^jcddlZddlmZddlmZddlmZmZmZmZddl Z ddl mZ ddl mZddlmZmZddlmZddlmZmZmZdd lmZmZdd lmZmZerdd lmZe e!efZ"d e#e"d e#e!fdZ$d e#e"de#e!d e e%e%ffdZ&de#e!d e#e"fdZ'de#e"d e e%e#e"ffdZ(de#e"de)e%e%fd e jTe jVfdZ, d/de#e"de)e%e%fde-de-d ef dZ. d0dede%de%de/de/d e/d e)e#e"e%ffd!Z0d"e!d e e%e%ffd#Z1 d1d$e!d"e!d%e-de-d e)e#e!e#e!e e!efff d&Z2d'e e!efd e-fd(Z3 d2d)d*d+e!de/de/d e/d,e%d-e%d e)e%e%ffd.Z4y)3N)datetime)Path) TYPE_CHECKINGAnyUnioncast)COCO_RAW_SEGMENTATION)approximate_mask_with_polygonsmap_detections_class_id) Detections) mask_to_rlepolygon_to_mask rle_to_mask)contains_holescontains_multiple_segments)read_json_filesave_json_file)DetectionDatasetcoco_categoriesreturncJt|dDcgc]}|d c}Scc}w)Nc |dS)Nid)categorys k/var/www/ramen.bs-engineer-server.com/venv/lib/python3.12/site-packages/supervision/dataset/formats/coco.pyz,coco_categories_to_classes.. s Xd^)keyname)sorted)rrs rcoco_categories_to_classesr"s14ST     s target_classesc|Dcic] }|d|d }}t|Dcic] \}}||| c}}Scc}wcc}}w)Nr r enumerate)rr#rsource_class_to_indextarget_class_indextarget_class_names rbuild_coco_class_index_mappingr*$sl:I-5(4.( 6?~5N  1  1 /02DD  s9>classescXt|Dcgc]\}}|dz|ddc}}Scc}}w)aConvert a list of class names to COCO ``categories`` entries. Category ids are emitted 1-indexed to comply with the COCO specification and tools such as CVAT, which require ``category_id`` values to start at ``1``. The id assigned to the class at position ``class_index`` is ``class_index + 1``, keeping it consistent with the ``category_id`` written by [`detections_to_coco_annotations`](#detections_to_coco_annotations). Args: classes: Class names ordered by their internal (0-indexed) class id. Returns: A list of COCO category dictionaries with 1-indexed ``id`` values. Examples: ```python from supervision.dataset.formats.coco import classes_to_coco_categories classes_to_coco_categories(classes=["cat", "dog"]) # [ # {"id": 1, "name": "cat", "supercategory": "common-objects"}, # {"id": 2, "name": "dog", "supercategory": "common-objects"}, # ] ``` zcommon-objects)rr supercategoryr%)r+ class_index class_names rclasses_to_coco_categoriesr10s?@(1'9   $K /-   s&coco_annotationsc\i}|D]$}|d}||vrg||<||j|&|S)Nimage_id)append)r2 annotations annotationr4s r"group_coco_annotations_by_image_idr8TsO.0K&1 j) ; &$&K !H$$Z0 1 rimage_annotations resolution_whcr|d|d}}tj||ft}g}|D]n}|jd}|s |j |j 7|jddr |j t |d|it|ts |j |j t|dtr|n|g}|j } |D]} tjtj| tjd} | jdk(r+tjd |jd d d v| t| |j!tz} |j | qtj|tS)Nr-rdtype segmentationiscrowdcounts)rler:)zISkipping empty polygon while loading COCO segmentation for annotation id=r.rC stacklevel)polygonr:)npzerosboolgetr5copyr isinstancelistreshapeasarrayint32sizewarningswarnrastype) r9r:heightwidth empty_maskmasksimage_annotationr>polygons object_maskrG polygon_arrays rcoco_annotations_to_masksr^`s"!$mA&6EF(*&%(MJ E-%"'++N;  LL* +     1 - LL X 6mT  ,- LL* + #-l1ot#D<<. oo'  G35:: 7"((3W4M!!Q& %%5%9%9$%?$@C   ?%]fTl K "  [!K%"N ::e4 ((r with_masks use_iscrowdc |stjS|Dcgc]}|d }}|Dcgc]}|d }}tj|tj}|ddddfxx|ddddfz cc<i}|rc|Dcgc]}|d } }|Dcgc]}|d  } }t tj| t tj| t }|rt|| } nWd} tjt|t} t|D]\} }|jd g| | <| |t<ttj|t || | Scc}wcc}wcc}wcc}w)aConvert COCO annotation dicts for a single image into a `Detections` object. .. warning:: The returned ``Detections.class_id`` contains **raw COCO** ``category_id`` values, not the final 0-indexed internal class ids. Callers **must** pass the result through :func:`map_detections_class_id` with the appropriate ``source_to_target_mapping`` (built by :func:`build_coco_class_index_mapping`) before the ``class_id`` values are meaningful. Skipping the remap step yields 1-based ids in a field that the rest of supervision treats as 0-based. Args: image_annotations: List of COCO annotation dicts for one image. resolution_wh: ``(width, height)`` of the image, used for mask decoding. with_masks: Whether to decode segmentation fields into binary masks. use_iscrowd: When ``True``, store ``iscrowd`` and ``area`` in ``Detections.data``. Returns: Detections with ``class_id`` set to raw COCO ``category_id`` values. Call :func:`map_detections_class_id` on the result before use. When ``with_masks=False``, ``detections.data[COCO_RAW_SEGMENTATION]`` is populated as an object array (shape ``(N,)``) holding the raw polygon list or RLE dict per annotation; consumed by :func:`detections_to_coco_annotations` for a coordinate-preserving round-trip. category_idbboxr<NrCrr?area)r?re)r9r:r>)class_idxyxymaskdata)r emptyrHrPfloat32dictintfloatr^lenobjectr&rKr )r9r:r_r`rZ class_ids xyxy_listrgrir?rerhraw_segsk_anns rcoco_annotations_to_detectionsrvs@ !!AR,<'ICTT.>!&)TIT$&JJy $KDAaCLDAaCL LACD@Q ,< Y '  BSS-= (SSJJwc2DPU9V (/} -/HH ! "&- !!23 7GAt((>26HQK 7&. "# IS14d AU  Ts E2 E7 E< F detectionsr4 annotation_idmin_image_area_percentagemax_image_area_percentageapproximation_percentagec g}|D]p\}}} } } } | td|d|dz |d|dz } } g}|ttjtj |}d| vr0t t j| dj}n#t t|xs t|}|r=ttt t|d t|jddd }nt|||| }|r'|Dcgc]}t|j}}nt!j"d |ddnt t j| j%ddj}| j%t&}|at)|rVt+|t,r|}nCt+|tr(|r&t+|dtt.fs t|g}n t|}t1t j| j%d| | zj}||t | dz|d|d| | g|||d}|j3||dz }s||fScc}w)a Convert `Detections` to COCO ``annotations`` entries. The internal 0-indexed ``Detections.class_id`` is serialized as a 1-indexed COCO ``category_id`` (``category_id = class_id + 1``). This complies with the COCO specification and tools such as CVAT, and stays consistent with the ids emitted by [`classes_to_coco_categories`](#classes_to_coco_categories), so a detection with internal ``class_id=k`` maps to ``category_id=k + 1``. Args: detections: The detections to convert. ``class_id`` must not be ``None``. image_id: COCO ``image_id`` shared by every produced annotation. annotation_id: First annotation id to assign; incremented per detection. min_image_area_percentage: Lower bound on detection area / image area, used only when approximating masks with polygons. max_image_area_percentage: Upper bound on detection area / image area, used only when approximating masks with polygons. approximation_percentage: Polygon-simplification ratio in ``[0, 1)``. Returns: A ``(coco_annotations, next_annotation_id)`` tuple, where ``next_annotation_id`` is one greater than the last id assigned. Raises: ValueError: If any detection has ``class_id`` equal to ``None``. Note: For ``iscrowd=0`` annotations, ``segmentation`` is a ``list[list[float]]`` where each inner list encodes one polygon part as flat ``[x1, y1, x2, y2, ...]`` coordinates. A single object with *N* disjoint parts produces *N* inner lists. When ``iscrowd`` is not in ``detections.data``, masks with holes or multiple disjoint segments are auto-encoded as RLE (``iscrowd=1``); simple single-region masks use polygon format (``iscrowd=0``). Supply ``data={"iscrowd": np.array([0])}`` to force polygon output regardless of mask topology. Examples: ```python import numpy as np from supervision import Detections from supervision.dataset.formats.coco import ( detections_to_coco_annotations, ) detections = Detections( xyxy=np.array([[0, 0, 10, 10]], dtype=np.float32), class_id=np.array([0], dtype=int), ) annotations, next_id = detections_to_coco_annotations( detections=detections, image_id=1, annotation_id=1 ) annotations[0]["category_id"] # 1 ``` Nz1Detections must include class_id for COCO export.rCrr-r?)rhF)rh compressed)r@rR)rhryrzr{z5Skipping COCO polygon segmentation for annotation id=z1 because mask approximation returned no polygons.rEre)rr4rbrcrer>r?) ValueErrorrnptNDArrayrHbool_rmrPitemrrrNr shaper flattenrSrTrKr rJrMrltuplernr5)rwr4rxryrzr{r2rgrh_rfri box_width box_heightr> mask_boolr?r[praw_segrecoco_annotations rdetections_to_coco_annotationsrs~@(*,6G(dAxD  PQ Q $Q$q' 147T!W3D: GI  S[[2D9ID bjji9>>@A" 2B1yA "S ;I%#P!BQ0   :".G.G-E ?G#H!D$5#HL#HMM+_-00$% "**TXXi%;<AACDGhh45G"tG}gt,#*Lw-&wqzD%=A%)M?L#'=LBJJtxx J8N'OPUUWX x=1,!Wd1gy*=(  0 OGP ] **O$Is I=annotations_pathct|}t|d}t|d|}|jDcic]\}}|| c}}Scc}}w)ay Generates a mapping from sequential class indices to original COCO class ids. This function is essential when working with models that expect class ids to be zero-indexed and sequential (0 to 79), as opposed to the original COCO dataset where category ids are non-contiguous ranging from 1 to 90 but skipping some ids. Use Cases: - Evaluating models trained with COCO-style annotations where class ids are sequential ranging from 0 to 79. - Ensuring consistent class indexing across training, inference and evaluation, when using different tools or datasets with COCO format. - Reproducing results from models that assume sequential class ids (0 to 79). How it Works: - Reads the COCO annotation file in its original format (`annotations_path`). - Extracts and sorts all class names by their original COCO id (1 to 90). - Builds a mapping from COCO class ids (not sequential with skipped ids) to new class ids (sequential ranging from 0 to 79). - Returns a dictionary mapping: `{new_class_id: original_COCO_class_id}`. Args: annotations_path: Path to COCO JSON annotations file (e.g., `instances_val2017.json`). Returns: A mapping from new class id (sequential ranging from 0 to 79) to original COCO class id (1 to 90 with skipped ids). categoriesrrr#)rr"r*items)r coco_datar+ class_mappingrtvs rget_coco_class_index_mappingresY>/0I(<9PQG2!,/M+002 3TQAqD 33 3s Aimages_directory_path force_masksc 0t|}t|d}t|d|}|d}t|d}g} i} t |j } |D]} | d| d | d }}} |j | d g}tt |t | z } t |j }|| k(rtd | d| d| |jvrtd | d|d| d|jrtd | d|d|xstd|D}t|||f||}t||}| j||| |< || | fS#ttf$r}td | d ||d}~wwxYw)a Load COCO annotations and convert them to `Detections`. If `force_masks` is `False`, masks are still loaded for images whose annotations include a `segmentation` field. This keeps mask handling consistent with other dataset loaders that infer masks from annotation content. Args: images_directory_path: Path to the image directory. annotations_path: Path to COCO JSON annotations. force_masks: If `True`, always attempt to load masks. use_iscrowd: If `True`, include `iscrowd` and `area` in detection data. Returns: A tuple of `(classes, image_paths, annotations)`. Raises: ValueError: If any annotation's ``file_name`` resolves to the images directory itself, to a path outside the images directory (e.g. via ``../`` traversal or an absolute path), or to a subdirectory instead of a regular image file. Note: Each annotation's ``file_name`` is validated against ``images_directory_path`` before loading. Annotations that reference paths outside the directory are rejected to prevent path-traversal attacks when loading user-supplied annotation files. Symlinked images pointing outside the resolved images directory are also rejected.  file_pathrrrimagesr6)r2 file_namerWrVrz COCO annotation refers to image z", which produces an invalid path: Nz1, which resolves to the images directory itself (z$). Expected a path to an image file.z, which resolves to u" — outside the images directory rDz, which resolves to directory z#. Expected a path to an image file.c32K|]}t|yw)N)_with_seg_mask).0r7s r z(load_coco_annotations..s( +5N: &( s)r9r:r_r`)source_to_target_mappingrw)rr"r*r8rresolverKstrOSErrorrparentsis_diranyrvr r5)rrrr`rr+class_index_mapping coco_imagescoco_annotations_groupsrr6images_directory_resolved coco_image image_name image_width image_heightr9 image_pathresolved_image_pathexcr_r7s rload_coco_annotationsrsZF)9:I(<9PQG8!,/H%K@"=1FK $%: ; C C E!3- { # w  x ".K 477 48H"M34tJ7GGH  "&z"2":":"<  "; ;2:.A-./  %,?,G,G G2:.A23467q:   % % '2:.A))<(=>))  ! C( 9J( % 4/& 5!#  -%8!  j!", Jg3-j FK ''W$ 2:.A--0E3  s(E//F>FFr7c6t|jdS)Nr>)rJrK)r7s rrrs  ~. //rdatasetrannotation_pathstarting_image_idstarting_annotation_idc X|dks|dkrtd|d|t|jjddddddg}g}g} t |j } ||} } |D]\} }}|j \}}}t| jt| j}| d|||tjjd d }| j|t|| | ||| \}} |j|| dz } i|| | |d }t||| | fS)ul Save a DetectionDataset to a COCO-format ``annotations.json`` file. Args: dataset: The DetectionDataset to write. annotation_path: Output path for the COCO ``annotations.json``. min_image_area_percentage: Lower bound on detection area / image area; used only for segmentation datasets. max_image_area_percentage: Upper bound on detection area / image area; used only for segmentation datasets. approximation_percentage: Polygon-simplification ratio in ``[0, 1)``; used only for segmentation datasets. starting_image_id: First image id to assign in the exported file. Defaults to ``1``. Override when exporting multiple splits into a coordinated COCO collection so ids remain unique across the set. starting_annotation_id: First annotation id to assign in the exported file. Defaults to ``1``. Override for the same multi-split reason as ``starting_image_id``. Returns: A ``(next_image_id, next_annotation_id)`` tuple. The returned values are one greater than the highest ids written, so they can be fed directly back into ``starting_image_id`` and ``starting_annotation_id`` when exporting another split into a coordinated COCO collection (see ``DetectionDataset.as_coco`` for the chaining pattern). When the dataset is empty the starting ids are returned unchanged. .. note:: This function ensures globally unique integer ``id`` values across splits. It does **not** ensure unique ``file_name`` values — the ``file_name`` field is set to the bare image basename, so splits that share filenames (e.g. ``000001.jpg`` in both train and valid) will have duplicate ``file_name`` values when their COCO files are merged. Use distinct output directories or rename images before merging if downstream tools require unique ``file_name`` keys. Example: ```python import supervision as sv from supervision.dataset.formats.coco import save_coco_annotations ds = sv.DetectionDataset.from_yolo( images_directory_path="train/images", annotations_directory_path="train/labels", data_yaml_path="data.yaml", ) next_img_id, next_ann_id = save_coco_annotations( dataset=ds, annotation_path="out/train/annotations.json" ) # next_img_id and next_ann_id are the first unused ids — pass them # to the next split to keep ids globally unique across files. ``` r-ztstarting_image_id and starting_annotation_id must be >= 1 (COCO spec requires 1-indexed ids); got starting_image_id=z, starting_annotation_id=T)rexist_okz,https://creativecommons.org/licenses/by/4.0/z CC BY 4.0)rurlr )r+z%m/%d/%Y,%H:%M:%S)rlicenserrVrW date_captured)rwr4rxryrzr{)infolicensesrrr6r)rrparentmkdirr1r+rstemsuffixrnowstrftimer5rextendr)rrryrzr{rrrr2rrr4rxrimager7rrrrrrannotation_dicts rsave_coco_annotationsrsz1 6 : &$&&@)?(A C     &&td&CA HK0IO/1GmH)0% E:',{{$ k1Z(--.tJ/?/F/F.GH #" %\\^445HI   :&)G!'&?&?%= * & 0A /4%' O?o> ] ""r)T)?g?)FT)rrrr-r-)5rSrpathlibrtypingrrrrnumpyrH numpy.typingrsupervision.configr supervision.dataset.utilsr r supervision.detection.corer &supervision.detection.utils.convertersr rr!supervision.detection.utils.masksrrsupervision.utils.filerrsupervision.dataset.corerrlrCocoDictrNr"rmr*r1r8rrrr^rJrvrnrrrrrrrrrs2242 YA9 S>X49 (^ 59#Y  #s(^ !S !d8n!H 8n  #tH~  .)H~.)6;CHo.)[[.)j EH~Ec?EE E  EX(+'*&* I+I+I+I+ % I+ % I+ $ I+ 4>3 I+X$43$44S>$4T h(h(h(h( h(  49d3ic:o!6 67 h(V0tCH~0$0(+'*&)"#r# r#r# %r# % r# $ r#  r# r# 38_r#r