9iz BSSKrSSKrSSKrSSKrSSKJr SSKJr SSKJ r J r J r J r J r SSKrSSKJr SSKJrJrJr SSKJrJrJr S \ \\ \ 4S \ \\ \ 44S jrS rS rSrS4SjrSr Sr!Sr"Sr#"SS\5r$"SS\5r%Sr&Sr'Sr(Sr)Sr*Sr+Sr,\RZ"\,S\,S 3\R\S!9r/S"\4S#jr0S$r1S%\%4S&jr2S'\S \4S(jr3S)\4S*jr4S"\S \ \S44S+jr5S5S"\S,\6S-\6S \ \ \\\64S44S.jjr7S6S/\S0\6S1\6S,\6S \4 S2jjr8S3r9g)7N)Counter)deepcopy)AnyDictListTupleUnion)Image) BaseModelcomputed_fieldmodel_validator)calculate_bbox_areacalculate_overlap_ratio"calculate_projection_overlap_ratiolayout_det_resreturnc[U5nUSVs/sHo"SS:wdM UPM nn[5n[[U55Hn[US-[U55HnXT;dXd;aM[ X5SX6SS5nUS:dM.[ X5S5n[ X6S5n X5SS:Xd X6SS:XaX5SX6S:waMvX:aUR U5 MUR U5 M M [U5V Vs/sHupX;dM UPM snn US'U$s snfs snn f) a Remove overlapping boxes from layout detection results based on a given overlap ratio. Args: layout_det_res (Dict[str, List[Dict]]): Layout detection result dict containing a 'boxes' list. Returns: Dict[str, List[Dict]]: Filtered dict with overlapping boxes removed. boxeslabel reference coordinatesmallgffffff?image)rsetrangelenrradd enumerate) rlayout_det_res_filteredboxrdropped_indexesij overlap_ratio box_area_i box_area_jidxs n/var/www/html/land-doc-ocr/venv/lib/python3.13/site-packages/paddlex/inference/pipelines/paddleocr_vl/uilts.pyfilter_overlap_boxesr+"s`'~6.w7 7w<;;V7  eO 3u: q1uc%j)A#q';3&(>Ms"0,1GH 0,1GH HW%0EHW4E4Phw'58G+<<+#''*#''*!*&&e,(,0J,(G$ #"5 .(s D7D7 D<(D<cp[U[R5(aU$[R"U5$)z Convert the input to a PIL Image. Args: img (PIL.Image or numpy.ndarray): Input image. Returns: PIL.Image: PIL Image object. ) isinstancer fromarrayimgs r* to_pil_imager1Ls)#u{{## ??3 cp[U[R5(a[R"U5$U$)z Convert the input to a numpy array. Args: img (PIL.Image or numpy.ndarray): Input image. Returns: numpy.ndarray: Numpy array image. )r-r nparrayr/s r* to_np_arrayr6[s(#u{{##xx} Jr2cUVs/sHn[U5RPM nnUVs/sHn[U5RPM nn[U5n[ U5nXE4$s snfs snf)z Calculate width (max of all) and height (sum) for a vertical merge of images. Args: images (List[PIL.Image or np.ndarray]): List of images. Returns: Tuple[int, int]: (width, height) of merged image. )r1widthheightmaxsum)imagesr0widthsheightswhs r*calc_merged_whrAjsc28 8#l3%%F 839:6C|C ''6G: F A G A 4K 9:s A$A)cU(dg[U5S:Xa[US5$[U[5(aU/[U5S- -n[U5[U5S- :wa [ S5e[ US5n[ S[U55Hn[ X5nXS- n[URUR5nURUR-n[R"SXg4S5nUS:Xa#XbR- S-n XdR- S-n O&US :XaXbR- n XdR- n OS=pURX)S45 URXJUR45 UnM [U5$) a Merge images vertically with given alignment. Args: images (List[PIL.Image or np.ndarray]): List of images to merge. aligns (str or List[str]): Alignment(s) for each merge step ('center', 'right', 'left'). Returns: np.ndarray: Merged image as numpy array. Nrrz,The length of aligns must be len(images) - 1RGBrErEcenterrright) rr6r-str ValueErrorr1rr:r8r9r newpaste) r<alignsmergedr$img2alignr?r@new_imgx1x2s r* merge_imagesrS{s`  6{a6!9%%&#S[1_- 6{c&kAo%GHH &) $F 1c&k "FI&1u   djj ) MMDKK '))EA6?; H ll"q(Bjj.Q&B g \\!BZZBKB f1g& d/0!#" v r2cn ^(^)/n0n[U5H&upEUSU;aXSU'MURXE45 M( /n/n/n/n Sm(U(4Sjn Sn [U5GHun upEU(d U/nU/n/n MX,S- upUSnUSnUSnUSn[UUS5nUS:H=(ai US :H=(a] UU:H=(aQ USUS :=(a? USUS :=(a- USUS - [US US- US US- 5S -:nUS:=(a US ;=(a UU:H=(a~ US US:=(al [ USUS - 5[US US- US US- 5S-:=(a/ T("USUS5T("US US 5- =(a U "XMU5nU(aSnOU(a U "UU5nOSnU(dU(a6URU5 URU5 U RU5 GMURX45 U/nU/n/n GM U(aURX45 /nUH1unn[ U5[U5nnURUUUU45 M3 /n[ 5m)SnU[U5:Ga SnUGHunnnnUU:XdM[U)4SjU55(dM-SnUV s/sH oU SPM nn U(aUO/n[U5un n!U S:waU!U - O [S5n"U"S :a[[U5HKun#n$UU$R5nUU$SUS'SUS'URU5 T)RU$5 MM Ox[UU5n%[U5H]un#n$UU$R5nU#S:XaU%OSUS'U#S:XaUOSUS'USUS'URU5 T)RU$5 M_ /n&[US-U5Hn'U'U;dM U&RU'5 M U&H(n'URUU'5 T)RU'5 M* US-n O U(aGMXC;a*UT);a$URX45 T)RU5 US- nU[U5:aGM U$s sn f)a/ Merge blocks based on alignment and overlap logic, except for those with labels in non_merge_labels. Args: blocks (List[Dict]): List of block dicts. non_merge_labels (List[str]): Block labels that should not be merged. Returns: List[Dict]: List of processed (and possibly merged) blocks. rc"[X- 5S:*$)Nabs)a1a2s r* is_aligned merge_blocks..is_aligneds27|q  r2cZ>T"USUS5(agT"USUS5(agg)NrleftrrGrF) block_bbox prev_bboxr[s r* get_alignment#merge_blocks..get_alignments6 jmYq\ 2 2  1 y| 4 4r2cX!SnX Sn[USUS5n[USUS5n[USUS5n[USUS5nXVXx/n [U5H$upXU4;aMU Sn [X5S:dM$ g g)Nr"rrrTF)minr:r r) block_idxprev_idxblocksrar`rQy1rRy2min_boxr) other_block other_bboxs r*overlapwith_other_box+merge_blocks..overlapwith_other_boxs$U+ &u- 1z!} - 1z!} - 1z!} - 1z!} -2" )& 1 C(++$U+J&w;a? !2 r2rr" horizontalrtextrreg333333?)rr?rFNFc3,># UH oT;v M g7fNr_).0r$ used_indicess r* merge_blocks..s#Q=a\$9=sTr0inf merge_alignsgroup_id)r appendrr:rXrfrrallrAfloatcopyrrSr)*rinon_merge_labelsblocks_to_mergenon_merge_blocksr)block merged_groups current_groupcurrent_indicescurrent_alignsrbror$rh prev_blockra prev_labelr` block_labeliou_his_crossis_updown_align align_mode group_ranges group_indicesrLstartend result_blocks group_foundimgsr{r?r@ aspect_ratior%rg merged_img insert_listn_idxr[rws* @@r* merge_blocksrscO'  >- -$)S !  " "C< 0 ( MMON! %_5?L!. v']);sUC?@"/M5L C F  1= -E3 ve|#Q=#Q Q Q" 2?@-Qq %(-@)/vR %d+1()Qq1uE%L 1$(1-(@ 9 &y 1 6 6 8'-i'8'?e 04n-%,,U3$((3 )A".dL!AJ(1-(@ 9 &y 1 6 6 856!Vze @AQ Dn-,9!,<j)%,,U3$((3 )A! "519c2E 00#**513)E!(()9%)@A $$U+)Ag?2>@    "s,'>  !1!6 7   S ! qO F P GAsR2c ^SSKmSU4SjjnUVs/sHn[U5PM snupVpxXu- n X- n UR5n TRXU4Xx4SSS9 TRn Sn U"X,[ X5SS 9upn[ S [R"X-55nXYU- S --nXjU-S --nTRU UUU4U US UTRS 9 U $s snf)a$ Fill a rectangular area in the image with a white background and write the given token string. Args: image (np.ndarray): Image to paint on. box (tuple): (x1, y1, x2, y2) coordinates of rectangle. token_str (str): Token string to write. Returns: np.ndarray: Modified image. rN?c>SupEUnXT- S:a>XE-S- nT RXUSS9uupn XU-:a XU-:aUnUnOUnXT- S:aM>UWW 4$)N)g? g{Gz?rr) thickness) getTextSize) rrfontFace square_size fill_ratior^rG optimal_scalemidr?r@_cv2s r*get_optimal_font_scale+paint_token..get_optimal_font_scaleNs  lT!<1$CqIIFQA++*4L0L # lT!a""r2rD)colorr)rrr)rrr)lineType)r) rintr rectangleFONT_HERSHEY_SIMPLEXrfr:mathfloorputTextLINE_AA)rr" token_strrvrQrjrRrkbox_wbox_hr0fontthickness_scale_ratio font_scaletext_wtext_hfont_thicknesstext_xtext_yrs @r* paint_tokenr@s# '**cc!fc*NBB GE GE **,CMM#Bx"BMO  # #D!7U*s"JDJJz'IJKN6>a' 'F 6>a' 'FKK     J;+sCc2SnSSKnURS5 0nUupgp/n U"[U55n URU 5 [ U5HupU SupnnX:dMX:dMUU::dM$UU ::dM,U R U 5 [ UU- UU- 5S:aMUX- X- UU- UU- /nS[X5-S-n[UUU5nS U S S 3UU'M [ U5VVs/sHunnUU ;dMUS PM nnnXU4$s snnf) a Replace figures in a table area with tokens, return new image and token map. Args: table_block_img (np.ndarray): Table image. table_box (list): Table bounding box [x_min, y_min, x_max, y_max]. figures (List[Dict]): List of figure dicts (must contain 'coordinate', 'path'). Returns: Tuple[np.ndarray, Dict[str, str], List[str]]: - New table image, - Token-to-img HTML map, - List of figure paths dropped. c1Skn/nSn[U5U:aC[[U55U-(dURU5 US- n[U5U:aMCU$)N>019rr)rrrHr})numexclude_digitsseqr$s r*gen_random_map0tokenize_figure_of_table..gen_random_mapsT( #hnAK.0 1 FA#hn r2rNir[F]z ) randomseedrshuffler r}rfrHr)table_block_img table_boxfiguresrr token_map table_x_min table_y_min table_x_max table_y_max drop_idxes random_map figure_idfigure figure_x_min figure_y_min figure_x_max figure_y_maxdraw_boxrr$f drop_figuress r*tokenize_figure_of_tabler~sO  KKI9B6KkJG -J NN:&w/ AG AU> L,  '+ + +   i (<,. |0KLrQ**{*{* H s:#899C?I)/8YOO%/v/?s#CIi )0*+4G*<P*<$!QZIAfI*URS5nSUS3nTRX RS55$)Nrrrr)groupget)matchtoken_idtokenfigure_token_maps r*repl(untokenize_figure_of_table..repls7;;q>XJa ##E;;q>::r2z \[F(\d+)\])resub) table_res_strrrpatterns ` r*untokenize_figure_of_tablers ; G 66' //r2c\rSrSr%SrSr\\S'Sr\\S'\\S'\\S'\\S'\\S '\ \S 'S r \ \S 'S r \ \S 'S r \ \S'\"SS9\S\S\4Sj55rSrg) TableCelliaK TableCell represents a single cell in a table. Attributes: row_span (int): Number of rows spanned. col_span (int): Number of columns spanned. start_row_offset_idx (int): Start row index. end_row_offset_idx (int): End row index (exclusive). start_col_offset_idx (int): Start column index. end_col_offset_idx (int): End column index (exclusive). text (str): Cell text content. column_header (bool): Whether this cell is a column header. row_header (bool): Whether this cell is a row header. row_section (bool): Whether this cell is a row section. rrow_spancol_spanstart_row_offset_idxend_row_offset_idxstart_col_offset_idxend_col_offset_idxrrF column_header row_header row_sectionbefore)modedatarc [U[5(amSU;aU$USRSS5n[U5(d<UR SS5nU(aUH nX$SS-- nM UR 5nX!S'U$)z Create TableCell from dict, extracting 'text' property correctly. Args: data (Any): Input data. Returns: Any: TableCell-compatible dict. rrbboxrtext_cell_bboxesN )r-rrrpopstrip)clsrrr text_cellsels r*from_dict_formatTableCell.from_dict_formats dD ! !~ <##GR0Dt99!XX&8$? (7 c 11)zz|L r2r_N)__name__ __module__ __qualname____firstlineno____doc__rr__annotations__rrHrboolrrr classmethodrr__static_attributes__r_r2r*rrs HcHc IM4JK(#CC$r2rcx\rSrSr%Sr/r\\\S'Sr \ \S'Sr \ \S'\ \ S\\\4Sj55rS rg ) TableDataiz TableData holds a table's cells, row and column counts, and provides a grid property. Attributes: table_cells (List[TableCell]): List of table cells. num_rows (int): Number of rows. num_cols (int): Number of columns. table_cellsrnum_rowsnum_colsrcD[UR5VVs/sH:n[UR5Vs/sHn[SUUS-UUS-S9PM snPM< nnnURHn[[ UR UR5[ URUR55HYn[[ URUR5[ URUR55H nXCUU'M M[ M U$s snfs snnf)zn Returns a 2D grid of TableCell objects for the table. Returns: List[List[TableCell]]: Table as 2D grid. rr)rrrrrr) rrrrrrfrrrr)selfr$r% table_datacells r*gridTableData.grids(4==)  *t}}- .A)*'(1u)*'(1u . *   $$DD--t}}=D++T]];114==A//?A(,qM!$  %-   sDD DDr_N)rrrrrrrrrrrrr propertyr#rr_r2r*rrsU$&Ki%HcHc d4 ?+r2rzzzzzzz+(?:|||||)z.*?(?=z|$))flagssc $SSR[[[[[ [ /5-S-n[R"X5n[R"X5nUVs/sHoDR5(dMUPM nnX#4$s snf)z Extract OTSL tags and text parts from the input string. Args: s (str): OTSL string. Returns: Tuple[List[str], List[str]]: (tokens, text_parts) (|)) joinOTSL_NL OTSL_FCEL OTSL_ECEL OTSL_LCEL OTSL_UCEL OTSL_XCELrfindallsplitr )r'rtokens text_partsrs r*otsl_extract_tokens_and_textr7Csz ))WiIy)T U V   ZZ #F'%J%/AZE;;=%ZJA  Bs &B B c ^[m[R"UU4Sj5VVs/sHup#U(aM[U5PM nnn/nSnSnU(Ga%[ SU55nUH:n [ U 5U:dMU R [5 [ U 5U:aM&M< /n Sn UHn U Hn U R U 5 U [ U5:dM%X U :XdM/U S- n U [ U5:dMEX [[[[[[4;dMmU R X 5 U S- n M U R [5 U [ U5:dMX [:XdMU S- n M U nSn Sn[U5GH=unnSnU[[4;aSnSnSnU[:wa XS-nSnUU-[ U5:aXU-OSnSnUS-[ U5:aU[ XFS-5:a XFS-UnU[[4;aUU "XGS-U[[/5- nU[[4;aUU"XGUS-[[/5- nUR [UR5UUUUU-UUU-S 95 U[[[[[4;aUS- nU[:XdGM6US- nSnGM@ XT4$s snnf) z Parse OTSL text and tags into TableCell objects and tag structure. Args: texts (List[str]): List of tokens and text. tokens (List[str]): List of OTSL tags. Returns: Tuple[List[TableCell], List[List[str]]]: (table_cells, split_row_tokens) c>UT:H$rur_)z split_words r*"otsl_parse_texts..fs Zr2rc38# UHn[U5v M g7frurrvrows r*rx#otsl_parse_texts..os<+;Cs3xx+;rcxSnUnXUU;a*US- nUS- nU[X5:aU$XUU;aM*U$Nrrr?)r5c_idxr_idx which_tokensspan c_idx_iters r* count_right%otsl_parse_texts..count_rightsY mJ'<7 !OJ AIDS// mJ'<7  r2ctSnUnXUU;a(US- nUS- nU[U5:aU$XUU;aM(U$rEr?)r5rFrGrHrI r_idx_iters r* count_down$otsl_parse_texts..count_downsX  '<7 !OJ AIDS[(  '<7  r2rr)rrrrrrrr)r- itertoolsgroupbylistr:rr}r/r.r0r1r2r rr )textsr5xysplit_row_tokensrrGrFmax_colsrA new_textstext_idxrrKrOr$rr cell_textrr right_offsetnext_right_cellnext_bottom_cellr;s @r*otsl_parse_textsr_XsJ%%f.GHHDA QH K E E<+;<<#Cc(X% 9%c(X%$ #C  'c%j(U_-EMH#e*,!!!!! I2"((9 A    W %#e*$G)CA #$$U#4 Iy) )HHLy !a%L  ,-|+;c%j+H,&'b  " qy3/003/ :;;'7 'B5'I$9i"88K$aiI8N Iy#99J$UQYI8N   "*%%).',x'7).',x'7  Iy)Y J J QJE 7? QJEEU$V  ((Is KKr!c VURnURn[UR5S:XagSnURn[ U5HnUS- n[ U5HnXEUnUR URpURURpX:wdX:waMD[R"URR55n UR(aSOSn U nUS:a USUS3- nU S:a US U S3- nUS US U S U S 3- nM US - nM SUS3nU$)zv Export TableData to HTML table. Args: table_data (TableData): TableData object. Returns: str: HTML string. rrzthtdrz rowspan=""z colspan="<>zzz
)rrrrr#rrrrrhtmlescaperrr r)r!nrowsncolsbodyr#r$r%r"rowspanrowstartcolspancolstartcontentcelltag opening_tags r*export_to_htmlrrs>   E   E : ! !"a' D ??D 5\ uA"gajD!%0I0IX!%0I0IX} kk$))//"34G"00ddG$IK{G9A66 {G9A66 a }AgYb ; ;D !"TF( #D Kr2otsl_strc^[U[5(deUR5n[U;a U[-$UR [5n/nUHnU(dM [ R U5nU(dM*[U5nSn[U5H&upxUR[5(dM!US-nM( URXEUS.5 M U(d[$U(a[SU55OSn U(a[SU55OSn U n [X5n [S5n U n[XS-5H#m[U4SjU55nX:dMUn TnM% /nUHQnUSn[U5nUU:aUS UnO[ /UU- -nUU-nURS R#U55 MS [R#U5[-$) z Pad OTSL string to a square (rectangular) format, ensuring each row has equal number of cells. Args: otsl_str (str): OTSL string. Returns: str: Padded OTSL string. rr) raw_cells total_lenmin_lenc3*# UH oSv M g7f)rwNr_r@s r*rx%otsl_pad_to_sqr_v2..s>Xcy>Xc3*# UH oSv M g7frvNr_r@s r*rxrys=HSK(Hrzrzc3F># UHn[UST- 5v M g7fr|rW)rvrAr8s r*rxrys# S(3S%5%=!>!>(s!ruNr)r-rHr r-r4OTSL_FIND_PATTERNr3rr startswithr.r}r:rrr;r/r,)rslinesrow_datalinerurvrwr$cell_strglobal_min_width max_total_len search_start search_endmin_total_cost optimal_widthcurrent_total_costrepaired_linesrAcells current_len new_cellspaddingr8s @r*otsl_pad_to_sqr_v2rs h $ $$ $~~Hh'!! NN7 #EH %--d3   N $Y/KA""9--a%0 # P  BJs>X>>PQAIC=H==qM#L%5J5\NM|!^4 S( SS  ./N!M 5 NK %j  &n}-I k][%@AGIbggi01 << '' 11r2 otsl_contentc[U5n[U5up[X!5up4[[ U5U(a[ SU55OSUS9n[ U5$)z Convert OTSL-v1.0 string to HTML. Only 6 tags allowed: , , , , , . Args: otsl_content (str): OTSL string. Returns: str: HTML table. c38# UHn[U5v M g7frur?r@s r*rx'convert_otsl_to_html..9s;*:3c#hh*:rCr)rrr)rr7r_rrr:rr)rr5 mixed_textsrrWr!s r*convert_otsl_to_htmlr*s_&l3L6|DF$4[$I!K%&?O#;*:;;UVJ * %%r2c[U5n[SUS-S-5H!nX-S:XdM USUnX1U--U:XdMUs $ g)z Find the shortest substring that repeats to form the entire string. Args: s (str): Input string. Returns: str or None: Shortest repeating substring, or None if not found. rrrN)rr)r'nr$ substrings r*!find_shortest_repeating_substringr?sT AA 1a1fqj ! 5A:"1IF#q(  " r2rw min_repeatsc0[[U5U-US- S5HvnX*SnURXB-5(dM"SnUnURU5(a#USU*nUS- nURU5(aM#[U5XS-- nUSUXE4s $ g)a Detect if string ends with a repeating phrase. Args: s (str): Input string. min_len (int): Minimum length of unit. min_repeats (int): Minimum repeat count. Returns: Tuple[str, str, int] or None: (prefix, unit, count) if found, else None. rrNr)rrendswith)r'rwrr$unitcounttemp_s start_indexs r*find_repeating_suffixrRs3q6k*GaK <v ::d( ) )EF//$''! //$''a&EI.K\k?D/ /= r2roline_thresholdchar_thresholdcUR5nU(dU$SU;aF[U5S:a7[USSS9nU(a%Uupgn[U5U-[U5S-:aU$SU;a?[U5U:a0[U5nU(a[U5[U5-nX:aU$UR S5V s/sH)oR5(dMU R5PM+ n n U (dU$[U 5n X:aU$[ U 5n U R S5SupX:a X- S :aU $U$s sn f) as Detect and truncate character-level, phrase-level, or line-level repetition in content. Args: content (str): Input text. line_threshold (int): Min lines for line-level truncation. char_threshold (int): Min repeats for char-level truncation. min_len (int): Min length for char-level check. Returns: Union[str, str]: (truncated_content, info_string)  drV)rwrrsrrg?)r rrrr4r most_common)rorrrwstripped_content suffix_matchprefixrepeating_unitrrr total_lines line_countsmost_common_lines r*truncate_repetitive_contentrmsD}}  ##,<(=(C,-=qVWX ,8 )FE>"U*S1A-BS-HH  ##,<(=(G:;KL ()S-@@E&%%'.mmD&9 J&9dZZ\\TZZ\&9E J e*K#%.K)55a8; E$7C#? N Ks >D=D=cfSSKn[UR5S:XaURXR5nOUR 5nUR [R:waUR[R5nUR5nUR5nX4:XaU$X$- X4- - S-nUR[R5nURUSSUR5upgURU5nUcU$URU5uppX X-2XU -24n U $)NrrerE)rrshapecvtColorCOLOR_BGR2GRAYrdtyper4uint8astyper:rf thresholdTHRESH_BINARY_INV findNonZero boundingRect)r0rgraymax_valmin_valrrbinarycoordsrUrVr?r@croppeds r* crop_marginrs 399~||C!3!34xxz zzRXX{{288$hhjGhhjG Nw0 1C 7D ;;rxx D dCc.C.CDIA __V $F ~ !!&)JA!ae)QQY&'G Nr2)rF)rrV)rrr):rfrQrr collectionsrrrtypingrrrrr numpyr4PILr pydanticr r r layout_parsing.utilsrrrrHr+r1r6rArSrrrrrrr-r.r/r0r1r2NON_CAPTURING_TAG_GROUPcompileDOTALLr~r7r_rrrrrrrrrr_r2r*rs  00??'#d4j)'# #tDz/'#T   "%PZz;|84v0*3 3l. .d       GJJv&=%>cB")) C*p)f"y"J727272t&s&*sDy1A(23 +. 5c3  %&8VX/ /"%/