def bonne_taille(filename):
repo = "separations/"+filename+"/"
liste_images = os.listdir(repo)
a_conserver = []
for img in liste_images :
a_conserver, repo = crop.main(repo+img, repo, a_conserver)
for i in os.listdir(repo):
if i not in a_conserver:
try :
os.remove(repo+"/"+i)
except :
...
else :
resize.main(repo + i)
bw.binarize(repo + i)
def trainingSetFromImage(trainImage, target):
grayImage = binarize.toGrayScale(trainImage)
binarizedImage = binarize.binarize(grayImage)
tmpBinImage = copy.deepcopy(binarizedImage)
height, width = binarizedImage.shape
imageRegionSet = segmentation.segmentation(binarizedImage)
i = 1
font = cv2.FONT_HERSHEY_SIMPLEX
for imageRegion in imageRegionSet:
dim = imageRegion[2]
(x, y, w, h) = dim
cv2.rectangle(tmpBinImage, (x, height - y - h), (x + w, height - y), (255, 255, 255), 2)
cv2.putText(tmpBinImage, str(i), (x, height - y - h - 20), font, 0.4, (255, 255, 255), 1, cv2.LINE_AA)
i = i + 1
cv2.imshow("Training Image", tmpBinImage)
k = cv2.waitKey(0)
cv2.destroyAllWindows()
print 'Number of Segments : ' + str(len(imageRegionSet))
while True:
print "(i)Include Training Set (r)Retry (q)Quit"
option = str(raw_input("Option : "))
if option == 'q':
break
elif option == 'i':
includeTrainingSet(imageRegionSet, target)
break
elif option == 'r':
break
return option
def upload_multipart():
if 'uploadFile' not in request.files:
make_response(jsonify({'result': 'uploadFile is required.'}))
file = request.files['uploadFile']
filename = file.filename
if '' == filename:
make_response(jsonify({'result': 'filename must not empty.'}))
filename = werkzeug.utils.secure_filename(filename)
file_path = os.path.join(UPLOAD_DIR, filename)
file.save(file_path)
img_rows, img_cols = 400, 400
img_channels = 1
img = cv2.imread(file_path, cv2.IMREAD_COLOR)
os.remove(file_path)
binary = cv2.resize(binarize(img), (img_rows, img_cols))
img2 = np.zeros((img_rows, img_cols, img_channels))
img2[:, :, 0] = binary
img2 = np.expand_dims(img2, axis=0)
with graph.as_default():
pred = model.predict(img2)
pred = pred.flatten()
# predicted_class_indices = np.argmax(pred)
labels = ['FBMessanger', 'Instagram', 'Invalid', 'LINE', 'Twitter']
pred_per = list(map(lambda x: x*100, pred))
return make_response(jsonify(dict(zip(labels, pred_per))))
def test(imageFileName, image=None, flag=0):
if flag == 0:
image = getImage(imageFileName)
grayImage = binarize.toGrayScale(image)
binarizedImage = binarize.binarize(grayImage)
cv2.imshow('image', binarizedImage)
regions = []
res, contours, hierarchy = cv2.findContours(binarizedImage,
cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
for contour in contours:
[x, y, w, h] = cv2.boundingRect(contour)
if h < 30 and w < 30:
continue
else:
regions.append((binarizedImage[y:y + h, x:x + w], (x, y, w, h)))
classifier = learning.getClassifier(svm_nn)
for region in regions:
testData = learning.getResizedImage(region[0])
(x, y, w, h) = region[1]
result = classifier.predict([testData])
cv2.rectangle(image, (x, y), (x + w, y + h), (0, 0, 0), 2)
cv2.putText(image, str(result[0]), (x, y - 20),
cv2.FONT_HERSHEY_SIMPLEX, 0.4, (0, 0, 0), 1, cv2.LINE_AA)
print result
cv2.imshow('test', image)
cv2.waitKey(0)
cv2.destroyAllWindows()
def test_booleans(self):
self.assertEqual(b'\x00', binarize(False, endian=Endian.little, width=1))
self.assertEqual(b'\x01', binarize(True, endian=Endian.little, width=1))
self.assertEqual(b'\x00', binarize(False, endian=Endian.big, width=1))
self.assertEqual(b'\x01', binarize(True, endian=Endian.big, width=1))
self.assertEqual(b'\x00\x00', binarize(False, endian=Endian.little, width=2))
self.assertEqual(b'\x01\x00', binarize(True, endian=Endian.little, width=2))
self.assertEqual(b'\x00\x00', binarize(False, endian=Endian.big, width=2))
self.assertEqual(b'\x00\x01', binarize(True, endian=Endian.big, width=2))
def test(models_dir, image_path):
im = cv2.imread(image_path, cv2.IMREAD_UNCHANGED)
bw = binarize.binarize(im, algorithm=binarize.ntirogiannis2014)
AH = algorithm.dominant_char_height(bw)
print('AH =', AH)
possible_AHs = np.array([int(d) for d in os.listdir(models_dir) if d.isdigit()])
size = possible_AHs[np.abs(possible_AHs - AH).argmin()]
model_dir = os.path.join(models_dir, str(size))
P_l_T, P_h_T, pca, A_l_T, Q_T, A_h_T = load_model(model_dir)
def open_pdf(link, cnum):
"""
This module opens the link in selenium
Takes a screenshot and saves it
Crops the captcha part
Binarizes the captcha
SOlves the captcha to get the text
Fills in the text and clicks submit
If captcha is correct the PDF opens for preview.
If captcha is wrong it tries again until the pdf opens.
"""
ndir = getCurrentDirectory(cnum)
if not os.path.exists(ndir):
os.makedirs(ndir)
driver = captcha_download.download_captcha(link, ndir)
crop_captcha.crop_captcha()
binarize.binarize()
key = captcha_solver.resolve("modified.png")
key = simplify(key)
rval = captcha_download.fill_captcha(driver, key)
try:
p = driver.find_element_by_id("lblCaptchaMessage")
if p.text == 'Please enter correct captcha !':
print "Wrong captcha"
driver.close()
return 0
else:
# time.sleep(40)
while not os.path.isfile(fname):
pass
driver.close()
return 1
except:
while not os.path.isfile(fname):
pass
driver.close()
return 1
def main():
import sys
classifier = cv2.SVM()
classifier.load(sys.argv[1])
im = cv2.imread(sys.argv[2])
im = cv2.GaussianBlur(im, (3, 3), 0)
imshow_large(__file__, im)
cv2.waitKey()
unrotated = unrotate(im)
#
# Check if image is right way up, correct otherwise
#
imshow_large(__file__, unrotated)
key = cv2.waitKey()
if key & 0xff == ord("r"):
unrotated = cv2.flip(cv2.flip(unrotated, 0), 1)
imshow_large(__file__, unrotated)
cv2.waitKey()
binarized = binarize(unrotated)
rois = get_rois(binarized)
results = {}
grayscale = cv2.cvtColor(unrotated, cv.CV_BGR2GRAY)
for (x, y, width, height) in rois:
roi = grayscale[y:y + height, x:x + width]
vec = extract_features(roi)
label = classifier.predict(vec)
results[(x, y, width, height)] = "01234567890X"[int(label)]
scale = SCREEN_HEIGHT / unrotated.shape[0]
unrotated = cv2.cvtColor(grayscale, cv.CV_GRAY2BGR)
scaled = cv2.resize(unrotated, (0, 0), fx=scale, fy=scale)
for roi in rois:
x = int(roi[0] * scale)
y = int(roi[1] * scale)
width = int(roi[2] * scale)
height = int(roi[3] * scale)
cv2.rectangle(scaled, (x, y), (x + width, y + height), (0, 255, 0, 0),
1)
if results[roi] == "X":
continue
cv2.putText(scaled, results[roi], (x, y), cv.CV_FONT_HERSHEY_SIMPLEX,
1, (0, 255, 0, 0))
cv2.imshow(__file__, scaled)
cv2.waitKey()
def main():
import sys
classifier = cv2.SVM()
classifier.load(sys.argv[1])
im = cv2.imread(sys.argv[2])
im = cv2.GaussianBlur(im, (3,3), 0)
imshow_large(__file__, im)
cv2.waitKey()
unrotated = unrotate(im)
#
# Check if image is right way up, correct otherwise
#
imshow_large(__file__, unrotated)
key = cv2.waitKey()
if key & 0xff == ord("r"):
unrotated = cv2.flip(cv2.flip(unrotated, 0), 1)
imshow_large(__file__, unrotated)
cv2.waitKey()
binarized = binarize(unrotated)
rois = get_rois(binarized)
results = {}
grayscale = cv2.cvtColor(unrotated, cv.CV_BGR2GRAY)
for (x,y,width,height) in rois:
roi = grayscale[y:y+height,x:x+width]
vec = extract_features(roi)
label = classifier.predict(vec)
results[(x,y,width,height)] = "01234567890X"[int(label)]
scale = SCREEN_HEIGHT/unrotated.shape[0]
unrotated = cv2.cvtColor(grayscale, cv.CV_GRAY2BGR)
scaled = cv2.resize(unrotated, (0,0), fx=scale, fy=scale)
for roi in rois:
x = int(roi[0]*scale)
y = int(roi[1]*scale)
width = int(roi[2]*scale)
height = int(roi[3]*scale)
cv2.rectangle(scaled, (x,y), (x+width, y+height), (0,255,0,0), 1)
if results[roi] == "X":
continue
cv2.putText(scaled, results[roi], (x, y), cv.CV_FONT_HERSHEY_SIMPLEX, 1, (0, 255, 0, 0))
cv2.imshow(__file__, scaled)
cv2.waitKey()
def extractBlobs(img, doFilter=True, forOCR=False): #img=cv.cvtColor(cImg, cv.COLOR_BGR2GRAY) bImg=binarize(img, 0.5) if forOCR: mode=cv.RETR_EXTERNAL else: mode=cv.RETR_LIST contours,hierarchy = cv.findContours(bImg, mode, cv.CHAIN_APPROX_NONE) print(len(contours)) H, W = bImg.shape area=H*W bxs=[] for cnt in contours: if((not doFilter) or (cv.contourArea(cnt)>area/100 and cv.contourArea(cnt)<area/20)): x,y,w,h = cv.boundingRect(cnt) bxs.append((y,x,h,w)) return bxs
def main():
import sys
in_dir = sys.argv[1]
images = map(lambda x: P.join(in_dir, x), os.listdir(in_dir))
out_dir = sys.argv[2]
counters = {}
for subdir in "0123456789X":
path = P.join(out_dir, subdir)
if not P.isdir(path):
os.makedirs(path)
files = os.listdir(path)
indices = map(lambda x: int(P.splitext(x)[0]), files)
try:
counters[subdir] = max(indices) + 1
except ValueError:
counters[subdir] = 0
for fpath in images:
print fpath
im = cv2.imread(fpath)
im = cv2.GaussianBlur(im, (3, 3), 0)
imshow_large(__file__, im)
key = cv2.waitKey()
if key & 0xff == ord("q"):
sys.exit(0)
elif key & 0xff == ord("n"):
continue
unrotated = unrotate(im)
#
# Check if image is right way up, correct otherwise
#
imshow_large(__file__, unrotated)
key = cv2.waitKey()
if key & 0xff == ord("r"):
unrotated = cv2.flip(cv2.flip(unrotated, 0), 1)
imshow_large(__file__, unrotated)
cv2.waitKey()
elif key & 0xff == ord("n"):
continue
binarized = binarize(unrotated)
scale = SCREEN_HEIGHT / unrotated.shape[0]
colorbin = copy.deepcopy(unrotated)
colorbin = cv2.resize(colorbin, (0, 0), fx=scale, fy=scale)
rois = get_rois(binarized)
for (x, y, width, height) in rois:
x = int(x * scale)
y = int(y * scale)
width = int(width * scale)
height = int(height * scale)
cv2.rectangle(colorbin, (x, y), (x + width, y + height),
(255, 0, 0, 0), 1)
cv2.imshow(__file__, colorbin)
key = cv2.waitKey()
if key & 0xff == ord("q"):
break
elif key & 0xff == ord("n"):
continue
for (x, y, width, height) in rois:
colorbin2 = copy.deepcopy(colorbin)
x_ = int(x * scale)
y_ = int(y * scale)
width_ = int(width * scale)
height_ = int(height * scale)
cv2.rectangle(colorbin2, (x_, y_), (x_ + width_, y_ + height_),
(0, 0, 255, 0), 1)
sub = unrotated[y:y + height, x:x + width]
supers = cv2.resize(sub, (192, 192), interpolation=cv.CV_INTER_NN)
cv2.imshow(__file__, colorbin2)
cv2.imshow("supers", supers)
key = cv2.waitKey()
if key & 0xff == ord("q"):
sys.exit(0)
elif key & 0xff == ord("n"): # move on to the next image
break
elif (key & 0xff) in map(ord, "0123456789"):
digit = chr(key & 0xff)
subdir = P.join(out_dir, digit)
out_file = P.join(subdir, str(counters[digit]) + ".png")
cv2.imwrite(out_file, sub)
counters[digit] += 1
else:
subdir = P.join(out_dir, "X")
out_file = P.join(subdir, str(counters["X"]) + ".png")
cv2.imwrite(out_file, sub)
counters["X"] += 1
def kim2014(orig, O=None, split=False, n_points_w=None, n_tries=30):
lib.debug_imwrite('gray.png', binarize.grayscale(orig))
im = binarize.binarize(orig, algorithm=lambda im: binarize.sauvola_noisy(im, k=0.1))
global bw
bw = im
im_h, im_w = im.shape
AH, lines, _ = get_AH_lines(im)
if O is None:
O = np.array((im_w / 2.0, im_h / 2.0))
if split:
# Test if line start distribution is bimodal.
line_xs = np.array([line.left() for line in lines])
bimodal = line_xs.std() / im_w > 0.10
dual = bimodal and im_w > im_h
else:
dual = False
if dual:
print('Bimodal! Splitting page!')
pages = crop.split_lines(lines)
n_points_w = 1.2 * np.percentile(np.array([line.width() for line in lines]), 90)
n_points_w = max(n_points_w, 1800)
if lib.debug:
debug = cv2.cvtColor(bw, cv2.COLOR_GRAY2BGR)
for page in pages:
page_crop = Crop.from_lines(page).expand(0.005)
# print(page_crop)
page_crop.draw(debug)
lib.debug_imwrite('split.png', debug)
page_crops = [Crop.from_lines(page) for page in pages]
if len(page_crops) == 2:
[c0, c1] = page_crops
split_x = (c0.x1 + c1.x0) / 2
page_crops = [
c0.union(Crop(0, 0, split_x, im_h)),
c1.union(Crop(split_x, 0, im_w, im_h))
]
result = []
for i, (page, page_crop) in enumerate(zip(pages, page_crops)):
print('==== PAGE {} ===='.format(i))
lib.debug_prefix.append('page{}'.format(i))
page_image = page_crop.apply(orig)
page_bw = page_crop.apply(im)
page_AH, page_lines, _ = get_AH_lines(page_bw)
new_O = O - np.array((page_crop.x0, page_crop.y0))
lib.debug_imwrite('precrop.png', im)
lib.debug_imwrite('page.png', page_image)
bw = page_bw
dewarper = Kim2014(page_image, page_bw, page_lines, [page_lines],
new_O, page_AH, n_points_w)
result.append(dewarper.run_retry()[0])
lib.debug_prefix.pop()
return result
else:
lib.debug_prefix.append('page0')
dewarper = Kim2014(orig, im, lines, [lines], O, AH, n_points_w)
lib.debug_prefix.pop()
return dewarper.run_retry(n_tries=n_tries)
def test_negative_integers(self):
self.assertEqual(b'\x80', binarize(-128, endian=Endian.little, width=1))
self.assertEqual(b'\x7f', binarize(-129, endian=Endian.little, width=1))
self.assertEqual(b'\xff\xff', binarize(-1, endian=Endian.little, width=2))
self.assertEqual(b'\xfe\xff', binarize(-2, endian=Endian.little, width=2))
def test_positive_big_endian_integers(self):
# some "normal" integers
self.assertEqual(b'\x00\x00\x00\x00', binarize(0, endian=Endian.big, width=4))
self.assertEqual(b'\x00\x00\x00\x01', binarize(1, endian=Endian.big, width=4))
self.assertEqual(b'\x00\x00\x04\x00', binarize(1024, endian=Endian.big, width=4))
# 1->2 byte boundary
self.assertEqual(b'\x00\x00\x00\xff', binarize(255, endian=Endian.big, width=4))
self.assertEqual(b'\x00\x00\x01\x00', binarize(256, endian=Endian.big, width=4))
# 2->3 byte boundary
self.assertEqual(b'\x00\x00\xff\xff', binarize(65535, endian=Endian.big, width=4))
self.assertEqual(b'\x00\x01\x00\x00', binarize(65536, endian=Endian.big, width=4))
# 3->4 byte boundary
self.assertEqual(b'\x00\xff\xff\xff', binarize(16777215, endian=Endian.big, width=4))
self.assertEqual(b'\x01\x00\x00\x00', binarize(16777216, endian=Endian.big, width=4))
# 4->5 byte boundary
self.assertEqual(b'\xff\xff\xff\xff', binarize(4294967295, endian=Endian.big, width=4))
self.assertRaises(TypeError, binarize, 4294967296, endian=Endian.big, width=4)
# large width: 8->9 byte boundary
self.assertEqual(b'\xff\xff\xff\xff\xff\xff\xff\xff', binarize((2**64)-1, endian=Endian.big, width=8))
self.assertRaises(TypeError, binarize, 2**64, endian=Endian.big, width=8)
# "weird" width
self.assertEqual(b'\x00\x10\xbc', binarize(4284, endian=Endian.big, width=3))
self.assertEqual(b'\xff\xff\xff', binarize(16777215, endian=Endian.big, width=3))
self.assertRaises(TypeError, binarize, 16777216, endian=Endian.big, width=3)
def upscale(path, data_dir, factor):
assert factor == 2
im = cv2.imread(path, cv2.IMREAD_UNCHANGED)
im_h, im_w = im.shape
bw = binarize.binarize(im, algorithm=binarize.ntirogiannis2014)
AH = dominant_char_height(bw)
print('AH =', AH)
possible_AHs = np.array(
[int(d) for d in os.listdir(data_dir) if d.isdigit()])
size = possible_AHs[np.abs(possible_AHs - AH).argmin()]
D_T_coupled = np.load(os.path.join(data_dir, str(size), 'dict.npy'))
W_l = int(size / 3) | 1
W_h = 2 * W_l
step = 3
assert D_T_coupled.shape[1] == W_l * W_l + W_h * W_h
D_T = D_T_coupled[:, :W_l * W_l]
K = D_T.shape[0]
lam = 0.2 # weight of sparsity. TODO: confirm same as training data.
lo_patches = patches(im, W_l, step)
struct = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
gradient = cv2.morphologyEx(im, cv2.MORPH_GRADIENT, struct)
gradient_means, _ = binarize.mean_std(gradient, W_l)
patch_gradient = gradient_means[W_l // 2:-W_l // 2 + 1:step,
W_l // 2:-W_l // 2 + 1:step]
assert patch_gradient.shape == (lo_patches.shape[0], lo_patches.shape[1])
patch_mask = patch_gradient > np.percentile(patch_gradient, 50)
good_patches = np.nonzero(patch_mask)
cv2.imwrite('spp.png', -patch_mask.astype(np.uint8))
lo_patches_vec = lo_patches[good_patches].reshape(-1, W_l * W_l).astype(
np.float64)
means = lo_patches_vec.mean(axis=1)
X_T = lo_patches_vec - means[:, newaxis]
t = X_T.shape[0]
Z_T = np.zeros((t, K), dtype=np.float64)
last_objective = None
while True:
training.feature_sign_search_vec(X_T, Z_T, D_T, lam)
objective = np.square(X_T - Z_T.dot(D_T)).sum()
print('\ncurrent objective:', objective)
if last_objective is not None:
relative_err = abs(last_objective - objective) / last_objective
print('relative error:', relative_err)
if relative_err < 1e-4:
break
last_objective = objective
hi_patches = Z_T.dot(D_T_coupled)[:, W_l * W_l:] + means[:, newaxis]
hi_float = np.zeros((factor * im_h, factor * im_w), dtype=np.float64)
dest_patches = patches(hi_float, W_h, 2 * step)
patch_count = np.zeros((factor * im_h, factor * im_w), dtype=int)
count_patches = patches(patch_count, W_h, 2 * step)
for i, patch in enumerate(hi_patches):
dest_patches[good_patches[i]] += patch
count_patches[good_patches[i]] += 1
np.divide(hi_float, patch_count, hi_float, where=patch_count > 0)
hi_lanczos = cv2.resize(im, (0, 0),
None,
2.,
2.,
interpolation=cv2.INTER_LANCZOS4)
return np.where(patch_count > 0, hi_float,
hi_lanczos).clip(0, 255).astype(np.uint8)
def process_image(original, dpi=None):
original_rot90 = original
for i in range(args.rotate / 90):
original_rot90 = np.rot90(original_rot90)
# original_rot90 = cv2.resize(original_rot90, (0, 0), None, 1.5, 1.5)
im_h, im_w = original_rot90.shape[:2]
# image height should be about 10 inches. round to 100
if not dpi:
dpi = int(round(im_h / 1100.0) * 100)
print('detected dpi:', dpi)
split = im_w > im_h # two pages
cropped_images = []
if args.dewarp:
lib.debug_prefix.append('dewarp')
dewarped_images = dewarp.kim2014(original_rot90)
for im in dewarped_images:
bw = binarize.binarize(im, algorithm=binarize.sauvola, resize=1.0)
lib.debug_prefix.append('crop')
_, [lines] = crop(im, bw, split=False)
lib.debug_prefix.pop()
c = Crop.from_lines(lines)
if c.nonempty():
cropped_images.append(Crop.from_whitespace(bw).apply(im))
lib.debug_prefix.pop()
else:
bw = binarize.binarize(original_rot90, algorithm=binarize.adaptive_otsu, resize=1.0)
debug_imwrite('thresholded.png', bw)
AH, line_sets = crop(original_rot90, bw, split=split)
for lines in line_sets:
c = Crop.from_lines(lines)
if c.nonempty():
lib.debug = False
bw_cropped = c.apply(bw)
orig_cropped = c.apply(original_rot90)
angle = algorithm.skew_angle(bw_cropped, original_rot90, AH, lines)
if not np.isfinite(angle): angle = 0.
rotated = algorithm.safe_rotate(orig_cropped, angle)
rotated_bw = binarize.binarize(rotated, algorithm=binarize.adaptive_otsu)
_, [new_lines] = crop(rotated, rotated_bw, split=False)
# dewarped = algorithm.fine_dewarp(rotated, new_lines)
# _, [new_lines] = crop(rotated, rotated_bw, split=False)
new_crop = Crop.union_all([line.crop() for line in new_lines])
if new_crop.nonempty():
# cropped = new_crop.apply(dewarped)
cropped = new_crop.apply(rotated)
cropped_images.append(cropped)
out_images = []
lib.debug_prefix.append('binarize')
for i, cropped in enumerate(cropped_images):
lib.debug_prefix.append('page{}'.format(i))
if lib.is_bw(original_rot90):
out_images.append(binarize.otsu(cropped))
else:
out_images.append(
binarize.ng2014_fallback(binarize.grayscale(cropped))
)
lib.debug_prefix.pop()
lib.debug_prefix.pop()
return dpi, out_images
def main():
import sys
in_dir = sys.argv[1]
images = map(lambda x: P.join(in_dir, x), os.listdir(in_dir))
out_dir = sys.argv[2]
counters = {}
for subdir in "0123456789X":
path = P.join(out_dir, subdir)
if not P.isdir(path):
os.makedirs(path)
files = os.listdir(path)
indices = map(lambda x: int(P.splitext(x)[0]), files)
try:
counters[subdir] = max(indices) + 1
except ValueError:
counters[subdir] = 0
for fpath in images:
print fpath
im = cv2.imread(fpath)
im = cv2.GaussianBlur(im, (3,3), 0)
imshow_large(__file__, im)
key = cv2.waitKey()
if key & 0xff == ord("q"):
sys.exit(0)
elif key & 0xff == ord("n"):
continue
unrotated = unrotate(im)
#
# Check if image is right way up, correct otherwise
#
imshow_large(__file__, unrotated)
key = cv2.waitKey()
if key & 0xff == ord("r"):
unrotated = cv2.flip(cv2.flip(unrotated, 0), 1)
imshow_large(__file__, unrotated)
cv2.waitKey()
elif key & 0xff == ord("n"):
continue
binarized = binarize(unrotated)
scale = SCREEN_HEIGHT/unrotated.shape[0]
colorbin = copy.deepcopy(unrotated)
colorbin = cv2.resize(colorbin, (0,0), fx=scale, fy=scale)
rois = get_rois(binarized)
for (x,y,width,height) in rois:
x = int(x*scale)
y = int(y*scale)
width = int(width*scale)
height = int(height*scale)
cv2.rectangle(colorbin, (x,y), (x+width, y+height), (255,0,0,0), 1)
cv2.imshow(__file__, colorbin)
key = cv2.waitKey()
if key & 0xff == ord("q"):
break
elif key & 0xff == ord("n"):
continue
for (x,y,width,height) in rois:
colorbin2 = copy.deepcopy(colorbin)
x_ = int(x*scale)
y_ = int(y*scale)
width_ = int(width*scale)
height_ = int(height*scale)
cv2.rectangle(colorbin2, (x_,y_), (x_+width_, y_+height_), (0,0,255,0), 1)
sub = unrotated[y:y+height, x:x+width]
supers = cv2.resize(sub, (192, 192), interpolation=cv.CV_INTER_NN)
cv2.imshow(__file__, colorbin2)
cv2.imshow("supers", supers)
key = cv2.waitKey()
if key & 0xff == ord("q"):
sys.exit(0)
elif key & 0xff == ord("n"): # move on to the next image
break
elif (key & 0xff) in map(ord, "0123456789"):
digit = chr(key & 0xff)
subdir = P.join(out_dir, digit)
out_file = P.join(subdir, str(counters[digit]) + ".png")
cv2.imwrite(out_file, sub)
counters[digit] += 1
else:
subdir = P.join(out_dir, "X")
out_file = P.join(subdir, str(counters["X"]) + ".png")
cv2.imwrite(out_file, sub)
counters["X"] += 1
def preprocess(image):
height, width, channels = image.shape
grayImage = binarize.toGrayScale(image)
binarizedImage = binarize.binarize(grayImage)
segmentList = segmentation.segmentation(binarizedImage)
return segmentList
assert np.all(segments[:, 0] < segments[:, 1])
theta_p = np.zeros((len(ellipses), ))
s_p = np.full((len(ellipses), ), 5)
def V_pq_sites(p1, p2, l1, l2):
s_1, theta_1 = unpack_label(l1)
s_2, theta_2 = unpack_label(l2)
centroid_1, centroid_2 = centroids[(p1, p2), :]
d_pq_sq = np.square(centroid_1 - centroid_2).sum()
scale = np.exp(-k * d_pq_sq / (s_values[s_1]**2 + s_values[s_2]**2))
f_diff = abs(s_1 - s_2) + abs(theta_1 - theta_2)
mu = 0 if l1 == l2 else (lam_1 if f_diff <= 3 else lam_2)
return mu * scale
# graph.set_smooth_cost_function(V_pq_sites)
# graph.expansion()
if __name__ == '__main__':
lib.debug = True
orig = cv2.imread(sys.argv[1])
bw = binarize.binarize(orig, algorithm=binarize.sauvola)
lib.debug_imwrite('retinex.png', bw)
AH = algorithm.dominant_char_height(bw)
koo2010(bw, AH)
with open('bbox.txt', 'rb') as f:
bxs=pickle.load(f)
fname='cars_markus/image_0011.jpg'
# trImg = cv.imread('train_img.png',cv.IMREAD_GRAYSCALE)
cImg = cv.imread(fname)
img=cv.cvtColor(cImg, cv.COLOR_BGR2GRAY)
# bx = [226, 381, 75, 32]
# trImg = trImg[bx[0]:bx[0]+bx[2],bx[1]:bx[1]+bx[3]]
bx=bxs[0][1]
cimgD = cImg[:][bx[0]:bx[0]+bx[2],bx[1]:bx[1]+bx[3]]
# print(bxs)
imgD = img[bx[0]:bx[0]+bx[2],bx[1]:bx[1]+bx[3]];
bimgD=binarize(imgD, 0.5)
bxs=extractBlobs(imgD)
model = getModel();
# clbp = findLBP(trImg);
# chist = lbpHist(clbp);
# chist = chist.reshape((1, 69)).astype(np.float32);
# retval, results, neigh_resp, dists = model.find_nearest(chist, k = 1)
# string = chr(int((results[0][0])))
# print(string)
i=0
for cbx in bxs:
y,x,h,w=cbx
print(cbx)