def test_text_bold(self):
finder = TextFinder()
matches = finder.find(Text('Find the word'), Image('sentence_bold'))
self.assertEqual(len(matches), 1)
self.assertEqual(matches[0].x, 12)
self.assertEqual(matches[0].y, 13)
self.assertAlmostEqual(matches[0].width, 100, delta=5)
self.assertAlmostEqual(matches[0].height, 10, delta=5)
def test_text_larger(self):
finder = TextFinder()
matches = finder.find(Text('Find the word'), Image('sentence_larger'))
self.assertEqual(len(matches), 1)
# TODO: location too far due to poor text detection
#self.assertEqual(matches[0].x, 13)
self.assertEqual(matches[0].y, 13)
#self.assertAlmostEqual(matches[0].width, 100, delta=5)
self.assertAlmostEqual(matches[0].height, 10, delta=5)
def test_text_basic(self):
finder = TextFinder()
matches = finder.find(Text('Find the word here'),
Image('sentence_sans'))
self.assertEqual(len(matches), 1)
# TODO: location too far due to poor text detection
#self.assertEqual(matches[0].x, 11)
self.assertEqual(matches[0].y, 12)
self.assertAlmostEqual(matches[0].width, 115, delta=5)
self.assertAlmostEqual(matches[0].height, 10, delta=5)
def test_text_nomatch(self):
finder = TextFinder()
finder.params["find"]["similarity"].value = 0.25
i = 1
for tdetect in finder.algorithms["text_detectors"]:
# TODO: this is still not implemented
if tdetect == "components":
continue
for ocr in finder.algorithms["text_recognizers"]:
# TODO: this is still not implemented
if ocr == "beamSearch":
continue
# TODO: handle newer OpenCV bugs with some backends
import cv2
# TODO: deprecate OpenCV 3.X versions after time
if cv2.__version__.startswith("3.") and tdetect == "east":
continue
finder.configure_backend(tdetect, "tdetect")
finder.configure_backend(ocr, "ocr")
# also with customized synchronization to the configuration
finder.synchronize_backend(tdetect, "tdetect")
finder.synchronize_backend(ocr, "ocr")
matches = finder.find(Text('Nothing'), Image('all_shapes'))
# verify match accuracy
self.assertEqual(len(matches), 0)
# verify dumped files count and names
dumps = self._verify_and_get_dumps(7, i)
self._verify_dumped_images('Nothing', 'all_shapes', dumps,
"text")
hotmaps = sorted(self._get_matches_in('.*hotmap.*', dumps))
self.assertEqual(len(hotmaps), 4)
for j, hotmap in enumerate(hotmaps):
if j == 0:
self.assertIn('3hotmap', hotmap)
elif j == 1:
self.assertIn('3hotmap-1char', hotmap)
elif j == 2:
self.assertIn('3hotmap-2text', hotmap)
else:
self.assertIn('3hotmap-3ocr-%stext' % (j - 2), hotmap)
if j == 3 or j == 4:
# report achieved similarity in the end of the filename
self.assertRegex(hotmap, ".*-\d\.\d+.*")
self.assertTrue(
os.path.isfile(os.path.join(self.logpath, hotmap)))
shutil.rmtree(self.logpath)
i += 1
def test_invalid_backends(self):
"""Test that unsupported backends are detected when loading and saving."""
# test on load
stepsfile_contents = [
"item_for_contour.png some_contour_matchfile.match",
"some_text_content some_unknown_content.match"
]
self.assertRaises(UnsupportedBackendError, self._build_chain,
os.linesep.join(stepsfile_contents))
# test on save
finder = Finder(False, False)
finder.params["find"] = {"backend": "unknown"}
chain = self._build_chain("")
chain._steps.append(Text("", finder))
self.assertRaises(UnsupportedBackendError, chain.save, "foobar")
def test_text_same(self):
finder = TextFinder()
finder.params["find"]["similarity"].value = 1.0
i = 1
for tdetect in finder.algorithms["text_detectors"]:
# TODO: this is still not implemented
if tdetect == "components":
continue
for ocr in finder.algorithms["text_recognizers"]:
# TODO: this is still not implemented
if ocr == "beamSearch":
continue
# TODO: handle newer OpenCV bugs with some backends
import cv2
# TODO: OpenCV 4.2.0 Tesseract bindings output nothing
if cv2.__version__ == "4.2.0" and ocr == "tesseract":
continue
# TODO: deprecate OpenCV 3.X versions after time
elif cv2.__version__.startswith("3.") and tdetect == "east":
continue
# HMM misinterprets one char leading to 3/4 recognized chars
# Tesseract still has similarity 1.0 though
if ocr == "hmm":
finder.params["find"]["similarity"].value = 0.75
if tdetect == "east":
finder.params["find"]["similarity"].value = 0.4
else:
finder.params["find"]["similarity"].value = 1.0
finder.configure_backend(tdetect, "tdetect")
finder.configure_backend(ocr, "ocr")
# also with customized synchronization to the configuration
finder.synchronize_backend(tdetect, "tdetect")
finder.synchronize_backend(ocr, "ocr")
matches = finder.find(Text('Text'), Image('all_shapes'))
# verify match accuracy
self.assertEqual(len(matches), 1)
# the EAST network confuses the space among some squares with
# text and thus still read the output but in a larger rectangle
if tdetect != "east":
self.assertEqual(matches[0].x, 22)
self.assertEqual(matches[0].y, 83)
self.assertAlmostEqual(matches[0].width, 40, delta=3)
self.assertAlmostEqual(matches[0].height, 15, delta=3)
# verify dumped files count and names
dumps = self._verify_and_get_dumps(7, i)
self._verify_dumped_images('Text', 'all_shapes', dumps, "text")
hotmaps = sorted(self._get_matches_in('.*hotmap.*', dumps))
self.assertEqual(len(hotmaps), 4)
for j, hotmap in enumerate(hotmaps):
if j == 0:
self.assertIn('3hotmap', hotmap)
elif j == 1:
self.assertIn('3hotmap-1char', hotmap)
elif j == 2:
self.assertIn('3hotmap-2text', hotmap)
else:
self.assertIn('3hotmap-3ocr-%stext' % (j - 2), hotmap)
if j == 3 or j == 4:
# report achieved similarity in the end of the filename
self.assertRegex(hotmap, ".*-\d\.\d+.*")
self.assertTrue(
os.path.isfile(os.path.join(self.logpath, hotmap)))
shutil.rmtree(self.logpath)
i += 1