def train(self, charcol, meta, path=None):
self._check_meta(meta)
trainer = zinnia.Trainer()
zinnia_char = zinnia.Character()
for set_name in charcol.get_set_list():
for character in charcol.get_characters(set_name):
if (not zinnia_char.parse(character.to_sexp())):
raise TrainerError, zinnia_char.what()
else:
trainer.add(zinnia_char)
if not path:
if "path" in meta:
path = meta["path"]
else:
path = os.path.join(os.environ['HOME'], ".tegaki",
"models", "zinnia",
meta["name"] + ".model")
else:
path = os.path.abspath(path)
if not os.path.exists(os.path.dirname(path)):
os.makedirs(os.path.dirname(path))
meta_file = path.replace(".model", ".meta")
if not meta_file.endswith(".meta"): meta_file += ".meta"
trainer.train(path)
self._write_meta_file(meta, meta_file)
def recognize(self, recognizer, writing):
character = Character()
character.set_utf8("?")
character.set_writing(writing)
sexp = self.char_to_sexp(character)
zinnia_char = zinnia.Character()
if (not zinnia_char.parse(sexp)):
print zinnia_char.what()
exit(1)
results = recognizer.classify(zinnia_char, 10)
return [results.value(i) for i in range(0, (results.size() - 1))]
def _recognize(self, writing, n=10):
s = zinnia.Character()
s.set_width(writing.get_width())
s.set_height(writing.get_height())
strokes = writing.get_strokes()
for i in range(len(strokes)):
stroke = strokes[i]
for x, y in stroke:
s.add(i, x, y)
result = self._recognizer.classify(s, n + 1)
size = result.size()
return Results([(result.value(i), result.score(i)) \
for i in range(0, (size - 1))])
def train(self):
trainer = zinnia.Trainer()
zinnia_char = zinnia.Character()
for char_code, xml_list in self.train_xml_files_dict.items():
for xml_file in xml_list:
character = self.get_character(xml_file)
sexp = self.char_to_sexp(character)
if (not zinnia_char.parse(sexp)):
print zinnia_char.what()
exit(1)
else:
trainer.add(zinnia_char)
path = os.path.join(self.ROOT, "model")
trainer.train(path)
#!/usr/bin/python
import zinnia
input = "(character (width 1000)(height 1000)(strokes ((243 273)(393 450))((700 253)(343 486)(280 716)(393 866)(710 880))))"
try:
s = zinnia.Character()
r = zinnia.Recognizer()
r.open("handwriting-zh_CN.model")
if (not s.parse(input)):
print s.what()
result = r.classify(s, 10)
size = result.size()
for i in range(0, (size - 1)):
print result.value(i).decode('utf8'), '\t-\t', result.score(i)
s.clear()
s.set_width(300)
s.set_height(300)
s.add(0, 51, 29)
s.add(0, 117, 41)
s.add(1, 99, 65)
s.add(1, 219, 77)
s.add(2, 27, 131)
s.add(2, 261, 131)
s.add(3, 129, 17)
s.add(3, 57, 203)
s.add(4, 111, 71)
s.add(4, 219, 173)