def check_keys(my_data):
# test for various keyboard inputs
(event, background, draw_color, line_width, keep_going, screen, mat) = my_data
if event.key == pygame.K_q:
keep_going = False
elif event.key == pygame.K_c:
background.fill((255, 255, 255))
draw_pixelated(np.zeros((28, 28)), screen)
# s - for saving a digit and pattern
elif event.key == pygame.K_s:
answer = int(ask(screen, ""))
f_m = matrix_manipulate.focus_mat(mat)
shrinking_model.learn_pattern(f_m, answer, "one_zero_shrinking")
extended_model.learn_pattern(f_m, answer, "one_zero_extended")
multiplication_model.learn_pattern(f_m, answer, "one_zero_multiplication")
"""shrinking_model.learn_pattern(f_m, answer, "shrinking_model")
extended_model.learn_pattern(f_m, answer, "extended_model")
multiplication_model.learn_pattern(f_m, answer, "multiplication_model")"""
# t - for testing some real time hand write digit
elif event.key == pygame.K_t:
model_tester.test_model(my_data, 'one_zero_shrinking')
model_tester.test_model(my_data, 'one_zero_extended')
model_tester.test_model(my_data, 'one_zero_multiplication')
"""model_tester.test_model(my_data, 'multiplication_model')
model_tester.test_model(my_data, "shrinking_model")
model_tester.test_model(my_data, "extended_model")"""
background.fill((255, 255, 255))
draw_pixelated(np.zeros((28, 28)), screen)
my_data = (event, background, draw_color, line_width, keep_going)
return my_data
def main():
file_name = {
'images':
'C:\\Users\\asafm\\PycharmProjects\\Final-Project\\samples\\train-images.idx3-ubyte',
'labels':
'C:\\Users\\asafm\\PycharmProjects\\Final-Project\\samples\\train-labels.idx1-ubyte'
}
labels_array = np.array([])
data_types = {
0x08: ('ubyte', 'B', 1),
0x09: ('byte', 'b', 1),
0x0B: ('>i2', 'h', 2),
0x0C: ('>i4', 'i', 4),
0x0D: ('>f4', 'f', 4),
0x0E: ('>f8', 'd', 8)
}
for name in file_name.keys():
if name == 'images':
images_file = open(file_name[name], 'rb')
if name == 'labels':
labels_file = open(file_name[name], 'rb')
images_file.seek(0)
magic = st.unpack('>4B', images_file.read(4))
if (magic[0] and magic[1]) or (magic[2] not in data_types):
raise ValueError("File Format not correct")
nDim = magic[3]
print("Data is ", nDim, "-D")
# offset = 0004 for number of images
# offset = 0008 for number of rows
# offset = 0012 for number of columns
# 32-bit integer (32 bits = 4 bytes)
images_file.seek(4)
nImg = st.unpack('>I', images_file.read(4))[0] # num of images/labels
nR = st.unpack('>I', images_file.read(4))[0] # num of rows
nC = st.unpack('>I', images_file.read(4))[0] # num of columns
nBytes = nImg * nR * nC
labels_file.seek(
8) # Since no. of items = no. of images and is already read
print("no. of images :: ", nImg)
print("no. of rows :: ", nR)
print("no. of columns :: ", nC)
# Read all data bytes at once and then reshape
images_array = 255 - np.asarray(
st.unpack('>' + 'B' * nBytes, images_file.read(nBytes))).reshape(
(nImg, nR, nC))
labels_array = np.asarray(
st.unpack('>' + 'B' * nImg, labels_file.read(nImg))).reshape((nImg, 1))
for i in tqdm(range(60000)):
lable = labels_array[i][0]
mat = range(784)
mat = np.reshape(mat, (28, 28))
for j in range(28):
for h in range(28):
if images_array[i][j][h] == 255:
mat[j][h] = 0
else:
mat[j][h] = 1
m = matrix_manipulate.focus_mat(mat)
#sum_model.learn_pattern(m, lable, "sum_model")
shrinking_model.learn_pattern(m, lable, "shrinking_model")
#extended_model.learn_pattern(m, lable, "extended_model")
#multiplication_model.learn_pattern(m, lable, "multiplication_model")
print('Finish to load MNIST dataset')
def main():
for arg in sys.argv[1:]:
file_name = {
'images':
'C:\\Users\\asafm\\PycharmProjects\\Final-Project\\samples\\t10k-images.idx3-ubyte',
'labels':
'C:\\Users\\asafm\\PycharmProjects\\Final-Project\\samples\\t10k-labels.idx1-ubyte'
}
labels_array = np.array([])
data_types = {
0x08: ('ubyte', 'B', 1),
0x09: ('byte', 'b', 1),
0x0B: ('>i2', 'h', 2),
0x0C: ('>i4', 'i', 4),
0x0D: ('>f4', 'f', 4),
0x0E: ('>f8', 'd', 8)
}
for name in file_name.keys():
if name == 'images':
images_file = open(file_name[name], 'rb')
if name == 'labels':
labels_file = open(file_name[name], 'rb')
images_file.seek(0)
magic = st.unpack('>4B', images_file.read(4))
if (magic[0] and magic[1]) or (magic[2] not in data_types):
raise ValueError("File Format not correct")
nDim = magic[3]
print("Data is ", nDim, "-D")
# offset = 0004 for number of images
# offset = 0008 for number of rows
# offset = 0012 for number of columns
# 32-bit integer (32 bits = 4 bytes)
images_file.seek(4)
nImg = st.unpack('>I', images_file.read(4))[0] # num of images/labels
nR = st.unpack('>I', images_file.read(4))[0] # num of rows
nC = st.unpack('>I', images_file.read(4))[0] # num of columns
nBytes = nImg * nR * nC
labels_file.seek(
8) # Since no. of items = no. of images and is already read
print("no. of images :: ", nImg)
print("no. of rows :: ", nR)
print("no. of columns :: ", nC)
# Read all data bytes at once and then reshape
images_array = 255 - np.asarray(
st.unpack('>' + 'B' * nBytes, images_file.read(nBytes))).reshape(
(nImg, nR, nC))
labels_array = np.asarray(
st.unpack('>' + 'B' * nImg, labels_file.read(nImg))).reshape(
(nImg, 1))
sum = 0
num_of_samples = 0
for i in tqdm(range(10000)):
lable = labels_array[i][0]
if lable == 1 or lable == 0:
num_of_samples = num_of_samples + 1
mat = range(784)
mat = np.reshape(mat, (28, 28))
for j in range(28):
for h in range(28):
if images_array[i][j][h] == 255:
mat[j][h] = 0
else:
mat[j][h] = 1
m = matrix_manipulate.focus_mat(mat)
result = test_machine(m, arg)
if result == lable:
sum += 1
print(
'===============================================================')
print(
'Finish to Test the machine with MNIST dataset on {}'.format(arg))
print('Num of samples = ', num_of_samples)
print('num of correct answers = ', sum)
print('Accuracy of the machine is: ', (sum / num_of_samples) * 100,
'%')
print(
'===============================================================')
def test_model(my_data, model_name):
(event, background, draw_color, line_width, keep_going, screen,
mat) = my_data
f_m = focus_mat(mat)
if 'shrinking' in model_name:
pattern = shrinking_model.get_pattern(f_m)
elif 'extended' in model_name:
pattern = extended_model.get_pattern(f_m)
elif 'multiplication' in model_name:
pattern = multiplication_model.get_pattern(f_m)
elif 'sum' in model_name:
pattern = sum_model.get_pattern(f_m)
try:
cnx = connector.connect(user='******',
password='******',
database='hand_write_recognition')
cursor = cnx.cursor()
query = ("SELECT digit FROM {} WHERE pattern = {}".format(
model_name, pattern))
cursor.execute(query)
digits = []
for digit in cursor:
digits.append(digit)
guess1mount = guess2mount = 0
try:
print('*** TEST FOR {} ***'.format(model_name))
guess1is = Counter(digits).most_common(2).pop(0)[0][0]
guess1mount = Counter(digits).most_common(2).pop(0)[1]
print('Guess 1 is :', guess1is)
except IndexError:
print('I could not guess a single number as a first guess.')
try:
guess2is = Counter(digits).most_common(2).pop(1)[0][0]
guess2mount = Counter(digits).most_common(2).pop(1)[1]
print('Guess 2 is :', guess2is)
except IndexError:
print('I could not guess a single number as a second guess.')
sum = guess1mount + guess2mount
try:
print('Accuracy of :', guess1is, ': ', (guess1mount / sum) * 100,
'%')
print('Accuracy of :', guess2is, ': ', (guess2mount / sum) * 100,
'%')
except UnboundLocalError:
print('I could not calculate the accuracy')
cursor.close()
except connector.Error as err:
if err.errno == errorcode.ER_ACCESS_DENIED_ERROR:
print("Something is wrong with your user name or password")
elif err.errno == errorcode.ER_BAD_DB_ERROR:
print("Database does not exist")
else:
print(err)
else:
cnx.close()