def predict_images(model):
img_path = sys.argv[1]
#img_path = input('image path: ')
# img = openImage(img_path)
# ret,img = cv2.threshold(img,130,255,cv2.THRESH_BINARY_INV)
img = imp.load_image(img_path)
img = ndimage.median_filter(img, 3)
# imp.show_image(img)
list_img = s._main(img)
# imp.show_image_array(list_img)
for i in range(len(list_img)):
h, w = list_img[i].shape
if h + w <= 5:
continue
img = list_img[i]
imp.show_image_array(img)
img = resize(img, 28, 28)
img = np.array(img)
#print(img)
img = img / 255
#print(img)
img = np.reshape(img, (1, 28, 28, 1))
result = model.predict_classes(img)
#print('res: ',result)
#print(labels)
kio = le.inverse_transform(result)
print(kio[0])
with open('dataaa.csv') as csvfile:
readCSV = csv.reader(csvfile, delimiter=',')
filee = open("file.txt", "a")
for row in readCSV:
if row[0] == kio[0]:
nee = row[1]
filee.write(nee)
print('UNICODE Value:', nee)
break
filee.close()
def main():
sys.setrecursionlimit(9999)
from sklearn.neighbors import KNeighborsClassifier
#img_name = input('Input the image name to predict: ')
img_name = sys.argv[1]
img = openImage(img_name)
#showImage(img)
#gaussian filtering
#img = cv2.blur(img,(5,5))
ret, img = cv2.threshold(img, 130, 255, cv2.THRESH_BINARY)
#showImage(img)
img = ndimage.median_filter(img, 3)
#print('background noise removed...')
#showImage(img)
trainingSet = []
testSet = []
split = 0.99
loadDataset('dataset-1.csv', split, trainingSet, testSet)
#print('Train set: '+repr(len(trainingSet)))
#print('Test set: '+repr(len(testSet)))
predictions = []
k = 3
#print('test: ',testSet)
#.....preprocessing....
#....line_segmentation.......
lines = ls.segment(img)
for u in range(len(lines)):
#showImage(lines[u])
#.....character segmentation.....
list_img = s._main(lines[u])
for i in range(len(list_img)):
h, w = list_img[i].shape
if h + w <= 5:
continue
list_img[i] = resize(list_img[i], 28, 28)
#showImage(list_img[i])
try:
ret, list_img[i] = cv2.threshold(list_img[i], 130, 255,
cv2.THRESH_BINARY)
except:
print('')
img = np.array(list_img[i])
#showImage(img)
array = np.array(img)
array = np.reshape(array, (1, np.product(array.shape)))
#showImage(list_img[i])
#trainingSet = np.array(trainingSet)
#print(trainingSet,'\n')
knn = KNeighborsClassifier()
trainingSet = np.array(trainingSet)
length = len(trainingSet)
#print('len: ',length)
X = trainingSet[0:length, 0:784]
Y = trainingSet[0:length, 784]
#print('array: ',array)
X = [[float(x) for x in rec] for rec in X]
X = np.array(X)
#print('array:',array)
knn.fit(X, Y)
#print('shape of X: ',X.shape)
#print('shape of arr:',array.shape)
print('predicted: ', knn.predict(array))
g = knn.predict(array)
print('Value:', g[0])
with open('dataaa.csv') as csvfile:
readCSV = csv.reader(csvfile, delimiter=',')
filee = open("file.txt", "a")
for row in readCSV:
if row[0] == g[0]:
nee = row[1]
filee.write(nee)
print('UNICODE Value:', nee)
filee.close()
webbrowser.open('http://127.0.0.1/FINALS/final.html')
def main():
sys.setrecursionlimit(99999)
img_name = input('Input the image name containing training characters: ')
img = openImage(img_name)
showImage(img)
#.....preprocessing....
img = ndimage.median_filter(img, 3)
#img = cv2.GaussianBlur(img,(5,5),0)
#img = ske.skeleton(img)
#print('after skeletonizing..')
showImage(img)
#.....segmentation.....
list_img = s._main(img)
label_name = input('Input the label of the segmented characters: ')
csv_file_name1 = input('Dataset(without skeletonizing)PATH: ')
csv_file_name2 = input('Dataset(with skeletonizing)PATH: ')
#csv_file_name3 = input('Dataset(with zoning)PATH: ')
#.....writing to csv.....
csv_file_1 = open(csv_file_name1, 'a', newline = '')
csv_file_2 = open(csv_file_name2, 'a', newline = '')
#csv_file_3 = open(csv_file_name3, 'a', newline = '')
with csv_file_1:
with csv_file_2:
writer1 = csv.writer(csv_file_1)
writer2 = csv.writer(csv_file_2)
for i in range(len(list_img)):
try:
w,h = list_img[i].shape
except:
continue
if w+h <= 5:#to avoid very small segments
continue
#showImage(list_img[i])
if True:
try:
list_img[i] = resize(list_img[i],28,28)
except:
continue
#print('after resizing: ')
#showImage(list_img[i])
ret,skeleton = cv2.threshold(list_img[i],40,255,cv2.THRESH_BINARY_INV)
#print('binarization: ')
#showImage(list_img[i])
skeleton = ske.skeleton(skeleton)
print('after skeletonizing: ')
showImage(skeleton)
t = input('Want to store the feature values of this image? (Y / N)')
if t not in ['Y','y']:
continue
#label_name = input('Input the label of the segmented characters: ')
array1 = np.array(list_img[i])
array1 = np.append(array1, [[label_name]])
array1 = np.reshape(array1, (1,np.product(array1.shape)))
array2 = np.array(skeleton)
array2 = np.append(array2, [[label_name]])
array2 = np.reshape(array2, (1,np.product(array2.shape)))
writer1.writerows(array1)
writer2.writerows(array2)
print(((i+1)/len(list_img))*100, '% done...')