Ä ÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÿÇ ÒpÅ üzÅ fM× ˜M× ¤òÙ ¨Ú tessedit_load_sublangs jpn_vert # Important configurations for CJK mode # New Segmentation search params language_model_ngram_on 1 segsearch_max_char_wh_ratio 1.3 language_model_ngram_space_delimited_language F language_model_ngram_scale_factor 0.1 language_model_use_sigmoidal_certainty T language_model_ngram_nonmatch_score -20 # always force word_associator to run, whereas "enable_assoc" is conditioned on # permuter rejection, which may not be what we want until we finalize the # dictionary issue for CJK #force_word_assoc T # Chinese symbols are more complex than Latin. Adjust the blob filtering # thresholds so they do not get filtered out as noise or containers. # Also force word segmentation to reduce the length of blob sequences. edges_use_new_outline_complexity T edges_children_fix T edges_max_children_per_outline 10 edges_max_children_layers 5 edges_children_count_limit 1000 tosp_force_wordbreak_on_punct F textord_force_make_prop_words T textord_noise_sizelimit 0.1 textord_noise_normratio 6 textord_max_noise_size 7 textord_noise_rejwords F textord_no_rejects T textord_tabfind_vertical_text_ratio 0.1 textord_tabfind_aligned_gap_fraction 0.5 textord_tabvector_vertical_box_ratio 0.3 textord_tabvector_vertical_gap_fraction 1.5 # Make use of the fact that Chinese is monospaced, and incorporate that # into character segmentation cost for search, and inject this cost to # word rating. classify_integer_matcher_multiplier 10 #classify_class_pruner_multiplier 15 assume_fixed_pitch_char_segment T chop_enable F tessedit_char_blacklist ° segment_nonalphabetic_script 1 allow_blob_division F # Since we do not have a dictionary for Chinese yet, none of the results # will be found in dawg. We need to relax the stopping condition and turn # off dictionary based penalties. Instead, we get additional constraints # from script consistency. stopper_nondict_certainty_base -2 segment_penalty_dict_nonword 1.0 segment_penalty_garbage 1.0 # Force use of a single x-height mode when the text is written horizontally. # This information could come from the unicharset (script_has_xheight), but # it is better in the config file so as to be available when training. textord_single_height_mode T # Use character height as x-height, and estimate it from character pitch and # kerning width. textord_use_cjk_fp_model T Series À 0 Ÿ ?¬ Series Input À 0 Input 0 Series À ConvSeries Convolve À Convolve Tanh À ConvNL„ ŒöŸ‹=ˆÀ?XG±`ÜÊ·?ÇÄ,i{¤?䌼ðjåÆ?’¸úª°Á?f]ZÞ<ƒÀ?Ÿùf÷e Ï¿ÑÊ�]À¿c¿w"'°¿L<á�e„º¿”B.Š?5Y²éd®Ç¿ø÷ÇÂG²¿Ã"ð1<§Ä?½%` *¿¿’ÙweÖ Õ¿sê%.Ö?B‘ þÆ?íËøˆÏ¡?ÞèÏ,"׿þÍ™x,9?W‘õžÓ¿5HQàây?~ &5Ó¿Gh˜›;Ô¿½¦Ó+«±Õ?UðΑ6×»¿¦ÞˆËú«Ð?Í’;ÎØ?øÈñŸXØ¿šŽß¿ÀÑ?°ê¡W ˜¿ä”8@NÈ¿ÿîÎ6$â?Ôá†H‚»µ¿4 äƒýûà¿—Z‚ä\·Õ?Ùò�ŠÈ7§¿2¡¢‡·Þο2mÁYŠÈ¿Ç˜‡{Dlο�éG›Öµ¿8D)H¹?ªT’•VÖ¶?Ïø¥÷ Ð?€$³ÑJUÈ?‰õàß*Æ?=�q3dgª?±Å+4”›iôx‘É¿G OYɦ?À úéÚ´?üÐuPÂÆž?³”üͦÓ?5þ
ö7~‰w?Ü 9QѨŠ?Xt¾X?y×:Z)V´?‰Ä–·ºp½?ËžÛëÚ.Ò?ûóÇsEØž¿ôîGÂ0·?ø�¤Îà?^©—¿mÖ¿íÞi 󋽿€y�Ê…É?3ÕnLÅjÁ¿„Í&Ø¿üFÃv9t¡?-˜õXë׿´Ø5»4ÁÄ?_°3-›Þ?Ðÿ§ãÎß?ª×/Gü¿¤Ž*áÀ¿e¯9T9Õ¿nõ â€Ã¿¹O�ú¿Ð¿ÖÏ øÔ¿Ã|@!›Ð¿G¤ä3X®à?æé³'©Â?Ke8M´?A.Lá\¸Û¿Ç$°WÍ'ä¿¡.,LpäÔ¿õ5?mÍ6º¿òR÷Kà?›Sá`&Ì?n¤½h]¥É¿¿qÿäÂ?ëA±A—ö½?§•Z¥?ev€YÇ-Ê¿b€)v�®?ã6ퟴº?$w& ¬îÆ¿5FbhÜÃ?ôÛ²†õ~Ç?àÆ�„ÁÏ¿˜¥WË_ŒÕ¿Ø£-!‰Ñ¿uZV£Ì†Ì¿pjÎ:Íòǿԟ2‡Å¿€Ñ Ûíã ¿»ŒÞVa‡Þ?´>cÝ?¿Ÿå^a=Ð?ÜGÄ`žÚÿàºö)¸æ¹?ªÍQÍC°¿[•2&$à¿“±¥ä@íÂ?Æw—VHã?Z®—²¹…Ä?†3Õ}=%Þ¿ËÕ+áÙ˹?$hbî"?Ïîùî�¥Ç¿uy9&FÍ¿E„USv±¿úàš†Ø?,ï<8—@Ù¿¤×¶µp‚Ì¿Œëئãá?ƒm¤0”Ò¿þY>Á¹¿·À¦Œj×?¬×Ìj`Ö»¿¦£ŒÏ¦”�?š¹'H¿ÿä¹ó»¿Œ7‚eZð„?•*-þRIp¿¢1¿À›¿HǘÁÿ…¾`0›Â¿â÷IBh ¿È˜6½¿z9\4mÜ¿Iâ‘j½ä?[8 ¸ˆ ¿ÆÎ?aBÅ¿¤ˆ¢MÝ?^¸-¡bß¿S zWæ¶¿ãg{ç÷×?ìi3m Æ¿2ÑF�Á¿ oƒ:o9 MaxpoolÀ Maxpool XYTransposeÀ @ Q XYTransLSTM SummLSTMÀ @ Q Lfys64P „@ Q •}ˆÝ á¿õ\f:ÜÚ¿Wk<™¡ðå?N•ü6¿gdm´á?Ñr½ñ”*ª?²ƒÀmõ�å¿ mú.#åÞ¿.6c6œÆ¿eï>B*ŽÌ¿9©ê=Ä?Hìå½L<×?¹kLUÔ"Þ¿eF©C¨•Æ?™|Lj¾,Ð?$}âôÐ0q?6ÜŠ•kuî¿ÜB $†Ô¿�ÿ¬õ¦?B\ú} Ö?¾*P»Õ„Ð? o1ûØ'Æ?†äLnØ¿XÇU~¸Ñ?†ô¬¦×¿·ÉÝ zï¿Ô‰Ëè¿ÌdGÝaå¿ì''™î±? h{ªå æ?hñqÀ‘î¿’öˆœt·í?²³^àjä¬?�®Koý¥¿'yÚU/�¿=mwœ.ì¿™ ¼å¿"†ˆîÃØ?6DDZó«?"⦠ϼÑ?g:`V�Ð?|NáBqÁ?½íÈU{\Ù¿¹Íì©Ø¸¿œª}èÜ3æ?8mmbžØÑ¿‹ÈHx�çÊ¿âD–ë0O¾¿>ß›øãµ¿½.I,¨æ¿^x�›UX©?Šœ(AÓ?”ÒkXðxÔ?eH�ß,¨Ë?ÃÝÕCyTÕ¿`ŠSõFÌ¿¸jÑÉ?R©'YÏzÛ?ÞAX„a¾¿‚š<âq;¼?qyZg&Ö¿[Qh[Ú¹ä?•ìF;¯ÃÏ?ÚdI^ÉXÇ?é <�FÞä?+Eu'ÚÑ¿qÙSÁ!ð¿* *‡ƒÚï¿~xÚ–Ê¿ýRŸm½¿'Fþ뿦 &˧ŸÅ¿ XÛçâ?"%?ûþÒ?ô›¿{ä«?¦fg4vî?ô•‰Øâ¿IÜyìV�пA….øÒ¸Ë¿”ì� ò<Ü?ˆž)½ŽÒ¿&ý