def datasetPreprocessing(label_file, folder, output):
if folder[-1] != '/':
folder += '/'
if output[-1] != '/':
output += '/'
length = len(open(label_file).readlines())
with open(label_file) as fp:
print("Processing files:")
for i, line in enumerate(fp):
printProgressBar(i, length)
if line[0] != '#':
l = line.strip().split(" ")
if l[1] == 'ok':
impath = (folder + l[0].split('-')[0] + '/' +
l[0].split('-')[0] + '-' + l[0].split('-')[1] +
'/' + l[0] + '.png')
img = cv2.imread(impath)
if img is not None:
img = imageNorm(img,
60,
border=False,
tilt=True,
hystNorm=True)
cv2.imwrite(
output + "%s_%s.jpg" % (l[-1], time.time()), img)
def recognise(self, img):
""" Recognising word and printing it """
# Pre-processing the word
img = imageNorm(img, 60, border=False, tilt=True, hystNorm=True)
# Separate letters
img = cv2.copyMakeBorder(img,
0,
0,
30,
30,
cv2.BORDER_CONSTANT,
value=[0, 0, 0])
gaps = charSeg.segmentation(img, RNN=True, debug=True)
chars = []
for i in range(len(gaps) - 1):
char = img[:, gaps[i]:gaps[i + 1]]
# TODO None type error after treshold
char, dim = letterNorm(char, is_thresh=True, dim=True)
# TODO Test different values
if dim[0] > 4 and dim[1] > 4:
chars.append(char.flatten())
chars = np.array(chars)
word = ''
if len(chars) != 0:
pred = self.charClass.run(chars)
for c in pred:
word += idx2char(c)
# print("Word: " + word)
return word
def loadImages(dataloc, idx=0, num=None):
""" Load images and labels """
print("Loading words...")
dataloc += '/' if dataloc[-1] != '/' else ''
# Load images and short them from the oldest to the newest
imglist = glob.glob(os.path.join(dataloc, u'*.jpg'))
imglist.sort(key=lambda x: float(x.split("_")[-1][:-4]))
tmpLabels = [name[len(dataloc):] for name in imglist]
labels = np.array(tmpLabels)
images = np.empty(len(imglist), dtype=object)
if num is None:
upper = len(imglist)
else:
upper = min(idx + num, len(imglist))
num += idx
for i, img in enumerate(imglist):
# TODO Speed up loading - Normalization
if i >= idx and i < upper:
images[i] = imageNorm(cv2.cvtColor(cv2.imread(img),
cv2.COLOR_BGR2RGB),
height=60,
border=False,
tilt=True,
hystNorm=True)
printProgressBar(i - idx, upper - idx - 1)
print()
return (images[idx:num], labels[idx:num])
(0, 255, 0), 1)
implt(img, t="Separated characters")
return gaps
# In[5]:
img = cv2.imread("./output/words/normal/0.jpg")
implt(img)
# In[6]:
img = cv2.imread("./output/words/normal/85.jpg")
implt(img)
img = imageNorm(img, 60, border=False, tilt=True, hystNorm=True)
implt(img)
img = cv2.copyMakeBorder(img,
0,
0,
30,
30,
cv2.BORDER_CONSTANT,
value=[0, 0, 0])
implt(img)
# In[12]:
gaps = segmentation(img, step=2, RNN=True, debug=False)
# In[13]: