def test_se_tardata_sudokus(file_path_base):
try:
image_file_name = file_path_base + '.jpg'
image = load_image(image_file_name)
except:
image_file_name = file_path_base + '.png'
image = load_image(image_file_name)
with open(file_path_base + '.txt', 'rt') as f:
correct_parsing = f.read().strip()
predictions, sudoku, subimage = extract_sudoku(image,
classifier(),
force=True)
parsed_sudoku = predictions_to_suduko_string(predictions, oneliner=False)
if parsed_sudoku != correct_parsing:
predictions, sudoku, subimage = extract_sudoku(image.rotate(-90),
classifier(),
force=True)
parsed_sudoku = predictions_to_suduko_string(predictions,
oneliner=False)
assert parsed_sudoku == correct_parsing
try:
os.remove(image_file_name)
os.remove(file_path_base + '.txt')
print("Deleted temporary test dir.")
os.removedirs(os.path.dirname(file_path_base))
except:
pass
def test_se_tardata_sudokus(file_path_base):
try:
image_file_name = file_path_base + '.jpg'
image = load_image(image_file_name)
except:
image_file_name = file_path_base + '.png'
image = load_image(image_file_name)
with open(file_path_base + '.txt', 'rt') as f:
correct_parsing = f.read().strip()
predictions, sudoku, subimage = extract_sudoku(image, classifier(), force=True)
parsed_sudoku = predictions_to_suduko_string(predictions, oneliner=False)
if parsed_sudoku != correct_parsing:
predictions, sudoku, subimage = extract_sudoku(image.rotate(-90), classifier(), force=True)
parsed_sudoku = predictions_to_suduko_string(predictions, oneliner=False)
assert parsed_sudoku == correct_parsing
try:
os.remove(image_file_name)
os.remove(file_path_base + '.txt')
print("Deleted temporary test dir.")
os.removedirs(os.path.dirname(file_path_base))
except:
pass
def test_xanadoku_sudokus(sudoku_doc):
if not sudoku_doc.get('verified'):
assert True
else:
print(sudoku_doc.get('_id'))
image = download_image(sudoku_doc.get('raw_image_url'))
predictions, sudoku, subimage = extract_sudoku(image, classifier(), force=True)
parsed_sudoku = predictions_to_suduko_string(predictions, oneliner=True)
if parsed_sudoku != sudoku_doc.get('parsed_sudoku'):
predictions, sudoku, subimage = extract_sudoku(image.rotate(-90), classifier(), force=True)
parsed_sudoku = predictions_to_suduko_string(predictions, oneliner=True)
assert parsed_sudoku == sudoku_doc.get('parsed_sudoku')
def test_image_1_with_trained_classifier():
image = _get_img(1)
c = fit.KNeighborsClassifier(n_neighbors=5)
c = fit.fit_sudokuextract_classifier(c)
predictions, sudoku, subimage = extract_sudoku(image, c)
parsed_sudoku = predictions_to_suduko_string(predictions)
assert parsed_sudoku == _get_parsed_img(1)
def test_image_2_with_trained_classifier():
image = _get_img(2)
c = fit.KNeighborsClassifier(n_neighbors=10)
c = fit.fit_combined_classifier(c)
predictions, sudoku, subimage = extract_sudoku(image, c)
parsed_sudoku = predictions_to_suduko_string(predictions)
assert parsed_sudoku == _get_parsed_img(2)
def extraction_method_map(image,
classifier,
use_local_thresholding=False,
apply_gaussian=False,
n=5,
force=False):
for sudoku, subimage in _extraction_iterator_map(image,
use_local_thresholding,
apply_gaussian,
n=n):
try:
pred_n_imgs = classify_sudoku(sudoku, classifier)
preds = np.array([[pred_n_imgs[k][kk][0] for kk in range(9)]
for k in range(9)])
imgs = [[pred_n_imgs[k][kk][1] for kk in range(9)]
for k in range(9)]
if np.sum(preds > 0) >= 17 or force:
try:
s = Sudoku(predictions_to_suduko_string(preds))
s.solve()
except SudokuException:
if force:
return preds, imgs, subimage
except Exception:
pass
else:
return preds, imgs, subimage
except Exception as e:
pass
raise SudokuExtractError(
"Corner Method could not extract any Sudoku from this image.")
def test_xanadoku_sudokus(sudoku_doc):
if not sudoku_doc.get('verified'):
assert True
else:
print(sudoku_doc.get('_id'))
image = download_image(sudoku_doc.get('raw_image_url'))
predictions, sudoku, subimage = extract_sudoku(image,
classifier(),
force=True)
parsed_sudoku = predictions_to_suduko_string(predictions,
oneliner=True)
if parsed_sudoku != sudoku_doc.get('parsed_sudoku'):
predictions, sudoku, subimage = extract_sudoku(image.rotate(-90),
classifier(),
force=True)
parsed_sudoku = predictions_to_suduko_string(predictions,
oneliner=True)
assert parsed_sudoku == sudoku_doc.get('parsed_sudoku')
def main():
import argparse
parser = argparse.ArgumentParser(description="Running SudokuExtract as a command line tool")
group = parser.add_mutually_exclusive_group()
group.add_argument('-p', '--path', dest='path', type=str, default=None, help="Path to an image to parse.")
group.add_argument('-u', '--url', dest='url', type=str, default=None, help="Url to an image to parse.")
parser.add_argument('--oneliner', action='store_true', help="Print oneliner solution.")
args = parser.parse_args()
if args.path is not None:
image = load_image(args.path)
else:
image = download_image(args.url)
preds, images, subimage = extract_sudoku(image)
sudoku_string = predictions_to_suduko_string(preds, args.oneliner)
print(sudoku_string)
def extraction_method_map(image, classifier, use_local_thresholding=False, apply_gaussian=False, n=5, force=False):
for sudoku, subimage in _extraction_iterator_map(image, use_local_thresholding, apply_gaussian, n=n):
try:
pred_n_imgs = classify_sudoku(sudoku, classifier)
preds = np.array([[pred_n_imgs[k][kk][0] for kk in range(9)] for k in range(9)])
imgs = [[pred_n_imgs[k][kk][1] for kk in range(9)] for k in range(9)]
if np.sum(preds > 0) >= 17 or force:
try:
s = Sudoku(predictions_to_suduko_string(preds))
s.solve()
except SudokuException:
if force:
return preds, imgs, subimage
except Exception:
pass
else:
return preds, imgs, subimage
except Exception as e:
pass
raise SudokuExtractError("Corner Method could not extract any Sudoku from this image.")
def _process_an_image(image, correct_sudoku):
predictions, sudoku, subimage = extract_sudoku(image, classifier())
parsed_sudoku = predictions_to_suduko_string(predictions)
assert parsed_sudoku == correct_sudoku
def _process_an_image(image, correct_sudoku):
predictions, sudoku, subimage = extract_sudoku(image, classifier())
parsed_sudoku = predictions_to_suduko_string(predictions)
assert parsed_sudoku == correct_sudoku
