§
    9²¸jˆq  ã            
       ó&  — d Z ddlZddlmZ ddlZddlmZ ddlm	Z	 ddl
mZ ddlmZ dd	lmZmZmZ dd
lmZmZmZmZmZmZ ddlmZmZ ddlmZmZmZ  G d„ ded¬¦  «        Z 	 	 d&ddde!e"e"f         dz  de"dz  fd„Z#de!e$e"f         de%e"e"f         fd„Z&de'ej(                 de'e'e"                  de%e'ej(                 e'e'e"                  f         fd„Z)d„ Z*d'd„Z+	 	 	 d(d e,d!e,d"e%e"e"f         dz  fd#„Z-e G d$„ d%e	¦  «        ¦   «         Z.d%gZ/dS ))zImage processor class for EoMT.é    N)ÚUnion)Ú
functionalé   )ÚTorchvisionBackend)Ú'SemanticSegmentationPostProcessorOutput)ÚBatchFeature)Úget_size_with_aspect_ratioÚgroup_images_by_shapeÚreorder_images)ÚIMAGENET_DEFAULT_MEANÚIMAGENET_DEFAULT_STDÚChannelDimensionÚ
ImageInputÚPILImageResamplingÚSizeDict)ÚImagesKwargsÚUnpack)Ú
TensorTypeÚauto_docstringÚfilter_out_non_signature_kwargsc                   ó.   — e Zd ZU dZeed<   edz  ed<   dS )ÚEomtImageProcessorKwargsa{  
    do_split_image (`bool`, *optional*, defaults to `self.do_split_image`):
        Whether to split the input images into overlapping patches for semantic segmentation. If set to `True`, the
        input images will be split into patches of size `size["shortest_edge"]` with an overlap between patches.
        Otherwise, the input images will be padded to the target size.
    ignore_index (`int`, *optional*, defaults to `self.ignore_index`):
        Label to be assigned to background pixels in segmentation maps. If provided, segmentation map pixels
        denoted with 0 (background) will be replaced with `ignore_index`.
    Údo_split_imageNÚignore_index)Ú__name__Ú
__module__Ú__qualname__Ú__doc__ÚboolÚ__annotations__Úint© ó    ú~/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/transformers/models/eomt/image_processing_eomt.pyr   r   *   s:   € € € € € € ðð ð ÐÐÑØ˜‘*ÐÐÑÐÐr#   r   F)ÚtotalÚsegmentation_mapútorch.TensorÚinstance_id_to_semantic_idr   c                 ój  ‡ — |�t          j        ‰ dk    |‰ dz
  ¦  «        Š t          j        ‰ ¦  «        }|�|||k             }ˆ fd„|D ¦   «         }|rt          j        |d¬¦  «        }n$t          j        dg‰ j        ¢R ‰ j        ¬¦  «        }|�{t          j        |j        d         ‰ j        ¬¦  «        }t          |¦  «        D ]D\  }}||�|                     ¦   «         dz   n|                     ¦   «                  }|�|dz
  n|||<   ŒEn|}| 	                    ¦   «         | 
                    ¦   «         fS )Nr   é   c                 ó   •— g | ]}‰|k    ‘Œ	S r"   r"   )Ú.0Úir&   s     €r$   ú
<listcomp>zAconvert_segmentation_map_to_binary_masks_fast.<locals>.<listcomp>G   s   ø€ Ð@Ð@Ð@°Ð%¨Ò*Ð@Ð@Ð@r#   ©Údim©Údevice)ÚtorchÚwhereÚuniqueÚstackÚzerosÚshaper2   Ú	enumerateÚitemÚfloatÚlong)	r&   r(   r   Ú
all_labelsÚbinary_masksÚlabelsr-   ÚlabelÚclass_ids	   `        r$   Ú-convert_segmentation_map_to_binary_masks_fastrB   :   sg  ø€ ð
 ÐÝ œ;Ð'7¸1Ò'<¸lÐL\Ð_`ÑL`ÑaÔaÐå”Ð.Ñ/Ô/€JàÐØ 
¨lÒ :Ô;ˆ
à@Ð@Ð@Ð@°ZÐ@Ñ@Ô@€LØð aÝ”{ <°QÐ7Ñ7Ô7ˆˆå”{ AÐ#?Ð(8Ô(>Ð#?Ð#?ÐHXÔH_Ð`Ñ`Ô`ˆð "Ð-Ý”˜ZÔ-¨aÔ0Ð9IÔ9PÐQÑQÔQˆå! *Ñ-Ô-ð 	Oð 	O‰HˆAˆuØ1À|ÐG_°5·:²:±<´<À!Ñ3CÐ3CÐej×eoÒeoÑeqÔeqÔsˆHØ(4Ð(@˜ 1™˜ÀhˆF�1‰IˆIð	Oð ˆØ×ÒÑÔ §¢¡¤Ð.Ð.r#   Ú	size_dictÚreturnc                 ó.   — | d         }| d         p|}||fS )z.Returns the height and width from a size dict.Úshortest_edgeÚlongest_edger"   )rC   Útarget_heightÚtarget_widths      r$   Úget_target_sizerJ   Y   s&   € à˜oÔ.€MØ˜^Ô,Ð=°€Là˜,Ð&Ð&r#   ÚpatchesÚoffsetsc                 ó¾   — t          t          || ¦  «        ¦  «        }|                     d„ ¬¦  «         t          |Ž \  }}t          |¦  «        t          |¦  «        fS )z@Sorts patches and offsets according to the original image index.c                 ó   — | d         d         S )Nr   r"   )Úxs    r$   ú<lambda>z-reorder_patches_and_offsets.<locals>.<lambda>g   s   €   !¤ Q¤€ r#   )Úkey)ÚlistÚzipÚsort)rK   rL   ÚcombinedÚsorted_offsetsÚsorted_patchess        r$   Úreorder_patches_and_offsetsrX   a   s\   € õ
 •C˜ Ñ)Ô)Ñ*Ô*€HØ‡M‚MÐ'Ð'€MÑ(Ô(Ð(Ý%(¨( ^Ñ"€N�Nå�ÑÔ¥ nÑ!5Ô!5Ð5Ð5r#   c                 óæ   — | j         d         |j         d         cxk    r|j         d         k    sn t          d¦  «        ‚|                     |¦  «        ||k    z  }| |         ||         ||         fS )a	  
    Binarize the given masks using `object_mask_threshold`, it returns the associated values of `masks`, `scores` and
    `labels`.

    Args:
        masks (`torch.Tensor`):
            A tensor of shape `(num_queries, height, width)`.
        scores (`torch.Tensor`):
            A tensor of shape `(num_queries)`.
        labels (`torch.Tensor`):
            A tensor of shape `(num_queries)`.
        object_mask_threshold (`float`):
            A number between 0 and 1 used to binarize the masks.
    Raises:
        `ValueError`: Raised when the first dimension doesn't match in all input tensors.
    Returns:
        `tuple[`torch.Tensor`, `torch.Tensor`, `torch.Tensor`]`: The `masks`, `scores` and `labels` without the region
        < `object_mask_threshold`.
    r   z1mask, scores and labels must have the same shape!)r8   Ú
ValueErrorÚne)ÚmasksÚscoresr?   Úobject_mask_thresholdÚ
num_labelsÚto_keeps         r$   Úremove_low_and_no_objectsra   m   sy   € ð( ŒK˜ŒN˜fœl¨1œoÐ@Ð@Ò@Ð@°´¸a´Ò@Ð@Ð@Ð@ÝÐLÑMÔMÐMà�iŠi˜
Ñ#Ô# vÐ0EÒ'EÑF€Gà�Œ>˜6 'œ?¨F°7¬OÐ;Ð;r#   ç      à?çš™™™™™é?c                 ó  — | |k    }|                      ¦   «         }||         |k    }|                      ¦   «         }||z  }	|	                      ¦   «         }
|dk    o|dk    o|
dk    }|r||z  }|                     ¦   «         |k    sd}||	fS )Nr   F)Úsumr:   )Úmask_labelsÚ
mask_probsÚkÚmask_thresholdÚoverlap_mask_area_thresholdÚmask_kÚmask_k_areaÚoriginal_maskÚoriginal_areaÚ
final_maskÚfinal_mask_areaÚmask_existsÚ
area_ratios                r$   Úcheck_segment_validityrs   ‰   s¨   € à˜AÒ€FØ—*’*‘,”,€Kð ˜q”M ^Ò3€MØ!×%Ò%Ñ'Ô'€Mà˜-Ñ'€JØ —n’nÑ&Ô&€Oà ’/ÐO m°aÒ&7ÐO¸OÈaÒ<O€Kàð  Ø  =Ñ0ˆ
Ø�ŠÑ Ô Ð#>Ò>Ð>ØˆKà˜
Ð"Ð"r#   ri   rj   Útarget_sizec                 ó¢  — |€| j         d         n|d         }|€| j         d         n|d         }t          j        ||ft          j        | j        ¬¦  «        dz
  }	g }
|                      ¦   «         } |d d …d d f         | z                       d¦  «        }d}i }t          |j         d         ¦  «        D ]›}||                              ¦   «         }t          || |||¦  «        \  }}|sŒ5|r||v r||v r||         |	|<   ŒK|||<   ||	|<   t          ||                              ¦   «         d¦  «        }|
                     |||dœ¦  «         |dz  }Œœ|	|
fS )Nr*   r   é   )Údtyper2   é   ©ÚidÚlabel_idÚscore)r8   r3   r7   r<   r2   ÚsigmoidÚargmaxÚranger:   rs   ÚroundÚappend)rg   Úpred_scoresÚpred_labelsÚstuff_classesri   rj   rt   ÚheightÚwidthÚsegmentationÚsegmentsrf   Úcurrent_segment_idÚstuff_memory_listrh   Ú
pred_classrq   ro   Úsegment_scores                      r$   Úcompute_segmentsr�   Ÿ   s¸  € ð %0Ð$7ˆZÔ˜aÔ Ð ¸[È¼^€FØ#.Ð#6ˆJÔ˜QÔÐ¸KÈ¼N€Eå”; ¨˜µe´jÈÔIZÐ[Ñ[Ô[Ð^_Ñ_€LØ€Hð ×#Ò#Ñ%Ô%€JØ˜q˜q˜q $¨˜}Ô-°
Ñ:×BÒBÀ1ÑEÔE€Kð ÐØ(*Ðå�;Ô$ QÔ'Ñ(Ô(ð  ð  ˆØ  ”^×(Ò(Ñ*Ô*ˆ
õ #9Ø˜ Q¨Ð8Sñ#
ô #
Ñˆ�Zð ð 	Øàð 	C˜Z¨=Ð8Ð8ØÐ.Ð.Ð.Ø+<¸ZÔ+H�˜ZÑ(Øà0BÐ! *Ñ-à#5ˆ�ZÑ Ý˜k¨!œn×1Ò1Ñ3Ô3°QÑ7Ô7ˆØ�Šà(Ø&Ø&ðð ñ	
ô 	
ð 	
ð 	˜aÑÐÐØ˜Ð!Ð!r#   c                   óT  ‡ — e Zd ZeZej        ZeZ	e
ZdddœZdZdZdZdZdZdZdZdee         fˆ fd„Zdej        d	ed
ee         deeef         fd„Zdej        d	edej        fd„Ze	 	 d4dedeej                 dz  de eef         dz  dee         de!f
ˆ fd„¦   «         Z"	 d5dededz  de eef         dz  de#de$de%e&z  dz  de'e%df         dz  dee         de!fd„Z(ded         de#d	eddde#de)de#de#de#d e)ee)         z  dz  d!e)ee)         z  dz  d"e#dz  fd#„Z*d$ej        d%eeeeef                  d&eeeef                  d	e e%ef         deej                 f
d'„Z+d$ej        d&eeeef                  d	e e%ef         deej                 fd(„Z,	 	 d6d&eeeef                  d	e e%ef         dz  d)e#dd*fd+„Z-	 	 	 	 	 d7d&eeeef                  d.e)d/e)d0e)d1ee         dz  d	e e%ef         dz  fd2„Z. e/¦   «         	 	 d8d&eeeef                  d.e)d	e e%ef         dz  fd3„¦   «         Z0ˆ xZ1S )9ÚEomtImageProcessori€  ©rF   rG   FTNÚkwargsc                 ó:   •—  t          ¦   «         j        di |¤Ž d S )Nr"   )ÚsuperÚ__init__)Úselfr‘   Ú	__class__s     €r$   r”   zEomtImageProcessor.__init__ä   s&   ø€ Ø�‰ŒÔÐ"Ð"˜6Ð"Ð"Ð"Ð"Ð"r#   ÚimagesÚsizeÚimage_indicesrD   c                 ó$  — g g }}|j         \  }}}}|j        }	t          ||¦  «        }
t          j        |
|	z  ¦  «        }||	z  |
z
  }|dk    r||dz
  z  nd}t          |¦  «        D ]§}t          ||	|z
  z  ¦  «        }||	z   }||k    r|dd…dd…||…dd…f         }n|dd…dd…dd…||…f         }t          t          j	        |d¬¦  «        ¦  «        D ]8\  }}| 
                    |¦  «         | 
                    ||         ||g¦  «         Œ9Œ¨||fS )zCSlices an image into overlapping patches for semantic segmentation.r*   r   Nr/   )r8   rF   ÚmaxÚmathÚceilr   r!   r9   r3   Úunbindr�   )r•   r—   r˜   r™   rK   Úpatch_offsetsÚ_r…   r†   Ú
patch_sizeÚlonger_sideÚnum_patchesÚtotal_overlapÚoverlap_per_patchr-   ÚstartÚendÚbatch_patchÚ	batch_idxÚsingles                       r$   Ú_split_imagezEomtImageProcessor._split_imageç   su  € ð "$ R�ˆà$œlÑˆˆ1ˆf�eØÔ'ˆ
å˜& %Ñ(Ô(ˆÝ”i ¨jÑ 8Ñ9Ô9ˆØ# jÑ0°;Ñ>ˆØALÈqÂÀ˜M¨[¸1©_Ñ=Ð=ÐVWÐå�{Ñ#Ô#ð 	Mð 	MˆAÝ˜˜ZÐ*;Ñ;Ñ<Ñ=Ô=ˆEØ˜*Ñ$ˆCà˜Š~ˆ~Ø$ Q Q Q¨¨¨¨5°¨9°a°a°aÐ%7Ô8��à$ Q Q Q¨¨¨¨1¨1¨1¨e°C¨iÐ%7Ô8�å%.­u¬|¸KÈQÐ/OÑ/OÔ/OÑ%PÔ%Pð Mð MÑ!�	˜6Ø—’˜vÑ&Ô&Ð&Ø×$Ò$ m°IÔ&>ÀÀsÐ%KÑLÔLÐLÐLðMð ˜Ð%Ð%r#   c                 ó  — |j         \  }}}}t          |j        |j        p|j        dœ¦  «        \  }}t	          d||z
  ¦  «        }t	          d||z
  ¦  «        }	d|	d|f}
t
          j        j                             ||
dd¬¦  «        }|S )z5Pads the image to the target size using zero padding.r�   r   Úconstantg        )ÚmodeÚvalue)	r8   rJ   rF   rG   r›   r3   Únnr   Úpad)r•   r—   r˜   r    r…   r†   rH   rI   Úpad_hÚpad_wÚpaddingÚpadded_imagess               r$   Ú_padzEomtImageProcessor._pad  s    € à$œlÑˆˆ1ˆf�eå&5Ø"Ô0À$ÔBSÐBiÐW[ÔWiÐjÐjñ'
ô '
Ñ#ˆ�|õ �A�} vÑ-Ñ.Ô.ˆÝ�A�| eÑ+Ñ,Ô,ˆØ�e˜Q Ð&ˆåœÔ+×/Ò/°¸ÀjÐX[Ð/Ñ\Ô\ˆØÐr#   Úsegmentation_mapsr(   c                 ó<   •—  t          ¦   «         j        |||fi |¤ŽS )a'  
        segmentation_maps (`ImageInput`, *optional*):
            The segmentation maps to preprocess for corresponding images.
        instance_id_to_semantic_id (`list[dict[int, int]]` or `dict[int, int]`, *optional*):
            A mapping between object instance ids and class ids.
        )r“   Ú
preprocess)r•   r—   r·   r(   r‘   r–   s        €r$   r¹   zEomtImageProcessor.preprocess  s+   ø€ ð "�u‰wŒwÔ! &Ð*;Ð=WÐbÐbÐ[aÐbÐbÐbr#   Údo_convert_rgbÚinput_data_formatÚreturn_tensorsr2   ztorch.devicec                 ó  — |                       ||||¬¦  «        }|                     dd¦  «        }	|                     ¦   «         }
i } | j        |fi |
¤Ž\  }}||d<   ||d<   |�ú|                       |ddt          j        ¬¦  «        }|                     ¦   «         }|                     ddt          j        d	œ¦  «          | j        dd
|i|¤Ž\  }}d„ |D ¦   «         }g g }}t          |¦  «        D ]d\  }}t          |t          ¦  «        r	||         }n|}t          |||	¬¦  «        \  }}|                     |¦  «         |                     |¦  «         Œe||d<   ||d<   |rd„ |D ¦   «         |d<   t          ||g d¢¬¦  «        S )z/
        Preprocess image-like inputs.
        )r—   rº   r»   r2   r   NÚpixel_valuesrŸ   rv   F)r—   Úexpected_ndimsrº   r»   )Údo_normalizeÚ
do_rescaleÚresampler—   c                 ór   — g | ]4}|                      d ¦  «                             t          j        ¦  «        ‘Œ5S )r   )ÚsqueezeÚtor3   Úint64)r,   r&   s     r$   r.   zDEomtImageProcessor._preprocess_image_like_inputs.<locals>.<listcomp>N  sD   € ð +ð +ð +Ø@PÐ ×(Ò(¨Ñ+Ô+×.Ò.­u¬{Ñ;Ô;ð+ð +ð +r#   )r   rf   Úclass_labelsc                 ó6   — g | ]}t          j        |¦  «        ‘ŒS r"   )r3   Útensor)r,   rL   s     r$   r.   zDEomtImageProcessor._preprocess_image_like_inputs.<locals>.<listcomp>g  s"   € Ð$XÐ$XÐ$X¸w¥U¤\°'Ñ%:Ô%:Ð$XÐ$XÐ$Xr#   )rŸ   rf   rÇ   )ÚdataÚtensor_typeÚskip_tensor_conversionr"   )Ú_prepare_image_like_inputsÚpopÚcopyÚ_preprocessr   ÚFIRSTÚupdater   ÚNEARESTr9   Ú
isinstancerR   rB   r�   r   )r•   r—   r·   r(   rº   r»   r¼   r2   r‘   r   Úimages_kwargsrÊ   Úprocessed_imagesrŸ   Úprocessed_segmentation_mapsÚsegmentation_maps_kwargsr    rf   rÇ   Úidxr&   Úinstance_idr\   Úclassess                           r$   Ú_preprocess_image_like_inputsz0EomtImageProcessor._preprocess_image_like_inputs!  s;  € ð ×0Ò0Ø¨.ÐL]Ðflð 1ñ 
ô 
ˆð —z’z .°$Ñ7Ô7ˆØŸš™œˆØˆØ*:¨$Ô*:¸6Ð*SÐ*SÀ]Ð*SÐ*SÑ'Ð˜-Ø/ˆˆ^ÑØ -ˆˆ_ÑàÐ(Ø*.×*IÒ*IØ(Ø Ø$Ý"2Ô"8ð	 +Jñ +ô +Ð'ð (.§{¢{¡}¤}Ð$Ø$×+Ò+à$)Ø"'å 2Ô :ð	ð ñô ð ð .>¨TÔ-=ð .ð .Ø2ð.Ø6Nð.ð .Ñ*Ð'¨ð+ð +ØToð+ñ +ô +Ð'ð )+¨B˜ˆKÝ)2Ð3NÑ)OÔ)Oð -ð -Ñ%�Ð%ÝÐ8½$Ñ?Ô?ð =Ø"<¸SÔ"A�K�Kà"<�Kå!NØ$ØØ!-ð"ñ "ô "‘��wð ×"Ò" 5Ñ)Ô)Ð)Ø×#Ò# GÑ,Ô,Ð,Ð,ð #.ˆD�ÑØ#/ˆD�Ñ àð 	YØ$XÐ$XÈ-Ð$XÑ$XÔ$XˆD�Ñ!åØØ&Ø#SÐ#SÐ#Sð
ñ 
ô 
ð 	
r#   r'   Ú	do_resizerÂ   z7PILImageResampling | tvF.InterpolationMode | int | NonerÁ   Úrescale_factorrÀ   r   Údo_padÚ
image_meanÚ	image_stdÚdisable_groupingc           	      ót  ‡ ‡‡— g }t          ||¬¦  «        \  }}i }|                     ¦   «         D ]$\  Š}|r‰                      |‰|¬¦  «        }||‰<   Œ%t          ||¦  «        }|r©t          ||¬¦  «        \  }}g g }}|                     ¦   «         D ]i\  Š}ˆfd„|                     ¦   «         D ¦   «         }‰                      |‰|¦  «        \  }}|                     |¦  «         |                     |¦  «         Œjt          ||¦  «        \  }}|	rEt          ||¬¦  «        \  }}ˆ ˆfd„|                     ¦   «         D ¦   «         }t          ||¦  «        }t          ||¬¦  «        \  }}i }|                     ¦   «         D ]$\  Š}‰                      |||||
|¦  «        }||‰<   Œ%t          ||¦  «        }||fS )z4Preprocesses the input images and masks if provided.)râ   )Úimager˜   rÂ   c                 ó,   •— g | ]\  }\  }}|‰k    ¯|‘ŒS r"   r"   )r,   Úoriginal_idxÚ	img_shaper    r8   s       €r$   r.   z2EomtImageProcessor._preprocess.<locals>.<listcomp>Ž  s4   ø€ ð $ð $ð $Ù%A \±>°I¸qÐenÐrwÒewÐew�LÐewÐewÐewr#   c                 óD   •— i | ]\  }}|‰                      |‰¦  «        “ŒS r"   )r¶   )r,   r8   Ústacked_imagesr•   r˜   s      €€r$   ú
<dictcomp>z2EomtImageProcessor._preprocess.<locals>.<dictcomp>˜  s:   ø€ ð ð ð Ù;P¸5À.��t—y’y °Ñ6Ô6ðð ð r#   )r
   ÚitemsÚresizer   r«   ÚextendrX   Úrescale_and_normalize)r•   r—   rÝ   r˜   rÂ   rÁ   rÞ   rÀ   r   rß   rà   rá   râ   r‘   rŸ   Úgrouped_imagesÚgrouped_images_indexÚresized_images_groupedré   rK   Úoriginal_indicesÚsplit_patchesrL   Úpadded_groupedÚprocessed_images_groupedrÖ   r8   s   `  `                      @r$   rÐ   zEomtImageProcessor._preprocesso  sc  øøø€ ð" ˆå/DÀVÐ^nÐ/oÑ/oÔ/oÑ,ˆÐ,Ø!#ÐØ%3×%9Ò%9Ñ%;Ô%;ð 	;ð 	;Ñ!ˆE�>Øð aØ!%§¢°>ÈÐW_ Ñ!`Ô!`�Ø,:Ð" 5Ñ)Ð)ÝÐ 6Ð8LÑMÔMˆàð 
	XÝ3HÈÐbrÐ3sÑ3sÔ3sÑ0ˆNÐ0Ø%'¨�]ˆGØ)7×)=Ò)=Ñ)?Ô)?ð .ð .Ñ%��~ð$ð $ð $ð $ØEY×E_ÒE_ÑEaÔEað$ñ $ô $Ð ð *.×):Ò):¸>È4ÐQaÑ)bÔ)bÑ&�˜wØ—’˜}Ñ-Ô-Ð-Ø×$Ò$ WÑ-Ô-Ð-Ð-Ý$?ÀÈÑ$WÔ$WÑ!ˆF�Màð 	JÝ3HÈÐbrÐ3sÑ3sÔ3sÑ0ˆNÐ0ðð ð ð ð ØTb×ThÒThÑTjÔTjðñ ô ˆNõ $ NÐ4HÑIÔIˆFå/DÀVÐ^nÐ/oÑ/oÔ/oÑ,ˆÐ,Ø#%Ð Ø%3×%9Ò%9Ñ%;Ô%;ð 	=ð 	=Ñ!ˆE�>Ø!×7Ò7Ø 
¨N¸LÈ*ÐV_ñô ˆNð /=Ð$ UÑ+Ð+Ý)Ð*BÐDXÑYÔYÐà Ð.Ð.r#   Úsegmentation_logitsrŸ   Útarget_sizesc                 óÖ  — |j         d         }g }g }|D ]„}t          ||d         |d         ¦  «        \  }	}
|                     t          j        ||	|
f|j        ¬¦  «        ¦  «         |                     t          j        ||	|
f|j        ¬¦  «        ¦  «         Œ…t          |¦  «        D ]°\  }\  }}}||         d         ||         d         k    rE||         dd…||…dd…fxx         ||         z  cc<   ||         dd…||…dd…fxx         dz  cc<   Œl||         dd…dd…||…fxx         ||         z  cc<   ||         dd…dd…||…fxx         dz  cc<   Œ±g }t          t          ||¦  «        ¦  «        D ]p\  }\  }}||                     d¬¦  «        z  }t          j	        j
                             |d         ||         d	d
¬¦  «        d         }|                     |¦  «         Œq|S )a¾  
        Reconstructs full-size semantic segmentation logits from patch predictions.

        Args:
            segmentation_logits (`torch.Tensor`):
                A tensor of shape `(num_patches, num_classes, patch_height, patch_width)` representing predicted logits
                for each image patch.
            patch_offsets (`list[tuple[int, int, int]]`):
                A list of tuples where each tuple contains:
                - `image_index` (int): Index of the original image this patch belongs to.
                - `start` (int): Start pixel index of the patch along the long dimension (height or width).
                - `end` (int): End pixel index of the patch along the long dimension.
            target_sizes (`list[tuple[int, int]]`):
                list of original (height, width) dimensions for each image before preprocessing.
            size (`dict[str, int]`):
                A size dict which was used to resize.
        r*   rF   rG   r1   r   N)Úmin©N.ÚbilinearF©r˜   r®   Úalign_corners)r8   r	   r�   r3   r7   r2   r9   rS   Úclampr°   r   Úinterpolate)r•   rö   rŸ   r÷   r˜   Únum_classesÚaggregated_logitsÚpatch_countsÚ
image_sizer…   r†   Ú	patch_idxÚ	image_idxÚpatch_startÚ	patch_endÚreconstructed_logitsrÙ   Ú	logit_sumÚcountÚaveraged_logitsÚresized_logitss                        r$   Úmerge_image_patchesz&EomtImageProcessor.merge_image_patches¨  s†  € ð0 *Ô/°Ô2ˆØÐØˆà&ð 	nð 	nˆJÝ6°zÀ4ÈÔCXÐZ^Ð_mÔZnÑoÔo‰MˆF�EØ×$Ò$¥U¤[°+¸vÀuÐ1MÐViÔVpÐ%qÑ%qÔ%qÑrÔrÐrØ×Ò¥¤¨[¸&À%Ð,HÐQdÔQkÐ lÑ lÔ lÑmÔmÐmÐmõ ?HÈÑ>VÔ>Vð 	Jð 	JÑ:ˆIÑ:˜	 ;°	Ø˜IÔ& qÔ)¨L¸Ô,CÀAÔ,FÒFÐFØ! )Ô,¨Q¨Q¨Q°¸IÐ0EÀqÀqÀqÐ-HÐIÐIÔIÐM`ÐajÔMkÑkÐIÐIÑIØ˜YÔ'¨¨¨¨;°yÐ+@À!À!À!Ð(CÐDÐDÔDÈÑIÐDÐDÑDÐDà! )Ô,¨Q¨Q¨Q°°°°;¸yÐ3HÐ-HÐIÐIÔIÐM`ÐajÔMkÑkÐIÐIÑIØ˜YÔ'¨¨¨¨1¨1¨1¨k¸)Ð.CÐ(CÐDÐDÔDÈÑIÐDÐDÑDÐDð  "ÐÝ'0µÐ5FÈÑ1UÔ1UÑ'VÔ'Vð 		8ð 		8Ñ#ˆCÑ#�)˜UØ'¨%¯+ª+¸!¨+Ñ*<Ô*<Ñ<ˆOÝ"œXÔ0×<Ò<Ø 	Ô*Ø! #Ô&ØØ#ð	 =ñ ô ð
 ôˆNð !×'Ò'¨Ñ7Ô7Ð7Ð7à#Ð#r#   c                 ó4  — g }t          |¦  «        D ]…\  }}t          ||d         |d         ¦  «        \  }}||         dd…d|…d|…f         }	t          j        j                             |	d         |dd¬¦  «        d         }
|                     |
¦  «         Œ†|S )	zJRestores panoptic segmentation logits to their original image resolutions.rF   rG   Nrú   rû   Frü   r   )r9   r	   r3   r°   r   rÿ   r�   )r•   rö   r÷   r˜   r  rÙ   Úoriginal_sizerH   rI   Úcropped_logitsÚupsampled_logitss              r$   Úunpad_imagezEomtImageProcessor.unpad_imageá  sÅ   € ð ˆå"+¨LÑ"9Ô"9ð 	4ð 	4ÑˆC�Ý*DØ˜t OÔ4°d¸>Ô6Jñ+ô +Ñ'ˆM˜<ð 1°Ô5°a°a°a¸¸-¸ÈÈ,ÈÐ6VÔWˆNÝ$œxÔ2×>Ò>Ø˜yÔ)°ÀJÐ^cð  ?ñ  ô  àô Ðð ×!Ò!Ð"2Ñ3Ô3Ð3Ð3ØÐr#   Úreturn_segmentation_scoreszBlist[torch.Tensor] | list[SemanticSegmentationPostProcessorOutput]c                 ó¸  — |�|n| j         }|j        }|j        }|j        }t	          |¦  «        }t
          j        j                             ||d¬¦  «        }| 	                    d¬¦  «        ddd…f         }	| 
                    ¦   «         }
t          j        d|	|
¦  «        }|r|                      ||||¦  «        }n„g }t          t          |¦  «        ¦  «        D ]e}t
          j        j                             ||                              d¬¦  «        ||         dd	¬
¦  «        }|                     |d         ¦  «         Œfd„ |D ¦   «         }|sd„ |D ¦   «         }|S )a@  
        Converts the output of [`EomtForUniversalSegmentation`] into semantic segmentation maps.

        Args:
            outputs ([`EomtForUniversalSegmentation`]):
                Raw outputs of the model.
            target_sizes (`list[tuple[int, int]]`):
                A list of tuples (`tuple[int, int]`) containing the target size (height, width) of each image in the
                batch.
            size (`dict[str, int]`, *optional*):
                The size to which the intermediate masks are interpolated. Defaults to `self.size`.
            return_segmentation_scores (`bool`, *optional*, defaults to `False`):
                Whether to return segmentation scores alongside the segmentation map. When `True`, each element of
                the returned list is a [`SemanticSegmentationPostProcessorOutput`] with fields `segmentation`
                (class IDs, shape `(height, width)`) and `segmentation_scores` (shape `(num_classes, height, width)`).

        Returns:
            `list[torch.Tensor]` or `list[SemanticSegmentationPostProcessorOutput]`: When
            `return_segmentation_scores=False` (default), a list of length `batch_size` where each item is a
            segmentation map of shape `(height, width)` with class IDs. When `return_segmentation_scores=True`,
            a list of [`SemanticSegmentationPostProcessorOutput`] with fields `segmentation` (class IDs, shape
            `(height, width)`) and `segmentation_scores` (shape `(num_classes, height, width)`). In both cases,
            `(height, width)` corresponds to the target size.
        Nrû   ©r˜   r®   éÿÿÿÿr/   .zbqc, bqhw -> bchwr   Frü   c                 ó\   — g | ])}t          |                     d ¬¦  «        |dœ¬¦  «        ‘Œ*S )r   r/   )r‡   Úsegmentation_scores)rÊ   )r   r~   )r,   Úlogits     r$   r.   zIEomtImageProcessor.post_process_semantic_segmentation.<locals>.<listcomp>6  sS   € ð !
ð !
ð !
ð õ 4Ø&+§l¢l°q lÑ&9Ô&9ÐRWÐXÐXðñ ô ð!
ð !
ð !
r#   c                 ó   — g | ]	}|j         ‘Œ
S r"   )r‡   )r,   r:   s     r$   r.   zIEomtImageProcessor.post_process_semantic_segmentation.<locals>.<listcomp>>  s   € Ð$YÐ$YÐ$Y¸4 TÔ%6Ð$YÐ$YÐ$Yr#   )r˜   Úmasks_queries_logitsÚclass_queries_logitsrŸ   rJ   r3   r°   r   rÿ   Úsoftmaxr}   Úeinsumr  r   ÚlenÚ	unsqueezer�   )r•   Úoutputsr÷   r˜   r  r  r  rŸ   Úoutput_sizeÚmasks_classesÚmasks_probsrö   Úoutput_logitsrÙ   r  Úsemantic_segmentations                   r$   Ú"post_process_semantic_segmentationz5EomtImageProcessor.post_process_semantic_segmentationö  s¦  € ð> Ð'ˆtˆt¨T¬Yˆà&Ô;ÐØ&Ô;ÐØÔ-ˆå% dÑ+Ô+ˆÝ$œxÔ2×>Ò>Ø ØØð  ?ñ  
ô  
Ðð -×4Ò4¸Ð4Ñ<Ô<¸SÀ#À2À#¸XÔFˆØ*×2Ò2Ñ4Ô4ˆå#œlÐ+>ÀÈ{Ñ[Ô[Ðàð 	8Ø ×4Ò4Ð5HÈ-ÐYeÐgkÑlÔlˆMˆMàˆMå�SÐ!4Ñ5Ô5Ñ6Ô6ð 8ð 8�Ý!&¤Ô!4×!@Ò!@Ø'¨Ô,×6Ò6¸1Ð6Ñ=Ô=Ø% cÔ*Ø#Ø"'ð	 "Añ "ô "�ð ×$Ò$ ^°AÔ%6Ñ7Ô7Ð7Ð7ð!
ð !
ð 'ð	!
ñ !
ô !
Ðð *ð 	ZØ$YÐ$YÐCXÐ$YÑ$YÔ$YÐ!à$Ð$r#   rc   rb   Ú	thresholdri   rj   r„   c                 ó  — |�|n| j         }|j        }|j        }	|	j        d         }
|	j        d         dz
  }t	          |¦  «        }t
          j        j                             ||d¬¦  «        }|  	                    |||¦  «        }|	 
                    d¬¦  «                             d¦  «        \  }}g }t          |
¦  «        D ]Å}t          ||         ||         ||         ||¦  «        \  }}}|j        d         dk    rN|�||         n|j        dd…         \  }}t          j        ||f¦  «        dz
  }|                     |g dœ¦  «         ŒŠt!          |||||||�||         nd¬	¦  «        \  }}|                     ||dœ¦  «         ŒÆ|S )
zIPost-processes model outputs into final panoptic segmentation prediction.Nr   r  r*   rû   r  r/   ©r‡   Úsegments_info)rg   r‚   rƒ   r„   ri   rj   rt   )r˜   r  r  r8   rJ   r3   r°   r   rÿ   r  r  r›   r   ra   r7   r�   r�   )r•   r!  r÷   r(  ri   rj   r„   r˜   r  r  Ú
batch_sizer_   r"  Úmask_probs_batchÚpred_scores_batchÚpred_labels_batchÚresultsr-   rg   r‚   rƒ   r…   r†   r‡   rˆ   s                            r$   Ú"post_process_panoptic_segmentationz5EomtImageProcessor.post_process_panoptic_segmentationB  så  € ð Ð'ˆtˆt¨T¬Yˆà&Ô;ÐØ&Ô;Ðà)Ô/°Ô2ˆ
Ø)Ô/°Ô3°aÑ7ˆ
å% dÑ+Ô+ˆÝ$œxÔ2×>Ò>Ø ØØð  ?ñ  
ô  
Ðð  ×+Ò+Ð,@À,ÐPTÑUÔUÐØ/C×/KÒ/KÐPRÐ/KÑ/SÔ/S×/WÒ/WÐXZÑ/[Ô/[Ñ,ÐÐ,àˆå�zÑ"Ô"ð 	Vð 	VˆAÝ3LØ  Ô#Ð%6°qÔ%9Ð;LÈQÔ;OÐQZÐ\fñ4ô 4Ñ0ˆJ˜ [ð
 Ô Ô" aÒ'Ð'Ø3?Ð3K ¨Q¤ ÐQ[ÔQaÐbcÐbdÐbdÔQe‘�˜Ý$œ{¨F°E¨?Ñ;Ô;¸aÑ?�Ø—’°ÈrÐRÐRÑSÔSÐSØå%5Ø%Ø'Ø'Ø+Ø-Ø,GØ/;Ð/G˜L¨œO˜OÈTð&ñ &ô &Ñ"ˆL˜(ð �NŠN¨LÈ8ÐTÐTÑUÔUÐUÐUØˆr#   c           
      óô  — |�|n| j         }|j        }|j        }t          |¦  «        }t          j        j                             ||d¬¦  «        }|                      |||¦  «        }|j	        }	|j
        d         }
|j
        d         }g }t          |
¦  «        D �]Ý}||         }||         }|                     d¬¦  «        ddd…f                              d¦  «        \  }}|dk                         ¦   «         }|                     ¦   «                              d	¦  «        |                     d	¦  «        z                       d	¦  «        |                     d	¦  «                             d	¦  «        d
z   z  }||z  }t	          j        ||         |	¬¦  «        d	z
  }g g }}d}t          |¦  «        D ]®}||                              ¦   «         }t	          j        ||         dk    ¦  «        st||k    rn||||         d	k    <   |                     |||                              ¦   «         t-          |d¦  «        dœ¦  «         |d	z  }|                     ||         ¦  «         Œ¯|                     ||dœ¦  «         �Œß|S )zDPost-processes model outputs into Instance Segmentation Predictions.Nrû   r  r   éþÿÿÿr  r/   .r*   g�íµ ÷Æ°>r1   rx   ry   r*  )r˜   r  r  rJ   r3   r°   r   rÿ   r  r2   r8   r   r  r›   r;   r}   Úflattenre   r7   r:   Úallr�   r€   )r•   r!  r÷   r(  r˜   r  r  r"  r-  r2   r,  Únum_queriesr0  r-   Ú	mask_predÚ
mask_classr]   Úpred_classesÚ
pred_masksÚmask_scoresr‚   r‡   Úinstance_mapsrˆ   r‰   Újr|   s                              r$   Ú"post_process_instance_segmentationz5EomtImageProcessor.post_process_instance_segmentation{  sª  € ð Ð'ˆtˆt¨T¬Yˆà&Ô;ÐØ&Ô;Ðå% dÑ+Ô+ˆÝ$œxÔ2×>Ò>Ø ØØð  ?ñ  
ô  
Ðð  ×+Ò+Ð,@À,ÐPTÑUÔUÐà%Ô,ˆØ)Ô/°Ô2ˆ
Ø*Ô0°Ô4ˆàˆå�zÑ"Ô"ð !	Vñ !	VˆAØ(¨Ô+ˆIØ-¨aÔ0ˆJð $.×#5Ò#5¸"Ð#5Ñ#=Ô#=¸cÀ3ÀBÀ3¸hÔ#G×#KÒ#KÈBÑ#OÔ#OÑ ˆF�LØ# aš-×.Ò.Ñ0Ô0ˆJð %×,Ò,Ñ.Ô.×6Ò6°qÑ9Ô9¸J×<NÒ<NÈqÑ<QÔ<QÑQ×VÒVÐWXÑYÔYØ×"Ò" 1Ñ%Ô%×)Ò)¨!Ñ,Ô,¨tÑ3ñˆKð ! ;Ñ.ˆKå œ; |°A¤¸vÐFÑFÔFÈÑJˆLà&(¨"˜8ˆMØ!"ÐÝ˜;Ñ'Ô'ð 8ð 8�Ø# Aœ×+Ò+Ñ-Ô-�å”y ¨A¤°!Ò!3Ñ4Ô4ð 
8¸À)Ò9KÐ9KØ7I�L ¨A¤°!Ò!3Ñ4Ø—O’Oà"4Ø(4°Q¬×(<Ò(<Ñ(>Ô(>Ý%*¨5°!¡_¤_ðð ñô ð ð '¨!Ñ+Ð&Ø!×(Ò(¨°A¬Ñ7Ô7Ð7øà�NŠN¨LÈ8ÐTÐTÑUÔUÐUÑUØˆr#   ©NN)N)NF)rc   rb   rc   NN)rc   N)2r   r   r   r   Úvalid_kwargsr   ÚBILINEARrÂ   r   rà   r   rá   r˜   Údefault_to_squarerÝ   rÁ   rÀ   r   rß   r   r   r”   r3   ÚTensorr   rR   r!   Útupler«   r¶   r   r   Údictr   r¹   r   r   Ústrr   r   rÜ   r;   rÐ   r  r  r'  r1  r   r>  Ú__classcell__)r–   s   @r$   r�   r�   Õ   s0  ø€ € € € € à+€LØ!Ô*€HØ&€JØ$€IØ °#Ð6Ð6€DØÐØ€IØ€JØ€LØ€NØ€FØ€Lð# Ð(@Ô!Að #ð #ð #ð #ð #ð #ð& 5¤<ð &°xð &ÐPTÐUXÔPYð &Ð^cÐdhÐjnÐdnÔ^oð &ð &ð &ð &ð8˜5œ<ð ¨xð ¸E¼Lð ð ð ð ð ð 8<Ø<@ð	cð càðcð   ¤Ô-°Ñ4ðcð %)¨¨c¨¤N°TÑ$9ð	cð
 Ð1Ô2ðcð 
ðcð cð cð cð cñ „^ðcð. 59ðL
ð L
àðL
ð &¨Ñ,ðL
ð %)¨¨c¨¤N°TÑ$9ð	L
ð
 ðL
ð ,ðL
ð ˜jÑ(¨4Ñ/ðL
ð �c˜>Ð)Ô*¨TÑ1ðL
ð Ð1Ô2ðL
ð 
ðL
ð L
ð L
ð L
ð\7/à�^Ô$ð7/ð ð7/ð ð	7/ð
 Lð7/ð ð7/ð ð7/ð ð7/ð ð7/ð ð7/ð ˜D œKÑ'¨$Ñ.ð7/ð ˜4 œ;Ñ&¨Ñ-ð7/ð  ™+ð7/ð 7/ð 7/ð 7/ðr7$à"œ\ð7$ð ˜E # s¨C -Ô0Ô1ð7$ð ˜5  c œ?Ô+ð	7$ð
 �3˜�8Œnð7$ð 
ˆeŒlÔ	ð7$ð 7$ð 7$ð 7$ðrà"œ\ðð ˜5  c œ?Ô+ðð �3˜�8Œnð	ð
 
ˆeŒlÔ	ðð ð ð ð2 '+Ø+0ðJ%ð J%ð ˜5  c œ?Ô+ðJ%ð �3˜�8Œn˜tÑ#ð	J%ð
 %)ðJ%ð 
NðJ%ð J%ð J%ð J%ð` Ø #Ø-0Ø*.Ø&*ð7ð 7ð ˜5  c œ?Ô+ð7ð ð	7ð
 ð7ð &+ð7ð ˜C”y 4Ñ'ð7ð �3˜�8Œn˜tÑ#ð7ð 7ð 7ð 7ðr %Ð$Ñ&Ô&ð
 Ø&*ð?ð ?ð ˜5  c œ?Ô+ð?ð ð	?ð
 �3˜�8Œn˜tÑ#ð?ð ?ð ?ñ 'Ô&ð?ð ?ð ?ð ?ð ?r#   r�   r?  )rb   rc   )rb   rc   N)0r   rœ   Útypingr   r3   Útorchvision.transforms.v2r   ÚtvFÚimage_processing_backendsr   Úimage_processing_outputsr   Úimage_processing_utilsr   Úimage_transformsr	   r
   r   Úimage_utilsr   r   r   r   r   r   Úprocessing_utilsr   r   Úutilsr   r   r   r   rE  r!   rB   rF  rD  rJ   rR   rC  rX   ra   rs   r;   r�   r�   Ú__all__r"   r#   r$   ú<module>rS     s÷  ðð &Ð %à €€€Ø Ð Ð Ð Ð Ð à €€€Ø 7Ð 7Ð 7Ð 7Ð 7Ð 7à ;Ð ;Ð ;Ð ;Ð ;Ð ;Ø OÐ OÐ OÐ OÐ OÐ OØ 2Ð 2Ð 2Ð 2Ð 2Ð 2Ø aÐ aÐ aÐ aÐ aÐ aÐ aÐ aÐ aÐ aðð ð ð ð ð ð ð ð ð ð ð ð ð ð ð ð 5Ð 4Ð 4Ð 4Ð 4Ð 4Ð 4Ð 4ðð ð ð ð ð ð ð ð ð ðð ð ð ð ˜|°5ð ñ ô ð ð$ 9=Ø#ð/ð /Ø$ð/à $ S¨# X¤°Ñ 5ð/ð ˜‘*ð/ð /ð /ð /ð>'˜t C¨ Hœ~ð '°%¸¸S¸´/ð 'ð 'ð 'ð 'ð	6Ø�%”,Ôð	6Ø*.¨t°C¬y¬/ð	6à
ˆ4�”Ô˜t D¨¤IœÐ.Ô/ð	6ð 	6ð 	6ð 	6ð<ð <ð <ð8#ð #ð #ð #ð6  Ø),Ø*.ð3"ð 3"ð
 ð3"ð "'ð3"ð �s˜C�x” 4Ñ'ð3"ð 3"ð 3"ð 3"ðl ðeð eð eð eð eÐ+ñ eô eñ „ðeðP  Ð
 €€€r#   