o 0jsY@sddlmZmZmZmZmZddlZddlm Z ddl m Z ddl m Z ddlmZdd lmZmZdd lmZd d lmZd d lmZd dlmZmZmZmZmZmZddl m!Z!ej"GdddeZ#e dddGdddeZ$dS))AnyDictListOptionalUnionN)logging)pipeline_requires_extra)ImageBatchSampler) ReadImage) HPIConfigPaddlePredictorOption) benchmark)(AutoParallelImageSimpleInferencePipeline) BasePipeline) CropByPolys SortPolyBoxes SortQuadBoxescal_ocr_word_boxconvert_points_to_boxes rotate_image) OCRResultcseZdZdZddddddddedeedeedeeeefd eed e d ee eeefe fd dffd dZ de ejde ed e ejfddZded e fddZdee dee dee d efddZ      d(deedeedeedeedeed eed efd!d"Z           d)d#e ee eeje ejfdee dee dee deedeedeedeedeed eed$eed%ee d efd&d'ZZS)* _OCRPipelinez OCR PipelineNFdeviceengine engine_config pp_optionuse_hpip hpi_configconfigrrrr r!r"returnc shtjd5||||||d||dd|_|jr,|didddi} || |_|dd|_|jrH|d id d d i} || |_|d id d di} |d|_ |j dkr| dd|_ | dd|_ | dd|_ | dd|_ | dd|_| dd|_| dd|_t|_tdd |_nI|j d!kr| dd"|_ | dd#|_ | dd|_ | dd$|_ | dd|_| dd%|_| dd|_t|_td&d |_ntd'|j |j| |j |j |j |j |j|j|jd(|_|d id)d d*i} | d+d,|_| d-d.|_| dd|_|j| |jd/|_t|d0d1d2|_td3d4|_dS)6aInitializes the OCR pipeline. Args: config (Dict): Configuration dictionary containing various settings. device (Optional[str], optional): The device to use for prediction. Defaults to `None`. engine (Optional[str], optional): Inference engine. Defaults to `None`. engine_config (Optional[Dict[str, Any]], optional): Engine-specific config. Defaults to `None`. pp_option (Optional[PaddlePredictorOption], optional): Paddle predictor options. Defaults to `None`. use_hpip (bool, optional): Whether to use HPIP. Defaults to `False`. hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional): HPIP configuration. Defaults to `None`. ruse_doc_preprocessorTZ SubPipelinesZDocPreprocessorZpipeline_config_errorz+config error for doc_preprocessor_pipeline!use_textline_orientationZ SubModulesZTextLineOrientationZmodel_config_errorz,config error for textline_orientation_model!Z TextDetectionz config error for text_det_model! text_typegenerallimit_side_leni limit_typemaxmax_side_limitithreshg333333? box_threshg333333? input_shapeN unclip_ratiog@Zquad)Z det_box_typeZsealiming?g?polyzUnsupported text type {})r)r*r,r-r.r0r/ZTextRecognitionz config error for text_rec_model!Z score_threshrreturn_word_boxF)r/ batch_sizer)r4ZBGR)format) super__init__getr%Zcreate_pipelinedoc_preprocessor_pipeliner&Z create_modeltextline_orientation_modelr'text_det_limit_side_lentext_det_limit_typetext_det_max_side_limittext_det_threshtext_det_box_threshr/text_det_unclip_ratior _sort_boxesr_crop_by_polysr ValueErrorr5text_det_modeltext_rec_score_threshr3text_rec_modelr batch_samplerr img_reader) selfr#rrrr r!r"kwargsZdoc_preprocessor_configZtextline_orientation_configZtext_det_configZtext_rec_config __class__r6i/var/www/html/Deteccion_Ine/venv/lib/python3.10/site-packages/paddlex/inference/pipelines/ocr/pipeline.pyr8*s        z_OCRPipeline.__init__image_array_listrotate_angle_listc Cst|t|ksJdt|dt|d|D] }|dvs&Jd|qg}t||D]\}}|d}t||}||q.|S)a Rotate the given image arrays by their corresponding angles. 0 corresponds to 0 degrees, 1 corresponds to 180 degrees. Args: image_array_list (List[np.ndarray]): A list of input image arrays to be rotated. rotate_angle_list (List[int]): A list of rotation indicators (0 or 1). 0 means rotate by 0 degrees 1 means rotate by 180 degrees Returns: List[np.ndarray]: A list of rotated image arrays. Raises: AssertionError: If any rotate_angle is not 0 or 1. AssertionError: If the lengths of input lists don't match. zLength of image_array_list (z*) must match length of rotate_angle_list ())rrz&rotate_angle must be 0 or 1, now it's )lenziprappend) rJrOrPZangleZrotated_imagesZ image_arrayZrotate_indicatorZ rotate_angleZ rotated_imager6r6rNrs  z_OCRPipeline.rotate_imagemodel_settingscCs<|dr|jstddS|dr|jstddSdS)a8 Check if the input parameters are valid based on the initialized models. Args: model_info_params(Dict): A dictionary containing input parameters. Returns: bool: True if all required models are initialized according to input parameters, False otherwise. r%zRSet use_doc_preprocessor, but the models for doc preprocessor are not initialized.Fr&z^Set use_textline_orientation, but the models for use_textline_orientation are not initialized.T)r%rerrorr&)rJrVr6r6rNcheck_model_settings_valids z'_OCRPipeline.check_model_settings_validuse_doc_orientation_classifyuse_doc_unwarpingr&cCsL|dur |dur |j}n |dus|durd}nd}|dur |j}t||dS)a Get the model settings based on the provided parameters or default values. Args: use_doc_orientation_classify (Optional[bool]): Whether to use document orientation classification. use_doc_unwarping (Optional[bool]): Whether to use document unwarping. use_textline_orientation (Optional[bool]): Whether to use textline orientation. Returns: dict: A dictionary containing the model settings. NTF)r%r&)r%r&dict)rJrYrZr&r%r6r6rNget_model_settingssz_OCRPipeline.get_model_settingsr<r=r>r?r@rAcCsh|dur|j}|dur|j}|dur|j}|dur|j}|dur#|j}|dur*|j}t||||||dS)a  Get text detection parameters. If a parameter is None, its default value from the instance will be used. Args: text_det_limit_side_len (Optional[int]): The maximum side length of the text box. text_det_limit_type (Optional[str]): The type of limit to apply to the text box. text_det_max_side_limit (Optional[int]): The maximum side length of the text box. text_det_thresh (Optional[float]): The threshold for text detection. text_det_box_thresh (Optional[float]): The threshold for the bounding box. text_det_unclip_ratio (Optional[float]): The ratio for unclipping the text box. Returns: dict: A dictionary containing the text detection parameters. N)r)r*r-r,r.r0)r<r=r>r?r@rAr[)rJr<r=r>r?r@rAr6r6rNget_text_det_paramss(z _OCRPipeline.get_text_det_paramsinputrFr3c )#sd|||sddiV||||| | dur#jdur*jt|D]\} }|j}drKt j |||d}ndd|D}dd|D}t j |fi}d d|Dfd dDfd dt |j |j|D}r|D] }g|d <g|d <qt tt|}fdd|D}|rg}dg}|D]U}t |||g}g}t |D]!\}}|jdkr|jddkr|jddkr||||q||<|||d<||||dt|qdrdd|D}||}ndgt|}t|D]\}}||}|||||d|d<q)t|D]\}}|||||d||}|}ddtD} t| ddd}!fdd|!D}"tj|"dD]\}}#|!|d}$|#| |$d<qtt| D]c}%| |%d}#|#dkrrt|#dd||%|#dd\}&}'|d |&|d |'|d |#ddn |d |#d|d!|#d|d"|#d#|d$||%qqE|D].}jd%kr!t|d$}(|(|d&<r d'd|d D|d(<ntg|d&<t|Vqq1dS))aQ Predict OCR results based on input images or arrays with optional preprocessing steps. Args: input (Union[str, list[str], np.ndarray, list[np.ndarray]]): Input image of pdf path(s) or numpy array(s). use_doc_orientation_classify (Optional[bool]): Whether to use document orientation classification. use_doc_unwarping (Optional[bool]): Whether to use document unwarping. use_textline_orientation (Optional[bool]): Whether to use textline orientation prediction. text_det_limit_side_len (Optional[int]): Maximum side length for text detection. text_det_limit_type (Optional[str]): Type of limit to apply for text detection. text_det_max_side_limit (Optional[int]): Maximum side length for text detection. text_det_thresh (Optional[float]): Threshold for text detection. text_det_box_thresh (Optional[float]): Threshold for text detection boxes. text_det_unclip_ratio (Optional[float]): Ratio for unclipping text detection boxes. text_rec_score_thresh (Optional[float]): Score threshold for text recognition. return_word_box (Optional[bool]): Whether to return word boxes along with recognized texts. Returns: OCRResult: Generator yielding OCR results for each input image. rWz0the input params for model settings are invalid!Nr%)rYrZcSsg|]}d|iqSZ output_imgr6).0Zarrr6r6rN hz(_OCRPipeline.predict..cSg|]}|dqSr_r6r`itemr6r6rNrajscSrc)dt_polysr6rdr6r6rNrarrbcsg|]}|qSr6)rBrdrJr6rNratscs6g|]\}}}}||||jggggd qS)) input_path page_indexdoc_preprocessor_resrfrVtext_det_paramsr'rFr3 rec_texts rec_scores rec_polys vis_fonts)r')r`rhrirjrf)rVr3rJrkrFr6rNravs" Z text_wordZtext_word_regioncs g|] }t|dkr|qS)r)rS)r`idx) dt_polys_listr6rNras rrrfr&cSs,g|]}ttj|dtjddqS)Z class_ids)Zdtyper)intnpZasarrayZint64Zravel)r`Ztextline_angle_infor6r6rNras Ztextline_orientation_anglescSs.g|]\}}||jdt|jddqS)rr) sub_img_id sub_img_ratio)shapefloat)r`Zimg_idsub_imgr6r6rNras cSs|dS)Nrvr6)xr6r6rNsz&_OCRPipeline.predict..)keycsg|]}|dqS)rur6)r`rz)all_subs_of_imgr6rNras )r3rurec_resZ rec_scoreZrec_textrlrmroZvis_fontrnr( rec_boxescSsg|]}t|qSr6)r)r`liner6r6rNrasZtext_word_boxes) r\rXr]rFr3 enumeraterHrIZ instanceslistr:rErTZ input_pathsZ page_indexesrangerSrCsizerwrUextendr;rsortedrGrr'rrtarrayr))rJr^rYrZr&r<r=r>r?r@rArFr3_Z batch_dataZ image_arraysZdoc_preprocessor_resultsZdoc_preprocessor_imagesZ det_resultsresultsresindicesZall_subs_of_imgsZ chunk_indicesrpZ filtered_subsZfiltered_polysryr2ZanglesirfZsub_img_info_listZsorted_subs_infoZsorted_subs_of_imgr~ruZsnoZword_box_content_listZ word_box_listrr6)r}rqrVr3rJrkrFrNpredict$s#                      z_OCRPipeline.predict)NNNNNN) NNNNNNNNNNN)__name__ __module__ __qualname____doc__rrstrrrboolrr r8rrtZndarrayrsrrXr[r\rxr]rr __classcell__r6r6rLrNr&s  l $ " 1     rZocrzocr-core)Zaltc@s$eZdZdZeddZddZdS) OCRPipelineZOCRcCstS)N)rrgr6r6rN _pipeline_clsszOCRPipeline._pipeline_clscCs |ddS)Nr4r)r9)rJr#r6r6rN_get_batch_sizes zOCRPipeline._get_batch_sizeN)rrrentitiespropertyrrr6r6r6rNrs   r)%typingrrrrrnumpyrtutilsrZ utils.depsr Zcommon.batch_samplerr Z common.readerr modelsr rZutils.benchmarkrZ _parallelrbaser componentsrrrrrrresultrZ time_methodsrrr6r6r6rNs&          Z