o yq i- @sddlZddlZddlmZddlZddlmZmZddl m Z m Z m Z ddl mZmZmZddlmZmZe d r>ddlZe d Gd d d eZe d d ed ejdededejf ddZe d dddZdS)N)Dict)Image ImageDraw)class_requires_depsfunction_requires_depsis_dep_available) SIMFANG_FONT create_fontcreate_font_vertical) BaseCVResult JsonMixinzopencv-contrib-pythonc@sjeZdZdZdejdejfddZdeee j ffddZ deeeffdd Z deeeffd d Z d S) OCRResultz OCR resultpointsreturncCst|}ttt|ddd}d\}}}}|dd|ddkr)d}d}nd}d}|dd|ddkr>d}d}nd}d}t||||||||gtj}|S) a} Get the minimum area rectangle for the given points using OpenCV. Args: points (np.ndarray): An array of 2D points. Returns: np.ndarray: An array of 2D points representing the corners of the minimum area rectangle in a specific order (clockwise or counterclockwise starting from the top-left corner). cSs|dS)Nr)xrrm/home/app/PaddleOCR-VL/.venv_paddleocr/lib/python3.10/site-packages/paddlex/inference/pipelines/ocr/result.py.sz,OCRResult.get_minarea_rect..)key)rr rrr r) cv2Z minAreaRectsortedlistZ boxPointsnparrayZastypeint32)selfrZ bounding_boxZindex_aZindex_bZindex_cZindex_dboxrrrget_minarea_rect"s&  zOCRResult.get_minarea_rectcCsd|vr}g}g}dd|dD}dd|dD}t|D]]\}}|}tt|dd|ddd|dd |dd d}tt|dd|d dd|dd |d d d} |dksn| dkroq|||||qn|d }|d }|d d } | jdd\} } t| tj} t | }t j | | dft j dd}tdt|}t}|dr|dd}tt||D]\}\}}ztddtddtddf}t |}t|dkrEdd|D}|j||d|d||}tdt|ddd ft|ddd f}t |ddd f|ddd f<t |ddd ftd||ddd f<ndd|D}|j||dt|tr_|d}t| | f|||j }t |t j!"d}t#||gd|d t$||}WqYqt %t | |d}t &d| d| fd}|'|dd| | f|'t || d| d| f|d}d|i}|d r|j(d!i|d j)|S)"z Converts the internal data to a PIL Image with detection and recognition results. Returns: Dict[Image.Image]: A dictionary containing two images: 'doc_preprocessor_res' and 'ocr_res_img'. text_word_regioncSg|] }|D]}|qqSrr.0Zsublistitemrrr Os z%OCRResult._to_img..cSr#rrr$rrrr'Rs text_wordrr rr rec_polys rec_textsdoc_preprocessor_resZ output_imgZdtypeZ vis_fontsrcSsg|]\}}||fqSrrr%ryrrrr'{s)outlinewidthfillg?NcSs g|] \}}t|t|fqSr)intr.rrrr's )r3)rrTRGBr-r-r-model_settingsZ ocr_res_imguse_doc_preprocessorr)* enumerater5mathsqrtappendshaperZcvtColorZ COLOR_BGR2RGBrZ fromarrayrZonesuint8randomseedrDrawr ziprandintrlentolistZpolygonr!maxminmean isinstancetupledraw_box_txt_finepathrZreshapeZ polylinesZ bitwise_andblendnewZpasteupdateimg)rZboxesZtxtsr"r(idxZ word_regionZchar_box box_height box_widthimagehwZ image_rgbZimg_leftZ img_rightZ draw_leftZvis_fontr txtcolorZptsheightZbox_ptsimg_right_textZimg_showr9Z res_img_dictrrr_to_imgDs           0".  zOCRResult._to_imgcOsDi}|d|d<|d|d<|d|d<|ddr#|djd|d<|ddkr-|d nt|d |d <|d |d <|d|d<d |vrOt|d |d <|d |d <|d |d <|d|d<t|d|d<|ddkrt|dnt|d|d<t|d|d<d|vr|d|d<|d|d<tj|g|Ri|S)axConverts the instance's attributes to a dictionary and then to a string. Args: *args: Additional positional arguments passed to the base class method. **kwargs: Additional keyword arguments passed to the base class method. Returns: Dict[str, str]: A dictionary with the instance's attributes converted to strings. input_path page_indexr9r:r+res text_typeZsealdt_polystext_det_paramstextline_orientation_anglestext_rec_score_threshreturn_word_boxr* rec_scoresr) rec_boxestext_word_boxesr()strrrr_to_strrargskwargsdatarrrrks:               zOCRResult._to_strcOsi}|d|d<|d|d<|d|d<|ddr#|djd|d<|d|d<|d|d<|d |d <d |vr?|d |d <|d |d <|d |d <|d |d <|d|d<|d|d<|d|d<d|vrs|d|d<|d|d<tj|g|Ri|S)a[ Converts the object's data to a JSON dictionary. Args: *args: Positional arguments passed to the JsonMixin._to_json method. **kwargs: Keyword arguments passed to the JsonMixin._to_json method. Returns: Dict[str, str]: A dictionary containing the object's data in JSON format. r^r_r9r:r+r`rbrcrardrerfr*rgr)rhrir()jsonr_to_jsonrlrrrrqs*                zOCRResult._to_jsonN) __name__ __module__ __qualname____doc__rndarrayr!rrjrr]rkrqrrrrrs "T+rimg_sizer rY font_pathrc Csttt|dd|dddt|dd|ddd}ttt|dd|dddt|dd|ddd}|d|krw|dkrwtd||fd}t|}|rvt|||f|}t |d||d dd n#td||fd}t|}|rt |||f|}|j ddg|d |d t ddg|dg||gd|gg} t j|t j d } t| | } t j|t jd }tj|| |tjtjdd } | S)a Draws text in a box on an image with fine control over size and orientation. Args: img_size (tuple): The size of the output image (width, height). box (np.ndarray): A 4x2 numpy array defining the corners of the box in (x, y) order. txt (str): The text to draw inside the box. font_path (str): The path to the font file to use for drawing the text. Returns: np.ndarray: An image with the text drawn in the specified box. rr rrr7r8)rrrrr)r3 line_spacing)r3fontr,)flagsZ borderModeZ borderValue)r5r<r=floatrrPrrCr draw_vertical_textr textrZfloat32rrZgetPerspectiveTransformr@ZwarpPerspectiveZ INTER_NEARESTZBORDER_CONSTANT) rwr rYrxrTrUZimg_textZ draw_textr|Zpts1Zpts2Mr\rrrrMsFDD    rMrzrc CsR|\}}|D] }|j||f|||d||} | d| d} || |7}qdS)N)r|r3r r)rZgetbbox) Zdrawpositionrr|r3r{rr/charZbbox char_heightrrrrs r)rzr)r<rAtypingrnumpyrZPILrrZ utils.depsrrrZ utils.fontsr r r Z common.resultr rrrrLrvrjrMrrrrrs6 H 7