コード例 #1
0
ファイル: ocr_hocr.py プロジェクト: rytse/pottan-ocr
def ocr_images( images ):
    model = models.load_model( opt.crnn )
    maxWidth = max([i.shape[1] for i in images ])
    images = [ np.pad( i, [(2, 1), (0, maxWidth - i.shape[1] ), (0,0)], mode='constant', constant_values=1) for i in images ]
    out =  model.predict( np.array(images) )
    out = out.argmax(2)
    textResults = [ decodeStr( i, raw=False ) for i in out ]
    return textResults
コード例 #2
0
def ocr_images(model, images):
    #  import pdb; pdb.set_trace();
    #  from IPython import embed; embed()
    out = []
    for img in images:
        res = model.predict(np.array([img]))
        res = res.argmax(2)
        out.append(decodeStr(res[0], raw=False))
    return out
コード例 #3
0
ファイル: data_gen.py プロジェクト: ucodai/pottan-ocr
 def testEncoding(self, opt):
     """
     function for self testing.
     Encode each wrod, then decode it back
     """
     goodWords = []
     words = getTrainingTexts(self.WORD_LIST_FILE)
     for w in words:
         try:
             enc, encSize = encodeStr(w)
             goodWords.append(w)
         except Exception as e:
             print('Error encoding "%s"' % w, e)
             continue
         dec = decodeStr(enc)
         if (w != dec):
             raise ValueError('Encoding failed "%s" != "%s"' % (w, dec))
     if (opt.update):
         writeFile(self.WORD_LIST_FILE, '\n'.join(goodWords))