ڧgddlmZddlmZddlmZmZmZddlm Z m Z ddl m Z ddlmZddlmZd d lmZmZmZd d lmZd d lmZmZmZd d lmZmZmZgdZ Gdde j!Z"Gdde j!Z#dede$de"fdZ%GddeZ&eede&j'fdej(fdddej(ddee&d e)dee$d!eed"ede"f d#Z*dS)$) OrderedDict)partial)AnyDictOptional)nnTensor) functional)SemanticSegmentation)_log_api_usage_once)register_modelWeights WeightsEnum)_VOC_CATEGORIES)_ovewrite_value_paramhandle_legacy_interfaceIntermediateLayerGetter)mobilenet_v3_largeMobileNet_V3_Large_Weights MobileNetV3)LRASPP!LRASPP_MobileNet_V3_Large_Weightslraspp_mobilenet_v3_largec jeZdZdZ d dejdededededd f fd Zd ede e effd Z xZ S)ra Implements a Lite R-ASPP Network for semantic segmentation from `"Searching for MobileNetV3" `_. Args: backbone (nn.Module): the network used to compute the features for the model. The backbone should return an OrderedDict[Tensor], with the key being "high" for the high level feature map and "low" for the low level feature map. low_channels (int): the number of channels of the low level features. high_channels (int): the number of channels of the high level features. num_classes (int, optional): number of output classes of the model (including the background). inter_channels (int, optional): the number of channels for intermediate computations. backbone low_channels high_channels num_classesinter_channelsreturnNctt|||_t |||||_dS)N)super__init__r r LRASPPHead classifier)selfrrr r!r" __class__s b/var/www/html/ai-engine/env/lib/python3.11/site-packages/torchvision/models/segmentation/lraspp.pyr&zLRASPP.__init__#sI D!!!  $\=+~^^inputc||}||}tj||jdddd}t }||d<|S)NbilinearFsizemode align_cornersout)rr(F interpolateshaper)r)r-featuresr5results r+forwardzLRASPP.forward+s_==''ooh''mCek"##&6ZW\]]]u  r,)r) __name__ __module__ __qualname____doc__rModuleintr&r rstrr; __classcell__r*s@r+rrs   sv__ _14_EH_WZ_lo_ ______VS&[(9r,rc TeZdZdededededdf fd Zdeeefdefd ZxZ S) r'rr r!r"r#Nc ttjtj||ddtj|tjd|_tjtjdtj||ddtj |_ tj||d|_ tj||d|_ dS)NF)biasT)inplace) r%r&r SequentialConv2d BatchNorm2dReLUcbrAdaptiveAvgPool2dSigmoidscalelow_classifierhigh_classifier)r)rr r!r"r*s r+r&zLRASPPHead.__init__7s = Im^QU C C C N> * * GD ! ! !   ]   # # Im^QU C C C JLL   !i k1EE!yaHHr,r-c |d}|d}||}||}||z}tj||jdddd}||||zS)Nlowhighr/r0Fr1)rNrQr6r7r8rRrS)r)r-rUrVxss r+r;zLRASPPHead.forwardFsElV} HHTNN JJt   E M!#)BCC.zQV W W W""3''$*>*>q*A*AAAr,) r<r=r>rAr&rrBr r;rCrDs@r+r'r'6s IS I I3 I`c Ihl I I I I I I BT#v+. B6 B B B B B B B Br,r'rr!r#cH|j}dgdt|Dzt|dz gz}|d}|d}||j}||j}t |t |dt |di}t ||||S) Nrc:g|]\}}t|dd|S)_is_cnF)getattr).0ibs r+ z'_lraspp_mobilenetv3..Vs.\\\A8UZ@[@[\1\\\r,rGrUrV) return_layers)r9 enumeratelen out_channelsrrBr)rr! stage_indiceslow_poshigh_posrr s r+_lraspp_mobilenetv3rjRs HC\\8)<)<\\\\`cdl`m`mpq`q_rrMBGR HG$1LX&3M&xG eUXYaUbUbdj?klllH (L- E EEr,c `eZdZedeeddeddddd d id d d dZeZdS)rzJhttps://download.pytorch.org/models/lraspp_mobilenet_v3_large-d234d4ea.pthi) resize_sizei"(1)rGrGz]https://github.com/pytorch/vision/tree/main/references/segmentation#lraspp_mobilenet_v3_largezCOCO-val2017-VOC-labelsg33333L@gV@)miou pixel_accg㥛 @g{G(@z These weights were trained on a subset of COCO, using only the 20 categories that are present in the Pascal VOC dataset. ) num_params categoriesmin_sizerecipe_metrics_ops _file_size_docs)url transformsmetaN) r<r=r>rrr rCOCO_WITH_VOC_LABELS_V1DEFAULTr,r+rr`s}%g X7/SAAA!)u) !%,,   ,&GGGr,r pretrainedpretrained_backbone)weightsweights_backboneNT)rprogressr!rrrrkwargsc |ddrtdt|}t j|}|,d}t d|t |jd}n|d}t|d }t||}|*| | |d |S) a|Constructs a Lite R-ASPP Network model with a MobileNetV3-Large backbone from `Searching for MobileNetV3 `_ paper. .. betastatus:: segmentation module Args: weights (:class:`~torchvision.models.segmentation.LRASPP_MobileNet_V3_Large_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.segmentation.LRASPP_MobileNet_V3_Large_Weights` below for more details, and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. num_classes (int, optional): number of output classes of the model (including the background). aux_loss (bool, optional): If True, it uses an auxiliary loss. weights_backbone (:class:`~torchvision.models.MobileNet_V3_Large_Weights`, optional): The pretrained weights for the backbone. **kwargs: parameters passed to the ``torchvision.models.segmentation.LRASPP`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.segmentation.LRASPP_MobileNet_V3_Large_Weights :members: aux_lossFz&This model does not use auxiliary lossNr!rpT)rdilated)r check_hash) popNotImplementedErrorrverifyrrreryrrjload_state_dictget_state_dict)rrr!rrrmodels r+rrzsLzz*e$$L!"JKKK/66w??G189IJJ+M;GLYeLfHgHghh   !*:DIIIH + 6 6E g44hSW4XXYYY Lr,)+ collectionsr functoolsrtypingrrrtorchrr torch.nnr r6transforms._presetsr utilsr _apirrr_metar_utilsrrr mobilenetv3rrr__all__r@rr'rArjrrz IMAGENET1K_V1boolrr|r,r+rs######&&&&&&&&&&$$$$$$777777((((((7777777777######\\\\\\\\\\UUUUUUUUUU W V V     RY   FBBBBBBBB8 F+ FC FF F F F F&&&&& &&&4 <T U+-G-UV <@!%=W=e 333 7 833# 3 9: 3  3 333  333r,