o yq iSU@sddlmZmZmZmZmZddlZddlm Z ddl m Z ddl m Z ddlmZdd lmZdd lmZdd lmZd d lmZd dlmZd dlmZmZmZmZmZm Z ddl!m"Z"ej#GdddeZ$e dddGdddeZ%dS))AnyDictListOptionalUnionN)logging)pipeline_requires_extra)ImageBatchSampler) ReadImage) benchmark) HPIConfig)PaddlePredictorOption)(AutoParallelImageSimpleInferencePipeline) BasePipeline) CropByPolys SortPolyBoxes SortQuadBoxescal_ocr_word_boxconvert_points_to_boxes rotate_image) OCRResultcseZdZdZ    d%dedeedeededee eee fe fd df fd d Z d e ejd e ed e ejfddZded efddZdeedeedeed efddZ      d&deedeedeedeedeedeed efddZ           d'd e ee eeje ejfdeedeedeedeedeedeedeedeedeed!eed"eed efd#d$ZZS)( _OCRPipelinez OCR PipelineNFconfigdevice pp_optionuse_hpip hpi_configreturnc s\tj||||d|dd|_|jr&|didddi}|||_|dd|_|jrB|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$|_ |dd|_|dd%|_|dd|_t|_td&d |_ntd'|j |j||j |j |j |j |j|j|jd(|_|d id)d d*i} | d+d,|_| d-d.|_| dd|_|j| |jd/|_t|d0d1d2|_td3d4|_dS)5a Initializes the class with given configurations and options. Args: config (Dict): Configuration dictionary containing various settings. device (str, optional): Device to run the predictions on. Defaults to None. pp_option (PaddlePredictorOption, optional): PaddlePredictor options. Defaults to None. use_hpip (bool, optional): Whether to use the high-performance inference plugin (HPIP) by default. Defaults to False. hpi_config (Optional[Union[Dict[str, Any], HPIConfig]], optional): The default high-performance inference configuration dictionary. Defaults to None. )rrrr use_doc_preprocessorTZ SubPipelinesZDocPreprocessorZpipeline_config_errorz+config error for doc_preprocessor_pipeline!use_textline_orientationZ SubModulesZTextLineOrientationZmodel_config_errorz,config error for textline_orientation_model!Z TextDetectionz config error for text_det_model! text_typegenerallimit_side_leni limit_typemaxmax_side_limitithreshg333333? box_threshg333333? input_shapeN unclip_ratiog@quad)Z det_box_typeZsealiming?g?ZpolyzUnsupported text type {})r&r'r)r*r+r-r,ZTextRecognitionz config error for text_rec_model!Z score_threshrreturn_word_boxF)r, batch_sizer)r1ZBGR)format) super__init__getr"Zcreate_pipelinedoc_preprocessor_pipeliner#Z create_modeltextline_orientation_modelr$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 ValueErrorr2text_det_modeltext_rec_score_threshr0text_rec_modelr batch_samplerr img_reader) selfrrrrr Zdoc_preprocessor_configZtextline_orientation_configZtext_det_configZtext_rec_config __class__o/home/app/PaddleOCR-VL/.venv_paddleocr/lib/python3.10/site-packages/paddlex/inference/pipelines/ocr/pipeline.pyr4+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) rFrKrLZangleZrotated_imagesZ image_arrayZrotate_indicatorZ rotate_angleZ rotated_imagerIrIrJrs  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#)rFrRrIrIrJcheck_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)rFrUrVr#r"rIrIrJget_model_settingssz_OCRPipeline.get_model_settingsr8r9r:r;r<r=cCsh|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+r-)r8r9r:r;r<r=rW)rFr8r9r:r;r<r=rIrIrJget_text_det_paramss(z _OCRPipeline.get_text_det_paramsinputrBr0c %#s|||sddiV||||| | dur#jdur*jt|D]\} }|j}drKt j |||d}ndd|D}dd|D}t j |fi}d d|Dfd dDfd dt |j |j|D}t tt|}fd d|D}|rg}d g}|D]}t ||||||dtqdrdd|D}||}ndgt|}t|D]\}}||}|||||d|d<qt|D]\}}|||||d||}|}ddtD}t|ddd}fdd|D}tj|dD]\}}||d} ||| d<q8rUg|d<g|d<tt|D]c}!||!d}|dkrrt|dd ||!|dd\}"}#|d|"|d|#|d|dd n |d|d|d |d|d!|d"|d#||!q[q|D].}jd$krt|d#}$|$|d%<rd&d|dD|d'<ntg|d%<t|Vqq1dS)(aQ Predict OCR results based on input images or arrays with optional preprocessing steps. Args: input (Union[str, list[str], np.ndarray, list[np.ndarray]]): Input image of pdf path(s) or numpy array(s). use_doc_orientation_classify (Optional[bool]): Whether to use document orientation classification. use_doc_unwarping (Optional[bool]): Whether to use document unwarping. use_textline_orientation (Optional[bool]): Whether to use textline orientation prediction. text_det_limit_side_len (Optional[int]): Maximum side length for text detection. text_det_limit_type (Optional[str]): Type of limit to apply for text detection. text_det_max_side_limit (Optional[int]): Maximum side length for text detection. text_det_thresh (Optional[float]): Threshold for text detection. text_det_box_thresh (Optional[float]): Threshold for text detection boxes. text_det_unclip_ratio (Optional[float]): Ratio for unclipping text detection boxes. text_rec_score_thresh (Optional[float]): Score threshold for text recognition. return_word_box (Optional[bool]): Whether to return word boxes along with recognized texts. Returns: OCRResult: Generator yielding OCR results for each input image. rSz0the input params for model settings are invalid!Nr")rUrVcSsg|]}d|iqSZ output_imgrI).0ZarrrIrIrJ _z(_OCRPipeline.predict..cSg|]}|dqSr[rIr\itemrIrIrJr]ascSr_)dt_polysrIr`rIrIrJr]ir^csg|]}|qSrI)r>r`rFrIrJr]kscs6g|]\}}}}||||jggggd qS)) input_path page_indexdoc_preprocessor_resrbrRtext_det_paramsr$rBr0 rec_texts rec_scores rec_polys vis_fonts)r$)r\rdrerfrb)rRr0rFrgrBrIrJr]ms" cs g|] }t|dkr|qS)r)rO)r\idx) dt_polys_listrIrJr]s rr#cSsg|] }t|ddqS)Z class_idsr)int)r\Ztextline_angle_inforIrIrJr]srZtextline_orientation_anglescSs.g|]\}}||jdt|jddqS)rr) sub_img_id sub_img_ratio)shapefloat)r\Zimg_idZsub_imgrIrIrJr]s cSs|dS)NrqrI)xrIrIrJsz&_OCRPipeline.predict..)keycsg|]}|dqS)rprI)r\rt)all_subs_of_imgrIrJr]s )r0rprec_resZ text_wordZtext_word_regionZ rec_scoreZrec_textrhrirkZvis_fontrjr% rec_boxescSsg|]}t|qSrI)r)r\linerIrIrJr]sZtext_word_boxes)rXrTrYrBr0 enumeraterDrEZ instanceslistr6rArPZ input_pathsZ page_indexesrangerOr?extendrQr7rsortedrCrr$rnparrayr)%rFrZrUrVr#r8r9r:r;r<r=rBr0_Z batch_dataZ image_arraysZdoc_preprocessor_resultsZdoc_preprocessor_imagesZ det_resultsresultsindicesZall_subs_of_imgsZ chunk_indicesrlZanglesiresrbZsub_img_info_listZsorted_subs_infoZsorted_subs_of_imgrxrpZsnoZword_box_content_listZ word_box_listryrI)rwrmrRr0rFrgrBrJpredicts#               z_OCRPipeline.predict)NNFN)NNNNNN) NNNNNNNNNNN)__name__ __module__ __qualname____doc__rrstrrboolrrrr4rrZndarrayrorrTrWrXrsrYrr __classcell__rIrIrGrJr'sb $ " 1     rZocrzocr-core)Zaltc@s$eZdZdZeddZddZdS) OCRPipelineZOCRcCstS)N)rrcrIrIrJ _pipeline_clsszOCRPipeline._pipeline_clscCs |ddS)Nr1r)r5)rFrrIrIrJ_get_batch_sizes zOCRPipeline._get_batch_sizeN)rrrentitiespropertyrrrIrIrIrJrs   r)&typingrrrrrnumpyrutilsrZ utils.depsr Zcommon.batch_samplerr Z common.readerr Zutils.benchmarkr Z utils.hpirZutils.pp_optionrZ _parallelrbaser componentsrrrrrrresultrZ time_methodsrrrIrIrIrJs(            >