o eo @s ddlZddlZddlmZejejeZej eej dejej eddej d<ddl Z ddlZddlZddlZddlZddlZddlmmZddlmZddlmZddlmZmZeZGd d d eZ d d Z!e"d kre!e#dSdS)N)Imagez../..Z auto_growthZFLAGS_allocator_strategy)build_post_process) get_logger)get_image_file_listcheck_and_readc@speZdZddZddZddZddZd d Zd d Z dddZ ddZ ddZ ddZ ddZ ddZdS)TextRecognizercCs dd|jdD|_|j|_|j|_d|j|jd}|jdkr*d|j|jd}n|jdkr8d |j|jd}n|jd krFd |j|jd}n}|jd krTd |j|jd}no|jdkrbd|j|jd}na|jdkrpd|j|jd}nS|jdkr~d|j|jd}nE|jdkrd|j|jd}n7|jdkrd |j|jdd}n(|jdkrdd|jd}n|jdkrddi}n|jdkr|j|_d|j|jd}t||_ t |d t \|_ |_|_|_|j|_|j|_|jrd!dl}t}t }|jd |j|jd"d|j|d|jr|ndgd#d!t d$ |_dSdS)%NcSsg|]}t|qS)int).0vrrQD:\Projects\ConvertPro\env\Lib\site-packages\paddleocr/tools/infer/predict_rec.py (sz+TextRecognizer.__init__..,ZCTCLabelDecode)namecharacter_dict_pathuse_space_charSRNZSRNLabelDecodeRAREZAttnLabelDecodeNRTRZNRTRLabelDecodeSARZSARLabelDecode VisionLANZ VLLabelDecodeViTSTRZViTSTRLabelDecodeABINetZABINetLabelDecodeSPINZSPINLabelDecode RobustScannerT)rrrZ rm_symbolRFLZRFLLabelDecodePRENrZPRENLabelDecodeCANZCANLabelDecodeZrecrZdynamic)Zpreprocess_timeZinference_timeZpostprocess_time) Z model_nameZmodel_precisionZ batch_sizeZ data_shapeZ save_pathZinference_configZpidsZ process_nameZgpu_idsZ time_keyswarmuplogger)rec_image_shapesplit rec_batch_num rec_algorithmZrec_char_dict_pathrZrec_image_inverseinverserpostprocess_oputilityZcreate_predictorr predictor input_tensoroutput_tensorsconfig benchmarkuse_onnxauto_logosgetpidZget_infer_gpuidZ AutoLogger precisionZuse_gpuautolog)selfargsZpostprocess_paramsr-pidZgpu_idrrr __init__'s                zTextRecognizer.__init__cCsV|j\}}}|jdks|jdkrbt|tj}tt|}|jdkr.| ||gtj }n | ||gtj }t |}t |d}|d}|jdkrV|tjd}|S|tjdd}|S|jdkrt|tj}tj |||ftjd }|d }|d }|tjddf}|d 8}|d }|S||jd ksJt||}|jr|jjddd} | dur| dkr| }|jdd \} } | t| } t|| |kr|} n tt|| } |jdkr| |jd kr|jd } |jd }t || |f}|d }|dd }|d 8}|d }tj|||ftjd} || ddddd| f<| S)Nrrro@g`@?r interpolationfloat32?r8rrZdtype)r r#cv2cvtColorCOLOR_BGR2GRAYrZ fromarraynpuint8resizeZBICUBICZ ANTIALIASarray expand_dims transposeastyper> INTER_CUBICnewaxisshaper r,r(floatmathceilzeros)r2img max_wh_ratioimgCimgHimgWZ image_pilnorm_img resized_imagewhratioZ resized_w padding_imrrr resize_norm_imgsb               zTextRecognizer.resize_norm_imgcCsX|\}}}|dddddddf}tj|||ftjd}|d}|dd}|S)Nr6r<r>r7r?rCrH INTER_LINEARrLrKr2rT image_shaperVrWrXrZrrr resize_norm_img_vls   z!TextRecognizer.resize_norm_img_vlcCs|\}}}t||f}|jd}|jd}||dkr't||d|f} n*||dkr8t||d|f} n||dkrIt||d|f} nt|||f} t| } t| tj} | |ddd| jdf<|ddddtjf}|j\} } } d} t || | | f tj S)Nrr9r8rA) rFrSrOrCrHZasarrayrDrErNreshaperLr>)r2rTrcrVrWrXZ img_blackZim_heiZim_widZimg_newZimg_nprowcolcrrr resize_norm_img_srns$        z"TextRecognizer.resize_norm_img_srnc Cs|\}}}t|d|d}ttd||dfd}ttd||dfd} td||f} t| ddd||g} t| d|ddgddg} t | ddd||g} t| d|ddgddg} |tj ddf}| tj ddf} || | | gS)Nrr9int64r6r>ge) r rFrIrangererLonesZtriuZtileZtrilrN) r2rc num_headsmax_text_lengthrVrWrXZ feature_dimencoder_word_pos gsrm_word_posZgsrm_attn_bias_datagsrm_slf_attn_bias1gsrm_slf_attn_bias2rrr srn_other_inputssH         zTextRecognizer.srn_other_inputsc Csr|||}|tjddf}||||\}}}} |tj}| tj} |tj}|tj}||||| fS)N)rirFrNrtrLr>rk) r2rTrcrnrorYrprqrrrsrrr process_image_srn s      z TextRecognizer.process_image_srn?cCsH|\}}}}|jd}|jd} d} td|} | t|} t|| } | | dkr3t| | | } |durr?r7r@grB)rOr rPrQrRroundmaxminrCrHrLrFrNrKrmr>)r2rTrcwidth_downsample_ratiorVrWZimgW_minZimgW_maxr\r[ valid_ratioZ width_divisorr]Zresize_wrZZ resize_shaper^Z pad_shaperrr resize_norm_img_sars6           z"TextRecognizer.resize_norm_img_sarcCst|tj}t|tddgtj}t|tj}t |d}| d}dg}dg}tj|tjd}tj|tjd}t| dd}dt| dd}||8}||9}|S)Nd r6r7g_@rBr9) rCrDrErHtuplerMrFrIr>rJrKre)r2rTmeanstdZstdinvrrr resize_norm_img_spin;s  z#TextRecognizer.resize_norm_img_spincCsL|\}}}tj|||ftjd}|d}|dd}|d8}|d}|S)Nr<r>r7r?r@r`rbrrr resize_norm_img_svtrLs   z#TextRecognizer.resize_norm_img_svtrc Csz|\}}}tj|||ftjd}|d}|d}tgd}tgd}||d|d}|d}|d}|S)Nr<r>r:)g ףp= ?gv/?gCl?)gZd;O?gy&1?g?)NN.r7)rCrHrarLrFrIrK) r2rTrcrVrWrXrZrrrrr resize_norm_img_abinetWs     z%TextRecognizer.resize_norm_img_abinetc Cst|tj}|jrd|}|jddkrG|j\}}|j\}}}||ks(||krGt||d}t||d} tj|d|fd| ffddd} | }t |dd}| d}|S)Nr?rr9Zconstant)Zconstant_valuesr:r>) rCrDrEr$r rOrxrFpadrJrL) r2rTrcr\r[_rWrXZ padding_hZ padding_wZ img_paddedrrr norm_img_canis&   zTextRecognizer.norm_img_canc.Cs t|}g}|D]}||jdt|jdqtt|}ddgg|}|j}t}|j r9|j j t d||D]F} t|| |} g} |jdkrXg} g} g}g}|jdkr_g}|jdd\}}}||}t | | D]}|||jdd\}}|d |}t||}qrt | | D]}|jdkr|||||j\}}}}|tjddf}tj|dd }||| |q|jdkr|||||jd d }| |d| |d||d||d | |dq|jdkr|||||j}|tjddf}| |q|jdvr:|||||j}|tjddf}| |q|jdkrX||||}|tjddf}| |q|jdkrx|||||j}|tjddf}| |q|jdkr|j||||jdd\}}}}|tjddf}tj|dd }g}||| |g}tt ddd}tj|dd }||q|jdkr|||||}|tjddf}| |tj|jdd}tjddgdd}g} g}!| ||!|q|||||}|tjddf}| |qt| } | } |j r/|j j |jdkrt| } t| } t|}t|}| | | ||g}"|j!rli}#| |#|j"j#<|j$%|j&|#}$d|$di}%n|j$'}&t t|&D]}'|j$(|&|'}(|()|"|'qw|j$%g}$|j&D] })|)*}*|$|*q|j r|j j d|$di}%n|jdkr(t|}| tj|gtj+dg}"|j!ri}#| |#|j"j#<|j$%|j&|#}$|$d}%n~|j$'}&t t|&D]}'|j$(|&|'}(|()|"|'q|j$%g}$|j&D] })|)*}*|$|*q |j r"|j j |$d}%n9|jdkrt|}t|}| ||g}"|j!rWi}#| |#|j"j#<|j$%|j&|#}$|$d}%n |j$'}&t t|&D]}'|j$(|&|'}(|()|"|'qb|j$%g}$|j&D] })|)*}*|$|*q|j r|j j |$d}%n|jdkrt| } t|!}!| | |!g}"|j!ri}#| |#|j"j#<|j$%|j&|#}$|$}%n|j$'}&g}(t t|&D]}'|j$(|&|'}+|+)|"|'|(|+q|(|_"|j$%g}$|j&D] })|)*}*|$|*q|j r|j j |$}%nN|j!r,i}#| |#|j"j#<|j$%|j&|#}$|$d}%n5|j")| |j$%g}$|j&D] })|)*}*|$|*q<|j rS|j j t|$dkr]|$}%n|$d}%|,|%},t t|,D] }-|,|-||| |-<ql|j r|j j j-ddq?|t|fS)Nr9rgrrrAr8r;)ZaxisrjZSVTR)rrrrrrv)rz(rkrr>rB$ZpredictT)stamp).lenappendrOrPrFZargsortrIr"timer+r1timesstartrlryr#r rxr|rNrJrurrdrrrLrrmr_Z concatenatecopyrr,r(rr'runr)Zget_input_namesZget_input_handleZ copy_from_cpuZ copy_to_cpur>r%end).r2img_listZimg_numZ width_listrTindicesrec_resZ batch_numstZ beg_img_noZ end_img_noZnorm_img_batchZencoder_word_pos_listZgsrm_word_pos_listZgsrm_slf_attn_bias1_listZgsrm_slf_attn_bias2_listZ valid_ratiosrVrWrXrUinor\r[Zwh_ratiorYrr{Zword_positions_listZword_positionsZnorm_image_maskZ word_labelZnorm_img_mask_batchZword_label_listZinputsZ input_dictZoutputspredsZ input_namesir(Z output_tensoroutputZinput_tensor_iZ rec_resultZrnorrr __call__s                                                                                zTextRecognizer.__call__N)rv)__name__ __module__ __qualname__r5r_rdrirtrur|rrrrrrrrr r&sh7  #  rc CsTt|j}t|}g}g}td|jr4tjddgd tj }t dD] }||gt |j }q'|D]'}t|\}} } | sFt|}|durStd|q6||||q6z||\} } Wn!ty} zttt| tWYd} ~ nd} ~ wwt t|D]} td|| | | q|jr|jdSdS)NzIn PP-OCRv3, rec_image_shape parameter defaults to '3, 48, 320', if you are using recognition model with PP-OCRv2 or an older version, please set --rec_image_shape='3,32,320rr?)0i@rAr8zerror in loading image:{}zPredicts of {}:{})rZ image_dirrrinforrFrandomuniformrLrGrlr r"rrCZimreadformatr Exception traceback format_excexitrr+r1report)r3Zimage_file_listZtext_recognizerZvalid_image_file_listrrTrresZ image_fileflagrrErrrr mainssF      r__main__)$r.sysZPILrpathdirnameabspath__file____dir__rinsertjoinenvironrCnumpyrFrQrrZpaddleZtools.infer.utilityZinferr&Zppocr.postprocessrZppocr.utils.loggingrZppocr.utils.utilityrrrobjectrrr parse_argsrrrr s6       Q'