§
    l¬¸j­  ã                  ó¢   — d dl mZ d dlm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  G d„ de¦  «        ZdS )é    )Úannotations)Úcopy)ÚAnyN)Úadd_polygon_background)Úyolo)ÚDetectionTrainer)ÚSemanticSegmentationModel)ÚDEFAULT_CFGÚLOGGERÚRANK)ÚcolorsÚplt_settingsc                  ó„   ‡ — e Zd ZdZeddfdˆ fd„Zˆ fd„Zddd„Zd„ Zˆ fd„Z	dd„Z
d„ Z e¦   «         d„ ¦   «         Zˆ xZS )ÚSemanticSegmentationTraineraÐ  Trainer for YOLO semantic segmentation models.

    This trainer handles semantic segmentation training, including dataset building, model initialization, and
    validation setup.

    Examples:
        >>> from ultralytics.models.yolo.semantic import SemanticSegmentationTrainer
        >>> args = dict(model="yolo26n-sem.yaml", data="cityscapes8.yaml", epochs=3)
        >>> trainer = SemanticSegmentationTrainer(overrides=args)
        >>> trainer.train()
    NÚ	overridesúdict[str, Any] | NoneÚ
_callbacksúdict | Nonec                ó`   •— |€i }d|d<   t          ¦   «                              |||¦  «         dS )a  Initialize SemanticSegmentationTrainer.

        Args:
            cfg (dict): Configuration dictionary with default training settings.
            overrides (dict, optional): Dictionary of parameter overrides.
            _callbacks (dict, optional): Callback functions.
        NÚsemanticÚtask)ÚsuperÚ__init__)ÚselfÚcfgr   r   Ú	__class__s       €úv/root/private/project/pallet_detection_v2/.venv/lib/python3.11/site-packages/ultralytics/models/yolo/semantic/train.pyr   z$SemanticSegmentationTrainer.__init__   s<   ø€ ð ÐØˆIØ&ˆ	�&ÑÝ‰Œ×Ò˜˜i¨Ñ4Ô4Ð4Ð4Ð4ó    c                ó^   •— t          t          ¦   «                              ¦   «         ¦  «        S )zTParse the dataset YAML and add background metadata for polygon labels when required.)r   r   Úget_dataset©r   r   s    €r   r    z'SemanticSegmentationTrainer.get_dataset,   s!   ø€ å%¥e¡g¤g×&9Ò&9Ñ&;Ô&;Ñ<Ô<Ð<r   Tr   ú
str | NoneÚweightsÚverboseÚboolc                óœ   — t          || j        d         | j        d         |o
t          dk    ¬¦  «        }|r|                     |¦  «         |S )a}  Return a SemanticSegmentationModel with optional pretrained backbone.

        Args:
            cfg (str, optional): Path to model configuration file.
            weights (str | Path, optional): Path to model weights.
            verbose (bool): Whether to display model information.

        Returns:
            (SemanticSegmentationModel): Semantic segmentation model.
        ÚncÚchannelséÿÿÿÿ)r'   Úchr$   )r	   Údatar   Úload)r   r   r#   r$   Úmodels        r   Ú	get_modelz%SemanticSegmentationTrainer.get_model0   s[   € õ *Ø�D”I˜d”O¨¬	°*Ô(=ÀwÐG]ÕSWÐ[]ÒS]ð
ñ 
ô 
ˆð ð 	 Ø�JŠJ�wÑÔÐàˆr   c                óŠ   — t           j                             | j        | j        t          | j        ¦  «        | j        ¬¦  «        S )z<Return a SemanticSegmentationValidator for model evaluation.)Úsave_dirÚargsr   )r   r   ÚSemanticSegmentationValidatorÚtest_loaderr0   r   r1   Ú	callbacks)r   s    r   Úget_validatorz)SemanticSegmentationTrainer.get_validatorC   s;   € åŒ}×:Ò:ØÔ t¤}½4ÀÄ	¹?¼?ÐW[ÔWeð ;ñ 
ô 
ð 	
r   c                ón   •— | j         d         dk    r"t          ¦   «                              ¦   «          dS dS )z^Compute pixel-frequency class weights; skipped for binary (nc==1, unweighted BCE in the loss).r'   é   N)r+   r   Úset_class_weightsr!   s    €r   r8   z-SemanticSegmentationTrainer.set_class_weightsI   s8   ø€ àŒ9�TŒ?˜QÒÐÝ‰GŒG×%Ò%Ñ'Ô'Ð'Ð'Ð'ð Ðr   c                ót  — | j         d         }t          j        |t          j        ¬¦  «        }| j        j        }t          |dg ¦  «        }|s|S t          j        t          |¦  «        ¦  «        }|rQt          |¦  «        |k    r>t          j	        dt          |¦  «        dz
  |¦  «         
                    t          ¦  «        }t          |dd¦  «        }|D ]á}||                              d¦  «        }		 |                     ||	�t          |	¦  «        nd¬	¦  «        }
n# t          $ r Y ŒSw xY w|�d
|
t          j        |
|¦  «         <   |
dk    |
|k     z  |
d
k    z  }|                     ¦   «         rGt          j        |
|         d¬¦  «        \  }}|| 
                    t          ¦  «        xx         |z  cc<   Œâ|S )zSReturn per-class pixel counts from training masks, optionally sampled to max_masks.r'   )ÚdtypeÚlabelsr   r7   Úinclude_classNÚshape)Úimage_shapeéÿ   T)Úreturn_counts)r+   ÚnpÚzerosÚfloat32Útrain_loaderÚdatasetÚgetattrÚarangeÚlenÚlinspaceÚastypeÚintÚgetÚ	load_maskÚtupleÚ	ExceptionÚisinÚanyÚunique)r   Ú	max_masksr'   Úpixel_countsrE   r;   Úindicesr<   Úidxr=   ÚmaskÚvalidÚclassesÚcountss                 r   Úget_class_countsz,SemanticSegmentationTrainer.get_class_countsN   s¿  € àŒY�tŒ_ˆÝ”x ­"¬*Ð5Ñ5Ô5ˆØÔ#Ô+ˆÝ˜ (¨BÑ/Ô/ˆØð 	 ØÐÝ”)�C ™KœKÑ(Ô(ˆØð 	M�˜W™œ¨	Ò1Ð1Ý”k !¥S¨¡[¤[°1¡_°iÑ@Ô@×GÒGÍÑLÔLˆGÝ ¨¸$Ñ?Ô?ˆØð 	<ð 	<ˆCØ˜3”K—O’O GÑ,Ô,ˆEðØ×(Ò(¨È%ÐJ[½%À¹,¼,¸,ÐaeÐ(ÑfÔf��øÝð ð ð Ø�ðøøøàÐ(Ø69�•b”g˜d MÑ2Ô2Ð2Ñ3Ø˜Q’Y 4¨"¢9Ñ-°¸²Ñ=ˆEØ�yŠy‰{Œ{ð <Ý"$¤)¨D°¬KÀtÐ"LÑ"LÔ"L‘�˜Ø˜WŸ^š^­CÑ0Ô0Ð1Ð1Ô1°VÑ;Ð1Ð1Ñ1øØÐs   Ã4(DÄ
D*Ä)D*c                óš   — |t          |                     ¦   «         d¦  «        z  }dt          j        d|z   ¦  «        z  | j        j        z  S )zdCompute ENet inverse-log `(1/ln(1.02 + p))**cls_pw` weights (Paszke et al., 2016, arXiv:1606.02147).g      ð?gR¸…ëQð?)ÚmaxÚsumrA   Úlogr1   Úcls_pw)r   Úclass_countsÚps      r   Úcompute_class_weightsz1SemanticSegmentationTrainer.compute_class_weightsh   sE   € à�3˜|×/Ò/Ñ1Ô1°3Ñ7Ô7Ñ7ˆØ•b”f˜T A™XÑ&Ô&Ñ&¨4¬9Ô+;Ñ;Ð;r   c           	     ó(  — ddl m} t          j        d| j        dz  › d�¦  «         | j        d         }| j        d         }|                      d¬	¦  «        }|                     ¦   «         st          j        d
¦  «         dS | 	                    dddd¬¦  «        \  }}| 
                    t          |¦  «        |d„ t          |¦  «        D ¦   «         ¬¦  «        }|                     d¦  «         |                     d¦  «         |                     d¦  «         dt          |¦  «        cxk     rdk     rin nf|                     t          t          |¦  «        ¦  «        ¦  «         |                     t%          |                     ¦   «         ¦  «        dd¬¦  «         |D ]p}|                     ¦   «         }	|	dk    rT|                     |                     ¦   «         |                     ¦   «         dz  z   |	t1          |	¦  «        d›ddd¬¦  «         Œq|j                             ¦   «         D ]}
|
                     d¦  «         Œ| j        dz  }|                     |d¬ ¦  «         |                     ¦   «          | j        r|                      |¦  «         dS dS )!züPlot training labels class distribution for semantic segmentation.

        Samples up to 1000 mask files from the training dataset, accumulates per-class pixel
        counts, and plots a bar chart of class distribution saved to 'labels.jpg'.
        r   NzPlotting labels to z
labels.jpgz...r'   Únamesiè  )rS   z2No semantic mask files found, skipping label plot.r7   )é   é   T)ÚfigsizeÚtight_layoutc                óB   — g | ]}d „ t          |d¦  «        D ¦   «         ‘ŒS )c                ó   — g | ]}|d z  ‘ŒS )g     ào@© )Ú.0Úcs     r   ú
<listcomp>zOSemanticSegmentationTrainer.plot_training_labels.<locals>.<listcomp>.<listcomp>   s   € Ð6[Ð6[Ð6[ÀQ°q¸5±yÐ6[Ð6[Ð6[r   F)r   )rm   Úis     r   ro   zDSemanticSegmentationTrainer.plot_training_labels.<locals>.<listcomp>   s4   € Ð5oÐ5oÐ5oÐ`aÐ6[Ð6[Í&ÐQRÐTYÑJZÔJZÐ6[Ñ6[Ô6[Ð5oÐ5oÐ5or   )ÚcolorÚClassÚPixelsz"Training Labels Class Distributioné   éZ   é
   )ÚrotationÚfontsizeg       @ú,ÚcenterÚbottomrf   )ÚhaÚvarx   FéÈ   )Údpi)Úmatplotlib.pyplotÚpyplotr   Úinfor0   r+   r[   rQ   ÚwarningÚsubplotsÚbarÚrangeÚ
set_xlabelÚ
set_ylabelÚ	set_titlerH   Ú
set_xticksÚset_xticklabelsÚlistÚvaluesÚ
get_heightÚtextÚget_xÚ	get_widthrK   ÚspinesÚset_visibleÚsavefigÚcloseÚon_plot)r   Úpltr'   re   rT   Ú_ÚaxÚbarsr…   ÚheightÚspineÚfnames               r   Úplot_training_labelsz0SemanticSegmentationTrainer.plot_training_labelsm   s™  € ð 	(Ð'Ð'Ð'Ð'Ð'åŒÐK¨$¬-¸,Ñ*FÐKÐKÐKÑLÔLÐLØŒY�tŒ_ˆØ”	˜'Ô"ˆØ×,Ò,°tÐ,Ñ<Ô<ˆØ×ÒÑ!Ô!ð 	ÝŒNÐOÑPÔPÐPØˆFà—’˜Q ¨6À�ÑEÔE‰ˆˆ2Ø�vŠv•e˜B‘i”i Ð5oÐ5oÕejÐkmÑenÔenÐ5oÑ5oÔ5oˆvÑpÔpˆØ
�Š�gÑÔÐØ
�Š�hÑÔÐØ
�ŠÐ9Ñ:Ô:Ð:Ø�s�5‰zŒzÐÐÒÐ˜BÒÐÐÐÐØ�MŠM�%¥ E¡
¤
Ñ+Ô+Ñ,Ô,Ð,Ø×Ò�t E§L¢L¡N¤NÑ3Ô3¸bÈ2ÐÑNÔNÐNØð 
	ð 
	ˆCØ—^’^Ñ%Ô%ˆFØ˜ŠzˆzØ—’Ø—I’I‘K”K #§-¢-¡/¤/°CÑ"7Ñ7ØÝ˜6‘{”{Ð&Ð&ØØØð ñ ô ð øð ”Y×%Ò%Ñ'Ô'ð 	%ð 	%ˆEØ×Ò˜eÑ$Ô$Ð$Ð$à” Ñ,ˆØ�Š�E˜sˆÑ#Ô#Ð#Ø�	Š	‰ŒˆØŒ<ð 	 Ø�LŠL˜ÑÔÐÐÐð	 ð 	 r   )r   r   r   r   )NNT)r   r"   r#   r"   r$   r%   )N)Ú__name__Ú
__module__Ú__qualname__Ú__doc__r
   r   r    r.   r5   r8   r[   rc   r   rž   Ú__classcell__)r   s   @r   r   r      sô   ø€ € € € € ð
ð 
ð 'È4Ðkoð 5ð 5ð 5ð 5ð 5ð 5ð 5ð=ð =ð =ð =ð =ðð ð ð ð ð&
ð 
ð 
ð(ð (ð (ð (ð (ð
ð ð ð ð4<ð <ð <ð
 €\�^„^ð* ð * ñ „^ð* ð * ð * ð * ð * r   r   )Ú
__future__r   r   Útypingr   ÚnumpyrA   Úultralytics.data.utilsr   Úultralytics.modelsr   Úultralytics.models.yolo.detectr   Úultralytics.nn.tasksr	   Úultralytics.utilsr
   r   r   Úultralytics.utils.plottingr   r   r   rl   r   r   ú<module>r­      s  ðð #Ð "Ð "Ð "Ð "Ð "à Ð Ð Ð Ð Ð Ø Ð Ð Ð Ð Ð à Ð Ð Ð à 9Ð 9Ð 9Ð 9Ð 9Ð 9Ø #Ð #Ð #Ð #Ð #Ð #Ø ;Ð ;Ð ;Ð ;Ð ;Ð ;Ø :Ð :Ð :Ð :Ð :Ð :Ø 7Ð 7Ð 7Ð 7Ð 7Ð 7Ð 7Ð 7Ð 7Ð 7Ø ;Ð ;Ð ;Ð ;Ð ;Ð ;Ð ;Ð ;ðF ð F ð F ð F ð F Ð"2ñ F ô F ð F ð F ð F r   