^j|GddlmZddlZddlZddlmZddlmZmZddl Z ddl mZ ddl mZddlmZddlmZddlmZdd lmZmZdd lmZmZmZmZmZerdd lmZdd Z dd Z!ddZ" ddZ#ddZ$ddZ%ddZ& d d dZ' d! d"dZ( d# d$dZ) d% d&dZ* d% d'dZ+d(dZ,y))) annotationsN)Path) TYPE_CHECKINGAny)Image)ORIENTED_BOX_COORDINATES)approximate_mask_with_polygons) Detections)polygon_to_maskpolygon_to_xyxy)list_files_with_extensions read_txt_fileread_yaml_filesave_text_filesave_yaml_file)DetectionDatasetc "|\}}}}tjt|t|dz z t|t|dz z t|t|dz zt|t|dz zgtjS)Ndtype)nparrayfloatfloat32)valuesx_centery_centerwidthheights k/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/supervision/dataset/formats/yolo.py _parse_boxr!s(.%Hhv 88 (OeElQ. . (OeFma/ / (OeElQ. . (OeFma/ /  jj cttj|d|dg|d|dg|d|dg|d|dggS)Nrr)rrboxs r _box_to_polygonr()sS 88 a&#a& CFCF+c!fc!f-=AA?OP r"cltj|tjjddS)Nrr)rrrreshapers r _parse_polygonr-/s$ 88F"** - 5 5b! <Bs+ %8$DJJL!A%8 998s#4c tt|}t|ts%td|dt |j d|j d}t|trt|j}dd}|Dcgc] }|| }}t|rjt|s_t||Dcgc] \}}|s | c}}dd}t||Dcgc] \}}|r | c}}dd} td |d |d | d t|rt|d } nt|t} | D cgc]} t|| c} St|tr|D cgc] } t| c} Std|dt |j dcc}wcc}}wcc}}wcc} wcc} w)aReturn class names from a YOLO data.yaml file ordered by class index. Supports list and dict forms of the ``names`` field. Dict keys that are all int-like (plain ints or digit strings) are sorted numerically so class index 10 follows index 9. All-non-numeric keys are sorted lexicographically. Mixed numeric/non-numeric keys raise ``ValueError``. Boolean YAML keys (``true``/``false``) are excluded from numeric sorting because ``bool`` is a subclass of ``int`` in Python. Args: file_path: Path to the data.yaml file. Returns: Class names in class-index order. Raises: ValueError: If the YAML root is not a mapping, if ``names`` is neither a list nor a dict, or if the dict has mixed key types.  file_pathz"Expected mapping in data.yaml at 'z', got .namesct|tryt|tryt|tr |j }|j Sy)NFT) isinstancer4intstrstripisdigit)keystrippeds r _is_int_likez*_extract_class_names.._is_int_likedsC#t$#s##s#99;''))r"Nr%z'Expected 'names' dict in data.yaml at 'zI' to have either all numeric or all non-numeric keys, got a mix: numeric z and non-numeric z keys.ct|SN)rF)ks r z&_extract_class_names..ys SVr")rJz7Expected 'names' to be a list or dict in data.yaml at ')rJrreturnr4)rrEdict ValueErrortype__name__getlistkeysr9allzipsortedrG) rAdatarCrXrLrOint_likeil mixed_numeric mixed_other sorted_keysrJnames r _extract_class_namesrcFs(*I>D dD !0 <J''( +   HHW E%EJJL! .22LO22 x=X,/h,?F51b2QFrJM*-dH*=HBR1H!LK9)E(/):;-vO  x= +;@Hm-/@ DAq1F1I' v;! F12J/C   % ''C(@A [1_$F12J7G  !A B!((&*)=> ''01 xx,HHH]"**=M 288Q1aL C CD D HH%6bjjI$,,RA6BHHaV2::66)1 %& 8$TBB(  "((= ;;H x} MD xdD II s/I&c|r|rtjdtdt|gdDcgc] }t |}}t |}i}|D]} t | j} tjj|| d} tjj| stj|| <rtj| } t!| d } | j"\}}||f}| j$d vrt'd | d | j$d | xr|xs t)| }t+| |||}||| <|||fScc}w)aM Loads YOLO annotations and returns class names, images, and their corresponding detections. Args: images_directory_path: The path to the directory containing the images. annotations_directory_path: The path to the directory containing the YOLO annotation files. data_yaml_path: The path to the data YAML file containing class information. force_masks: If True, forces masks to be loaded for all annotations, regardless of whether they are present. This parameter has no effect when `is_obb=True`; mask generation is always disabled for OBB annotations. is_obb: If True, loads the annotations in OBB format. OBB annotations are defined as `[class_id, x, y, x, y, x, y, x, y]`, where pairs of [x, y] are box corners. Returns: A tuple containing a list of class names, a dictionary with image names as keys and images as values, and a dictionary with image names as keys and corresponding Detections instances as values. zl`force_masks=True` has no effect when `is_obb=True`; mask generation is always disabled for OBB annotations.r stacklevel) bmpdngjpgjpegmpopngtiftiffwebp) directory extensionsr@reT)rA skip_empty)RGBLz9Images must be 'RGB' or 'grayscale', but z mode is 'z'.)r<)r<r0rurv)warningswarn UserWarningr rGrcrstemrfrgjoinexistsr rsropenrsizemoderSr>r})images_directory_pathannotations_directory_pathdata_yaml_path force_masksrvrg image_pathsclassesr image_path image_stemannotation_pathimager<rzr{r0ru annotations r load_yolo_annotationsrsy<+  F   /+     D K$#^"PQh-.!122#Qs= Ccl|r:t|dkDr,t|jvrtdtdtd|r0|j$t j dtdtdg}|D]=\}}} } } } | td t| ttjfstd t| d t| } |r|tj| ttj } | jd k7r td| jdtdt!|| || }|j#||Dt%||||}|D].}t'|}t!|| ||}|j#|0t!|| |}|j#|@|S)a$Convert detections to YOLO annotation lines. Args: detections: The detections to serialize. Each detection must have a valid integer ``class_id``. When ``is_obb=True``, each non-empty detection must also carry ``detections.data['xyxyxyxy']`` with shape ``(N, 4, 2)``. image_shape: The ``(height, width, channels)`` shape of the source image, used to normalize coordinates to ``[0, 1]``. min_image_area_percentage: Minimum detection area as a fraction of the image area; smaller detections are omitted. Ignored when ``is_obb=True``. max_image_area_percentage: Maximum detection area as a fraction of the image area; larger detections are omitted. Ignored when ``is_obb=True``. approximation_percentage: Fraction of polygon points removed during contour approximation when saving mask annotations. Ignored when ``is_obb=True``. is_obb: If ``True``, serializes oriented bounding-box corners from ``detections.data['xyxyxyxy']`` as a 9-token YOLO OBB line ``class_id x1 y1 x2 y2 x3 y3 x4 y4``. Mask data is ignored. Returns: A list of YOLO annotation strings, one per detection (or one per polygon for instance-segmentation annotations). Raises: ValueError: If any detection has ``class_id=None`` or a non-integer ``class_id``. ValueError: If ``is_obb=True`` and any non-empty detection is missing ``'xyxyxyxy'`` in ``detections.data``. Examples: >>> import numpy as np >>> from supervision.detection.core import Detections >>> from supervision.dataset.formats.yolo import detections_to_yolo_annotations >>> detections = Detections( ... xyxy=np.array([[10, 10, 90, 90]], dtype=np.float32), ... class_id=np.array([0]), ... ) >>> detections_to_yolo_annotations(detections, image_shape=(100, 100, 3)) ['0 0.50000 0.50000 0.80000 0.80000'] rz`is_obb=True` requires `'z'` in `detections.data` with shape (N, 4, 2). Load OBB datasets via `DetectionDataset.from_yolo(..., is_obb=True)` or set `detections.data['z'']` (shape (N, 4, 2)) before exporting.zo`detections.mask` is ignored when `is_obb=True`; OBB annotations use corner coordinates from `detections.data['z']`.rrz*Class ID is required for YOLO annotations.z!?rzzRG}}& "==/*))A(BC88 )%' K   k *   5*C*C)A H $ /&w7,) +#  !!+. /)LkK   k *a0+b r"c $t|jdd|D]o\}}}t|j} t| } tj j || } t||j||||} t| | qy)aQSave dataset annotations in YOLO format. Args: dataset: The dataset whose annotations are saved. annotations_directory_path: Path to the directory where annotation ``.txt`` files are written; created automatically if absent. min_image_area_percentage: Minimum detection area as a fraction of the image area; smaller detections are omitted. Ignored when ``is_obb=True``. max_image_area_percentage: Maximum detection area as a fraction of the image area; larger detections are omitted. Ignored when ``is_obb=True``. approximation_percentage: Fraction of polygon points removed during contour approximation when saving mask annotations. Ignored when ``is_obb=True``. is_obb: If ``True``, writes oriented bounding-box annotations using the 9-token format ``class_id x1 y1 x2 y2 x3 y3 x4 y4``. Each non-empty detection must carry ``detections.data['xyxyxyxy']`` with shape ``(N, 4, 2)``. Examples: >>> from supervision.dataset.core import DetectionDataset >>> from supervision.dataset.formats.yolo import save_yolo_annotations >>> dataset = DetectionDataset(classes=["cat"], images={}, annotations={}) >>> save_yolo_annotations(dataset, "/tmp/labels") Tparentsexist_ok)ri)rrrrrrv)r<rAN) rmkdirrbrlrfrgrrrr) datasetrrrrrvrrrriyolo_annotations_nameyolo_annotations_pathr<s r save_yolo_annotationsrsD #$**4$*G)0E% E:*%** >* U " &(=! /! &?&?%=   U.CDEr"ct||d}t|jjddt ||y)N)ncrCTr)r\rA)r:rparentrr)rrr\s r save_data_yamlrs8g, 1D%%dT%B7r")r list[str]rQnpt.NDArray[np.float32])r'rrQr)r5zlist[npt.NDArray[np.number]]r0tuple[int, int]rQznpt.NDArray[np.bool_])r<rrQr4)rArGrQr)rirGrQrG)F) r<rr0rrur4rvr4rQr )FF) rrGrrGrrGrr4rvr4rQz2tuple[list[str], list[str], dict[str, Detections]]rN) rqznpt.NDArray[np.number]rprFrtuple[int, int, int]r/znpt.NDArray[np.number] | NonerQrG)gg?g?F)rr rrrrrrrrrvr4rQr)rrrrGrrrrrrrvr4rQNone)rrGrrrQr)- __future__rrfrpathlibrtypingrrnumpyr numpy.typingnptPILrsupervision.configrsupervision.dataset.utilsr supervision.detection.corer &supervision.detection.utils.convertersr r supervision.utils.filer rrrrsupervision.dataset.corerr!r(r-r6r>rcrlr}rrrrrr"r rs" %7D1S9  = * ;J  :<~ .J .J".J.J  .J  .Jj T-T- #T-T- T-  T- 8 T-v.2 3 33&3+ 3  32(+'*&* {{%{ %{ % { $ {  {{B(+'*&* 1E 1E #1E %1E % 1E $ 1E  1E 1Eh8r"