialise the list of crops
crops = []
height, width = image.shape[:2]
coords = [[0, 0, self.width, self.height],
[width - self.width, 0, width, self.height],
[width - self.width, height - self.height, width, height],
[0, height - self.height, self.width, height]]
# also compute the center crop
dw = int(0.5 * (width - self.width))
dh = int(0.5 * (height - self.height))
coords.append(([dw, dh, width - dw, height - dh]))
# loop over the coords and extract crops
for startx, starty, endx, endy in coords:
crop = image[starty:endy, startx:endx]
crop = cv2.resize(crop,
dsize=(self.width, self.height),
interpolation=self.inter)
crops.append(crop)
# check to see if horiz is True
if self.horiz:
mirrors = [cv2.flip(c, 1) for c in crops]
crops.extend(mirrors)
return np.array(crops)
def flip(self, image, dirction=None):
#flip the image
if dirction is None:
dirction = random.randint(-1, 1)
flipped = cv2.flip(image, dirction)
return flipped
def readFrame(frame_count_up, frame_count_down):
ret, frame = camera.read()
if ret:
frame = imutils.resize(frame, FRAME_WIDTH)
#(H, W) = frame.shape[:2]
(corners, ids,
_rejected) = cv2.aruco.detectMarkers(frame,
arucoDict,
parameters=arucoParams)
cv2.aruco.drawDetectedMarkers(frame, corners, ids, (0, 255, 0))
if ids is not None:
frame_count_up += detect(UP_MARKER, ids)
frame_count_down += detect(DOWN_MARKER, ids)
"""
for (markerCorner, _markerID) in zip(corners, ids):
corners = markerCorner.reshape((4,2))
(topLeft, _topRight, bottomRight, _bottomLeft) = corners
cX = int((topLeft[0] + bottomRight[0]) / 2.0)
cY = int((topLeft[1] + bottomRight[1]) / 2.0)
if cX < (W // 3) and cY < (H // 3):
paddle_a_up
if cX < (W // 3) and cY > (H // 3):
paddle_a_down
"""
frame = cv2.flip(frame, 1)
#cv2.rectangle(frame, (0, 0), (W // 3, H // 3), (255, 0, 0), 3)
#cv2.rectangle(frame, (0, (H // 3) * 2), (W//3, H), (0, 0, 255), 3)
cv2.imshow('Frame', frame)
return frame_count_up, frame_count_down
def faceRecAndVid():
FRAME_THICKNESS = 2
FONT_THICKNESS = 1
MODEL = 'hog' # Model for recognition cnn/hog
TOLERANCE = 0.5 # Similarity
COLOUR = [77, 77, 200] # shade of red
# Loads face encodings stored in .dat file
with open('FaceRecog/Student_Encodings.dat', 'rb') as f:
Loaded_face_encodings = pickle.load(f)
face_names = list(Loaded_face_encodings.keys())
face_encodings = np.array(list(Loaded_face_encodings.values()))
f.close()
Student_Names, Student_Face_Encodings = face_names, face_encodings
video = VideoCapture(0)
if video is None:
popUp('Error Getting Video')
else:
while True:
ret, image = video.read()
if ret == True:
image = resize(image, (0, 0), None, .75, .75, INTER_AREA)
image = cv2.flip(
image,
1) # not use cv2.rotate(img, deg) / -1 flips it over
locations = face_recognition.face_locations(image, model=MODEL)
encodings = face_recognition.face_encodings(image, locations)
for face_encoding, face_locations in zip(encodings, locations):
results = compare_faces(Student_Face_Encodings,
face_encoding, TOLERANCE)
match = None
if True in results:
# match is the name of the identity found
match = Student_Names[results.index(True)]
Attendance(match)
drawBox(image, FRAME_THICKNESS, FONT_THICKNESS, match,
face_locations)
alertSFX()
imshow('Test', image)
waitKey(500)
imshow('Test', image)
if waitKey(1) & 0xFF == ord("q"):
video.release()
cv2.destroyWindow('Test')
popUp('Done')
break
else:
video.release()
popUp('Error getting Video')
break
def bayer_aug(img, mode):
if mode == 'hor':
return cv2.flip(img, 1)[:, 1:-1]
elif mode == 'ver':
return img[::-1][1:-1, :]
elif mode == 'trans':
return img.T
return img
def preprocess_image(x, flip=False):
# x=cv2.cvtColor(x, cv2.COLOR_BGR2GRAY)
if flip:
x = cv2.flip(x, 1)
x = x[50:150, 0:320].copy()
x = cv2.resize(x, (160, 50))
x = x.reshape((50, 160, 3))
return x / 255.0
def showImage(self, img, flip=False):
img = cv2.resize(img, (240, 240))
if (flip):
img = cv2.flip(img, 1)
img = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
toshow = QImage(img.data, img.shape[1], img.shape[0],
QImage.Format_RGB888)
self.lb_video.setPixmap(QPixmap.fromImage(toshow))
def _cv2_trim(self) -> bool:
"""
Remove black borders from a cv2 image array.
:param cv2_image: cv2 image array
"""
logger.info("Trying to remove black borders with cv2")
og_w, og_h = self.cv2.shape[1], self.cv2.shape[0]
logger.debug("Original dimensions: %dx%d", og_w, og_h)
og_quotient = og_w / og_h
first_img = _remove_lateral_cv2(self.cv2)
tmp_img = cv2.transpose(first_img)
tmp_img = cv2.flip(tmp_img, flipCode=1)
if tmp_img is None:
raise exceptions.NothingFound("Possible all-black image found")
final = _remove_lateral_cv2(tmp_img)
out = cv2.transpose(final)
final_img = cv2.flip(out, flipCode=0)
new_w, new_h = final_img.shape[1], final_img.shape[0]
logger.debug("New dimensions: %dx%d", new_w, new_h)
new_quotient = new_w / new_h
if abs(new_quotient - og_quotient) > 0.9:
logger.info("Possible bad quotient found: %s -> %s", og_quotient,
new_quotient)
return False
width_percent = (100 / og_w) * new_w
height_percent = (100 / og_h) * new_h
if any(percent <= 65 for percent in (width_percent, height_percent)):
logger.info("Possible bad trim found: %s -> %s", width_percent,
height_percent)
return False
self.cv2 = final_img
return True
def _imflip_(img, direction='horizontal'):
"""Inplace flip an image horizontally or vertically.
Args:
img (ndarray): Image to be flipped.
direction (str): The flip direction, either "horizontal" or
"vertical" or "diagonal".
Returns:
ndarray: The flipped image (inplace).
"""
assert direction in ['horizontal', 'vertical', 'diagonal']
if direction == 'horizontal':
return cv2.flip(img, 1, img)
elif direction == 'vertical':
return cv2.flip(img, 0, img)
else:
return cv2.flip(img, -1, img)
def video_demo():
capture = cv.VideoCapture(0)
while True:
frame = capture.read()
frame = cv.flip(frame,1)#镜像翻转
cv.imshow("video",frame)
c = cv.waitKey(50)
if c == 27:
break
def video_demo():
capture = cv.VideoCapture(0) #括号中可以是视频路径
while (True):
ret, frame = capture.read()
frame = cv.flip(frame, 1) #摄像头镜像
cv.imshow("video", frame)
c = cv.waitKey(50)
if c == 27:
break
def flip_kernel(kernel):
"""flip a kernel (probably to prep for convolution).
:param img: the kernel
:type img: matrix
"""
k = np.copy(kernel)
return (cv2.flip(k, -1))
def flip(pic_path,save_path,min_flipValue, max_flipValue):
file=os.listdir(pic_path)
for i in file:
if (".jpg" in n)or(".png" in n)or(".JPG" in n)or(".jpeg" in n):
image = cv2.imread(pic_path+i)
xImg = cv2.flip(image,flipValue,dst=None)
cv2.imwrite(save_path+i[:-4]+"_"+str(flipValue)+"_"+str('flip')+".jpg",xImg)
else:
continue
def bayer_aug(self, img, mode):
mode_dict = {0: 'trans', 1: 'hor', 2: 'ver', 3: ''}
mode = mode_dict[mode]
if mode == 'hor':
return cv2.flip(img, 1)[:, 1:-1]
elif mode == 'ver':
return img[::-1][1:-1, :]
elif mode == 'trans':
return img.T
return img
def show_webcam(mirror=False):
cam = cv2.VideoCapture(0)
while True:
_, img = cam.read()
if mirror:
img = cv2.flip(img, 1)
cv2.imshow('my webcam', img)
if cv2.waitKey(1) == 27:
break # esc to quit
cv2.destroyAllWindows()
def Flip_img(img):
"""
翻转
flip(img,1)#1代表水平方向旋转180度
flip(img,0)#0代表垂直方向旋转180度
flip(img,-1)#-1代表垂直和水平方向同时旋转
"""
pic = cv.imread(img_path) #读入图片
h_pic = cv.flip(pic, 1) #水平翻转
cv.imshow("overturn-h", h_pic)
v_pic = cv.flip(pic, 0) #垂直翻转
cv.imshow("overturn-v", v_pic)
hv_pic = cv.flip(pic, -1) #水平垂直翻转
cv.imshow("overturn-hv", hv_pic)
cv.waitKey(0)
cv.destroyAllWindows()
def Perspective_transform(orig, screenCnt, ratio): # Funkcija za 2D transoformaciju pronađenog papira warped = four_point_transform(orig, screenCnt.reshape(4, 2) * ratio) # Pozivamo funkciju iz drugog programa za izravnjavanje slike warped = cv2.flip(warped, 1) # Zrcalimo izravnanu sliku po Y osi height, width, channels = warped.shape # Tražimo dimenzije novodobivene slike # Ako nam je širina veća od visine znači da je slika polegnuta na desnu stranu te je trebamo poravnati za 90 stupnjeva if width > height: warped = imutils.rotate_bound(warped, -90) Shadow_removal(warped) # Pozivamo funkciju za otklanjanje sjene sa slike
