j dddlZddlZddlmZddlmZddlZddlm Z dgZ Gdde Z dS)N)Path)Any)AbstractDataset OCRDatasetc <eZdZdZ d dededededdf fd ZxZS) raImplements an OCR dataset >>> from doctr.datasets import OCRDataset >>> train_set = OCRDataset(img_folder="/path/to/images", >>> label_file="/path/to/labels.json") >>> img, target = train_set[0] Args: img_folder: local path to image folder (all jpg at the root) label_file: local path to the label file use_polygons: whether polygons should be considered as rotated bounding box (instead of straight ones) **kwargs: keyword arguments from `AbstractDataset`. F img_folder label_file use_polygonskwargsreturnNc tj|fi|g|_tj}t |d5}t j|}dddn #1swxYwY|D]V\}} t|}tj tj |j|s5tdtj |j|t!| ddkr@|j|t%tjd|gfd| dD} |r d | D} d | dD} |j|t%tj| || fXdS) Nrbzunable to locate typed_wordsr)r)dtype)boxeslabelsc ng|]2}ttt|ddd3S)geometryNr)listmapfloat.0objs U/var/www/html/Carbon-Document/venv/lib/python3.11/site-packages/doctr/datasets/ocr.py z'OCRDataset.__init__..<s7aaasT#eS_RaR%899::aaacrg|]4}|dd|d|dg|dd|d|dgg5S)Nrr)rgeoms rrz'OCRDataset.__init__..?s["1"XQa148d1gtAw=OPrcg|] }|d S)valuer#rs rrz'OCRDataset.__init__..DsOOOSCLOOOr)super__init__datanpfloat32openjsonloaditemsrospathexistsjoinrootFileNotFoundErrorlenappenddictzerosasarray) selfr r r r np_dtypefr)img_name annotationsgeoms text_targets __class__s rr(zOCRDataset.__init__!s ..v...8: : *d # # q9Q<>"',,ty("C"CDD a'(_BGLLT\<]<](_(_```;}-..!33   (Drxh7W7W7W`b,c,c,c!deeeaakR_F`aaaE  % POK 4NOOOL I  h2:e83T3T3T]i(j(j(jk l l l l- m msA  A$'A$)F) __name__ __module__ __qualname____doc__strboolrr( __classcell__)rBs@rrrs  $# %m%m%m%m %m  %m  %m%m%m%m%m%m%m%m%m%mr) r-r0pathlibrtypingrnumpyr*datasetsr__all__rr#rrrOs  %%%%%% .4m4m4m4m4m4m4m4m4m4mr