def generate_training_result():
""" It should be very careful to use this method,
since it will replace all the result files.
"""
arg_exp = 20
dataset_matrix_hash = data_file_helper.get_dataset_matrix_hash(
data_file_helper.FONT_TRAINING_PATH, range(1, 10)
)
mb = MultipleSvm.train_and_save_variables(
Smo, dataset_matrix_hash, 200, 0.0001, 1000, arg_exp, data_file_helper.SUPPLEMENT_RESULT_PATH
)
if __name__ == "__main__":
from minitest import *
import data_file_helper
from picture_sudoku.helpers import numpy_helper
from picture_sudoku.cv2_helpers.display import Display
from picture_sudoku.cv2_helpers.image import Image
from picture_sudoku.helpers.common import Resource
from picture_sudoku import main_sudoku
def show_number_matrix(number_matrix):
binary_number_image = number_matrix.reshape((data_file_helper.IMG_SIZE, data_file_helper.IMG_SIZE))
number_image = numpy_helper.transfer_values_quickly(binary_number_image, {1: 255})
number_image = numpy.array(number_image, dtype=numpy.uint8)
# Image.save_to_txt(number_image,'test1.dataset')
Display.image(number_image)
""" how to use training, please see the generator.py"""
with test(MultipleSvm.dag_classify):
dataset_matrix_hash = data_file_helper.get_dataset_matrix_hash(data_file_helper.FONT_TRAINING_PATH, (9,))
mb = MultipleSvm.load_variables(Smo, data_file_helper.FONT_RESULT_PATH)
mb.dag_classify(dataset_matrix_hash[9][0]).must_equal(9)
with test("special number"):
file_name = "sample_18_01_26_resized.dataset"
issue_ragion = Image.load_from_txt(Resource.get_path("test", file_name))
transfered_ragion = main_sudoku.transfer_to_digit_matrix(issue_ragion)
mb = MultipleSvm.load_variables(Smo, data_file_helper.FONT_RESULT_PATH)
mb.dag_classify(transfered_ragion).must_equal(1)
# Display.binary_image(main_sudoku.resize_to_cell_size(issue_ragion))
