def train(self, training_data_path):
faces = []
labels = []
#cv2.namedWindow('Training...', cv2.WND_PROP_FULLSCREEN)
#cv2.setWindowProperty('Training...', cv2.WND_PROP_FULLSCREEN, cv2.WINDOW_FULLSCREEN)
#screen_width, screen_height = utils.get_screen_resolution()
#window_width = int(screen_width * 0.8)
#window_height = int(screen_height * 0.8)
#cv2.namedWindow('Training...', cv2.WINDOW_NORMAL)
#cv2.resizeWindow('Training...', window_width, window_height)
cv2.namedWindow('Training...')
cv2.moveWindow('Training...', 0, 0)
dirs = os.listdir(training_data_path)
print dirs
label = 0
for dir_name in dirs:
if dir_name.startswith('.'):
continue
subject = dir_name
print subject, label
subject_path = os.path.join(training_data_path, dir_name)
subject_image_files = os.listdir(subject_path)
# Read each image, detect the face, add the detected face to the
# subject's list of faces
for image_name in subject_image_files:
# Ignore system files like .DS_Store
if image_name.startswith('.') or image_name == 'name.txt':
continue
print image_name
image_path = os.path.join(subject_path, image_name)
image = cv2.imread(image_path)
# Display an image window to show the image
#cv2.imshow('Training...', image)
small_image = cv2.resize(image, None, fx=0.1, fy=0.1)
cv2.imshow('Training...', small_image)
cv2.waitKey(100)
# Detect and record face
face, rect = self.face_detector.detect(image)
if face is not None:
small_rectangle_coordinates = utils.scale_coordinates(
rect, 0.1)
utils.draw_rectangle(small_image,
small_rectangle_coordinates)
cv2.imshow('Training...', small_image)
cv2.waitKey(100)
faces.append(face)
labels.append(label)
label += 1
# Clean up after ourselves
cv2.destroyWindow('Training...')
cv2.waitKey(1)
cv2.destroyAllWindows()
self.face_recogniser.train(faces, numpy.array(labels))
def draw(self, app, cr):
utils.draw_rectangle(
cr, self._rect, self._fill_color, self._outline_color)
node_data = [
# Nodes, input, offset, text align
(self._component.inputs, True,
(NODE_RADIUS+2, 0), utils.TextHAlign.LEFT),
(self._component.outputs, False,
(-NODE_RADIUS-2, 0), utils.TextHAlign.RIGHT),
]
for nodes, is_input, offset, text_align in node_data:
for i, node in enumerate(nodes):
position = self.node_pos(is_input, i)
connected = node.is_connected()
fill_color = _node_color(node.value, connected)
utils.draw_circle(cr, position, NODE_RADIUS, fill_color, BLACK)
utils.draw_text(
cr, node.label, position + offset, size=10,
bold=True, h_align=text_align)
name = self._component.name
if name:
position = self.center - (0, self._rect.height/2+8)
utils.draw_text(cr, name, position, bold=True)
self._component.on_draw(app, cr)
def process(img_folder, label_boxes, hog, clf):
for f in file_iterator(img_folder, 'jpg'):
img = cv2.cvtColor(cv2.imread(f), cv2.COLOR_BGR2GRAY)
f = os.path.basename(f)
print ('Processing file %s'.format(f))
windows = detect(img, hog, clf)
label_box = label_boxes[f]
img_with_box = img.copy()
draw_rectangle(img_with_box, label_box, color=(255,55,0))
if len(windows) == 0:
cv2.imwrite(LOW_ACCURACY_IMG_PATH + f, img_with_box)
with open(ACCURACY_FILE, 'a') as af:
af.write(f + ',0\n')
continue
pred_box = get_corner(flattern(windows))
accuracy = cpt_accuracy(label_box, pred_box)
with open(ACCURACY_FILE, 'a') as af:
af.write(f + ',' + str(accuracy) + '\n')
draw_rectangle(img_with_box, pred_box, color=(0,55,255))
cv2.imwrite(PREDICT_PATH + f, img_with_box)
if accuracy < 0.5:
cv2.imwrite(LOW_ACCURACY_IMG_PATH + f, img_with_box)
apply_hard_negative(img, flattern(windows), label_box)
def web_identify():
form = IdentifyForm(request.form)
if request.method == 'POST':
if len(request.files) > 0:
file = request.files['file']
if utils.is_picture(file.filename):
try:
pilimage = Image.open(file.stream, 'r').convert('RGB')
except Exception as e:
flash("Error: {}".format(e))
return render_template('identify.html', form=form)
try:
faces = dlib_api.detect_and_identify_faces(
np.array(pilimage))
except Exception as e:
flash("Error: {}".format(e))
return render_template('identify.html', form=form)
font = utils.get_font(1024)
draw = ImageDraw.Draw(pilimage)
for p, rect, shape in faces:
s = font.getsize(p.name)
utils.draw_rectangle(draw, rect, 5, color=128)
draw.text(
((rect.right() - rect.left()) / 2 + rect.left() -
int(s[0] / 2), rect.top() - s[1] - 20),
p.name,
fill=128,
font=font)
del draw
tmpfile = os.path.join('static/tmp',
'{}.jpg'.format(uuid.uuid4()))
# resize
pilimage = utils.resize_image(pilimage, 1024)
with open(tmpfile, 'wb+') as fp:
pilimage.save(fp, 'JPEG')
flash('found {} faces... '.format(len(faces)) +
'\n'.join(("id: {}, name: {}".format(p.id, p.name)
for p, _, _ in faces)))
return render_template('identify.html',
form=form,
proc_image=tmpfile)
else:
flash('Error: Only images allowed!')
else:
flash('Error: All the form fields are required. ')
return render_template('identify.html', form=form)
def draw(self, app, cr, mouse_pos):
color = (0, 0, 1)
for component in self._components:
rect = component.display.rect
rect = shapes.Rectangle(
size=rect.size+(20, 20),
position=rect.position-(10, 10))
utils.draw_rectangle(
cr, rect, fill_color=None, outline_color=color)
if self._select_pos:
rect = _get_rectangle(self._select_pos, mouse_pos)
utils.draw_rectangle(
cr, rect, fill_color=None, outline_color=color)
def _capture_image(self, img):
img = cv2.resize(img, (self.widthImg, self.heightImg)) # RESIZE IMAGE
imgBlank = np.zeros((self.heightImg, self.widthImg, 3),
np.uint8) # CREATE A BLANK IMAGE FOR TESTING DEBUGING IF REQUIRED
imgGray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # CONVERT IMAGE TO GRAY SCALE
imgBlur = cv2.GaussianBlur(imgGray, (5, 5), 1) # ADD GAUSSIAN BLUR
thres = utils.get_track_bar_values() # GET TRACK BAR VALUES FOR THRESHOLDS
imgThreshold = cv2.Canny(imgBlur, thres[0], thres[1]) # APPLY CANNY BLUR
kernel = np.ones((5, 5))
imgDial = cv2.dilate(imgThreshold, kernel, iterations = 2) # APPLY DILATION
imgThreshold = cv2.erode(imgDial, kernel, iterations = 1) # APPLY EROSION
## FIND ALL COUNTOURS
imgContours = img.copy() # COPY IMAGE FOR DISPLAY PURPOSES
imgBigContour = img.copy() # COPY IMAGE FOR DISPLAY PURPOSES
contours, hierarchy = cv2.findContours(imgThreshold, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE) # FIND ALL CONTOURS
cv2.drawContours(imgContours, contours, -1, (0, 255, 0), 10) # DRAW ALL DETECTED CONTOURS
# FIND THE BIGGEST COUNTOUR
biggest, maxArea = utils.calc_biggest_contour(contours) # FIND THE BIGGEST CONTOUR
# If connected corners found -> capture image.
if biggest.size != 0:
biggest = utils.reorder_points(biggest)
cv2.drawContours(imgBigContour, biggest, -1, (0, 255, 0), 20) # DRAW THE BIGGEST CONTOUR
imgBigContour = utils.draw_rectangle(imgBigContour, biggest, 2)
pts1 = np.float32(biggest) # PREPARE POINTS FOR WARP
pts2 = np.float32(
[[0, 0], [self.widthImg, 0], [0, self.heightImg], [self.widthImg, self.heightImg]]) # PREPARE POINTS FOR WARP
matrix = cv2.getPerspectiveTransform(pts1, pts2)
imgWarpColored = cv2.warpPerspective(img, matrix, (self.widthImg, self.heightImg))
# REMOVE 20 PIXELS FORM EACH SIDE
imgWarpColored = imgWarpColored[20:imgWarpColored.shape[0] - 20, 20:imgWarpColored.shape[1] - 20]
imgWarpColored = cv2.resize(imgWarpColored, (self.widthImg, self.heightImg))
# APPLY ADAPTIVE THRESHOLD
imgWarpGray = cv2.cvtColor(imgWarpColored, cv2.COLOR_BGR2GRAY)
imgAdaptiveThre = cv2.adaptiveThreshold(imgWarpGray, 255, 1, 1, 9, 2)
imgAdaptiveThre = cv2.bitwise_not(imgAdaptiveThre)
imgAdaptiveThre = cv2.medianBlur(imgAdaptiveThre, 3)
# print("Done")
# Image Array for Display
img_all = ([img, imgGray, imgThreshold, imgContours],
[imgBigContour, imgWarpColored, imgWarpGray, imgAdaptiveThre])
return imgAdaptiveThre, img_all
else:
img_all = ([img, imgGray, imgThreshold, imgContours],
[imgBlank, imgBlank, imgBlank, imgBlank])
return None, img_all
def on_draw(component, app, cr):
lines = list(itertools.chain.from_iterable(
textwrap.wrap(line, width=max_width, tabsize=4)
for line in component.data['text'].split('\n')
))
if lines and not lines[-1]:
lines.pop()
lines = lines[-max_height:]
grid_size = app.grid_size
position = (
component.display.rect.bottom_left +
(grid_size//2, -grid_size//2)
)
rect = shapes.Rectangle(
position=position-(4, 12*max_height),
size=(max_width*7+8, max_height*12+4))
utils.draw_rectangle(cr, rect, (0.9, 0.9, 0.9), (0, 0, 0))
for i, line in enumerate(reversed(lines)):
pos = position - (0, i*12)
utils.draw_text(
cr, line, pos,
h_align=utils.TextHAlign.LEFT,
v_align=utils.TextVAlign.BOTTOM)
def cf(x):
return sv.predict(x)[0] in target_index
def score_cf(imag):
scores = np.array(sv.predict_softmax_score(imag))
return (np.argmax(scores)
in target_index) and (np.max(scores) > param['score_bottom_line'])
for i, img in enumerate(data.test_img):
print("{}th image in progress".format(i))
temp_img, matrix = sw.sliding_window(
img,
score_cf,
window_size=param['ss_dic']['window_size'],
stride=16,
boundary=param['sw_dic']['boundary'])
temp_img.save('./test_result/sw' + str(i) + '.jpg', 'jpeg')
img.save('./Images/sw' + str(i) + '.jpg', 'jpeg')
box_set = sw.make_box(matrix)
for box in box_set:
img = utils.draw_rectangle(img, box, label='block')
xml = utils.box_to_xml('sw' + str(i) + '.xml', path + str(i) + '.jpg',
img.size + (3, ), box_set)
ElementTree(xml).write('./Annotation/sw' + str(i) + '.xml')
txt = utils.box_to_txt([0 for _ in range(len(box_set))], box_set, img.size)
with open('./labels/sw' + str(i) + '.txt', 'w') as f:
f.write(txt)
img.save('./test_result/sw_box' + str(i) + '.jpg', 'jpeg')
def debug_box(webcam, X, Y):
img = load_webcam(webcam)
color = average_color(img, X, Y)
draw_rectangle(img, X, Y, color=color)
return send_file(img_to_io(img), mimetype='image/jpeg')
import cv2
import numpy as np
from utils import draw_rectangle,paint_landmark_points
location = "outputs\\part1_with_rectangle_"
image = cv2.imread("images\\aydemirakbas.png")
draw_rectangle(image,location+"aydemirakbas.png")
image = cv2.imread("images\\deniro.jpg")
draw_rectangle(image,location+"deniro.png")
image = cv2.imread("images\\kimbodnia.png")
draw_rectangle(image,location+"kimbodnia.png")
location = "outputs\\part1_with_landmarks_collage.png"
image = cv2.imread("images\\aydemirakbas.png")
image1 = paint_landmark_points(image)
image = cv2.imread("images\\deniro.jpg")
image2 = paint_landmark_points(image)
image = cv2.imread("images\\kimbodnia.png")
image3 = paint_landmark_points(image)
image = cv2.imread("images\\cat.jpg")
image4 = paint_landmark_points(image,np.load("images\\cat_landmarks.npy"))
image = cv2.imread("images\\gorilla.jpg")
image5 = paint_landmark_points(image,np.load("images\\gorilla_landmarks.npy"))
image = cv2.imread("images\\panda.jpg")
image6 = paint_landmark_points(image,np.load("images\\panda_landmarks.npy"))
top_image = cv2.hconcat([image6, image4, image5]) #Concatenate animal images horizontally
bottom_image = cv2.hconcat([image2, image1, image3]) #Concatenate human images horizontally
collage = cv2.vconcat([top_image, bottom_image]) # Concatenate top_image and bottom_image vertically
cv2.imwrite(location, collage) #Write image on specified place with "location"
RET, FRAME = VID_FEED.read()
if not RET:
print("Unable to capture video")
sys.exit()
if DRAW_HARRIS:
FRAME = draw_harris_kps(FRAME)
# MATCH_DATA = find_homographies(REF_DSC, FRAME)
MATCH_DATA, _ = get_homographies_contour(FRAME, REF_IMAGES, MATCH_DATA,
None)
for ind, homography in enumerate(MATCH_DATA):
# homography = match_tuple[0]
if homography is not None:
projection_matrix, R, T = get_matrix(CAM_PARAMS, homography)
print(R, T)
# FRAME = draw_corners(FRAME, homography)
FRAME = render(FRAME, OBJ, projection_matrix, REF_IMAGES[ind],
False)
if RECTANGLE:
FRAME, _ = draw_rectangle(homography, REF_IMAGES[ind],
FRAME)
cv2.imshow('frame', FRAME)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
VID_FEED.release()
cv2.destroyAllWindows()
def doc_scan_pipeline(input=PATH, output="./img/scanned_doc.jpg"):
img = cv2.imread(input)
# 0. Convert given image from BGR to RGB format
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
h, w, _ = img.shape
img = cv2.resize(img, (width, height))
# 1. Convert to grayscale
img_gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# 2. Add Gaussian blur
img_blur = cv2.GaussianBlur(img_gray, (5, 5), 1)
# 3. Add Canny edge detection
img_threshold = cv2.Canny(img_blur, 100, 200, L2gradient=True)
# 3.1 Apply dilation
kernel = np.ones((3, 3))
img_threshold = cv2.dilate(img_threshold, kernel, iterations=2)
# 4. Find all the contours
img_contours = img.copy()
img_big_contour = img.copy()
contours, hierarchy = cv2.findContours(img_threshold, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cv2.drawContours(image=img_contours,
contours=contours,
contourIdx=-1,
color=(0, 255, 0),
thickness=5)
# 5. Find the biggest contour
biggest, maxArea = biggest_contour(contours)
biggest = reorder(biggest)
cv2.drawContours(image=img_big_contour,
contours=biggest,
contourIdx=-1,
color=(0, 255, 0),
thickness=10)
# 5.1 Draw a rectangle, i.e., 4 lines connecting the 4 dots corresponding to the largest contour
img_big_contour = draw_rectangle(img_big_contour, biggest, thickness=2)
pts1 = np.float32(biggest)
pts2 = np.float32([[0, 0], [width, 0], [0, height], [width, height]])
# 6. Image Warp
# 6.1 Calculate a 3x3 perspective transform matrix
matrix = cv2.getPerspectiveTransform(pts1, pts2)
# 6.2 Apply the perspective matrix to the image
img_warp_coloured = cv2.warpPerspective(img, matrix, (width, height))
# 7. Adaptive thresholding
img_warp_gray = cv2.cvtColor(img_warp_coloured, cv2.COLOR_BGR2GRAY)
img_adaptive_th = cv2.adaptiveThreshold(img_warp_gray, 255, 1,
cv2.THRESH_BINARY, 5, 2)
# 7.1 Apply median blurring to remove tiny speckles of noise
img_adaptive_th = cv2.medianBlur(img_adaptive_th, 3)
# Save the document to disk
cv2.imwrite(output, img_adaptive_th)
# Add labels to each image
img = draw_text(img, "Original")
img_gray = draw_text(img_gray, "Grayscale")
img_blur = draw_text(img_blur, "Gaussian Blur", pos=(int(width / 4), 50))
img_threshold = draw_text(img_threshold,
"Canny Edge",
pos=(int(width / 4), 50))
img_contours = draw_text(img_contours, "Contours")
img_big_contour = draw_text(img_big_contour,
"Largest Contour",
pos=(int(width / 7), 50))
img_warp_coloured = draw_text(img_warp_coloured,
"Warp",
pos=(int(width / 3), 50))
img_adaptive_th = draw_text(img_adaptive_th,
"Adaptive Thresholding",
pos=(int(width / 7), 50),
font_scale=2,
font_thickness=6)
blank_img = np.zeros((height, width, 3), dtype=np.uint8)
image_list = [
img, img_gray, img_blur, img_threshold, img_contours, img_big_contour,
img_warp_coloured, img_adaptive_th
]
# Combine the images into a grid
# image_grid returns PIL image, np.asarray() can be used to convert it back to cv2 compatible format
grid = np.asarray(image_grid(image_list, width, height))
Runner script to test during development.
"""
import cv2 as cv
import numpy as np
import matplotlib.pyplot as plt
from augment import Pipeline
from horizontal import RandomHorizontalFlip
from scale import RandomScale, Scale
from translate import RandomTranslate, Translate
from rotate import RandomRotate
from utils import draw_rectangle
image = cv.imread("styrofoam cup8.jpg")[:, :, ::-1]
bbox = np.array([[402, 0, 799, 270, 1]]).astype('float64')
plt.imshow(draw_rectangle(image, bbox))
plt.show()
print(bbox, type(bbox))
transforms = Pipeline(
[RandomHorizontalFlip(1),
Translate(x=0.2, y=0.2),
RandomRotate((-2, 2))])
image, bbox = transforms(image, bbox)
print(bbox, type(bbox))
plt.imshow(draw_rectangle(image, bbox))
plt.show()