def check_on_length(wrong_file, right_file, ending=2000):
fWrong = open(wrong_file, "w")
fRight = open(right_file, "w")
cadidate = open("textOnly.txt")
cadidate = cadidate.readlines()
total = 0
wrong = 0
widgets = [
'Test: ',
Percentage(), ' ',
Bar(marker='0', left='[', right=']'), ' ',
ETA(), ' ',
FileTransferSpeed()
] # see docs for other options
pbar = ProgressBar(widgets=widgets, maxval=ending)
pbar.start()
iii = 0
for file in cadidate:
file = file.rstrip()
print(file)
pbar.update(iii)
iii += 1
file = file.replace(".txt", "")
img = cv2.imread("scanned\\" + file + ".png")
ftext = open("text\\" + file + ".txt", "r", encoding="utf-8")
## image segmentation
text = ftext.read()
wordsArray = nltk.word_tokenize(text)
thre = binarize(img)
rotated = textSkewCorrection(thre)
wordList = wordSegmentation(rotated)
print(file, len(wordsArray), len(wordList))
total += 1
if len(wordList) != len(wordsArray):
wrong += 1
stt = "{0:<20} {1:<20} {2:<20}".format(file, str(len(wordList)),
str(len(wordsArray)))
fWrong.write(stt + "\n")
print("************wrong Image************")
write_tokenz(wordsArray)
#showWordCuts(img,wordList)
break
else:
fRight.write(file + ".txt" + "\n")
print("************corrent Image************")
ftext.close()
if total == ending:
break
fWrong.close()
fRight.close()
pbar.finish()
def segmentationFromPath(imagePath, ticktock):
# return segmentationFromPath_ammar(imagePath)
img = cv2.imread(imagePath)
ticktock.tick("start time")
thre = binarize(img)
rotated = textSkewCorrection(thre)
wordsFromImage = charSegmentation(rotated)
return wordsFromImage, rotated
def main():
for i in range(0, 1):
file = random.choice(os.listdir("scanned\\"))
img = cv2.imread("capr2.png") #"scanned\\"+file)
thre = binarize(img)
rotated = textSkewCorrection(thre)
wordList = charSegmentation(rotated)
def extract(self):
folders = glob(f'{self.datasetPath}\\*')
for folder in folders:
print(f"currently in dire: {folder}")
for test_case in glob(folder + '/*'):
for f in glob(test_case + '/*.png'):
outputFolder = self.outPath + f[7:-4] + ".txt"
# characterImage=preprocessImage(f)
img = cv2.imread(f)
img = cv2.resize(img, (28, 28))
img = binarize(img)
self.writeFeatureVector(outputFolder, img.flatten())
def main():
for i in range(0, 10):
file = random.choice(os.listdir("scanned\\"))
img = cv2.imread("scanned\\" + file)
thre = binarize(img)
rotated = textSkewCorrection(thre)
wordList = wordSegmentation(rotated)
for word in wordList:
c1, c2 = word['rows']
r1_offset, r1_offset2 = word['columns']
if len(word['subwords']) == 0:
cv2.imshow("sub word ", rotated[c1:c2, r1_offset:r1_offset2])
cv2.waitKey(0)
else:
for subword in word['subwords']:
r1, r2 = subword
cv2.imshow("sub word ",
rotated[c1:c2, r1_offset + r1:r1_offset + r2])
cv2.waitKey(0)
check_on_length("wrong.txt", "right.txt", ending=5000)
f = open("right.txt", "r")
fw = open("output.txt", "w", encoding="utf-8")
files = f.read().split()
for file in files:
img = cv2.imread("scanned\\" + file.replace(".txt", "") + ".png")
ftext = open("text\\" + file, "r", encoding="utf-8")
## image segmentation
# text preprocessing
text = ftext.read().replace(":", "").replace(".", "").replace("-", "")
# tokenizations
wordsArray = nltk.word_tokenize(text)
for word in wordsArray:
fw.writelines([word, " ", str(word.encode("utf8-")), "\n"])
thre = binarize(img)
rotated = textSkewCorrection(thre)
wordList = charSegmentation(rotated)
createDataSet(rotated, wordList, wordsArray, nonConnChOneSide,
nonConnChTwoSide, englishName, file)
f.close()
fw.close()
def imageToFeatureVector(imagePath):
img = cv2.imread(imagePath)
img = cv2.resize(img, (28, 28))
img = binarize(img)
return img.flatten()