a h3 @s8ddlZddlmZddlmZddlmZmZmZm Z ddl Z ddl m Z m Z ddl mZddlmZdd lmZd d lmZd d lmZmZmZd d lmZd dlmZmZd dlmZm Z ddlmZ!ddl"m#Z#ddl$m%Z%ddl&m'Z'm(Z(ddgZ)e*e*e*ede j+fe j,dddZ-e*e*ede j+fe j,dddZ.e j+ddd Z/Gd!d"d"e j+Z0Gd#d$d$e(Z1Gd%d&d&e(Z2Gd'd(d(e j+Z3e ej4ej5fe*ede j+fd)d*d+Z6Gd,ddeZ7eed-e7j8fd.e j9fd/dd0de j9ddd1ee7e:ee*ee ee*eede j+fee'd2d3dZ;dS)4N) OrderedDict)partial)AnyCallableOptionalUnion)nnTensor)Conv2dNormActivation)ObjectDetection)_log_api_usage_once) mobilenet)register_modelWeights WeightsEnum)_COCO_CATEGORIES)_ovewrite_value_paramhandle_legacy_interface)mobilenet_v3_largeMobileNet_V3_Large_Weights)_utils)DefaultBoxGenerator)_validate_trainable_layers)SSDSSDScoringHead%SSDLite320_MobileNet_V3_Large_Weightsssdlite320_mobilenet_v3_large.) in_channels out_channels kernel_size norm_layerreturnc Cs(tt|||||tjdt||dS)N)r"groupsr#activation_layerr)r Sequentialr ReLU6Conv2d)r r!r"r#r*R/var/www/auris/lib/python3.9/site-packages/torchvision/models/detection/ssdlite.py_prediction_blocks r,)r r!r#r$c CsJtj}|d}tt||d||dt||dd|||dt||d||dS)Nrr)r"r#r&r )r"Zstrider%r#r&)rr(r'r )r r!r#Z activationZintermediate_channelsr*r*r+ _extra_block0s&   r-)convcCsP|D]B}t|tjrtjjj|jddd|jdurtjj |jdqdS)NgQ?)meanZstd) modules isinstancerr)torchinitZnormal_ZweightZbiasZ constant_)r.Zlayerr*r*r+ _normal_initIs    r5csVeZdZeeeeeedejfdfdd Zee e e e fdddZ Z S) SSDLiteHead.r num_anchors num_classesr#cs,tt|||||_t||||_dSN)super__init__SSDLiteClassificationHeadclassification_headSSDLiteRegressionHeadregression_head)selfr r8r9r# __class__r*r+r<Rs zSSDLiteHead.__init__xr$cCs||||dS)N)Zbbox_regression cls_logits)r@r>)rArEr*r*r+forwardYszSSDLiteHead.forward)__name__ __module__ __qualname__listintrrModuler<r dictstrrG __classcell__r*r*rBr+r6Qsr6cs:eZdZeeeeeedejfdfdd ZZ S)r=.r7csNt}t||D] \}}|t|||d|qt|t||dS)Nr r ModuleListzipappendr,r5r;r<)rAr r8r9r#rFchannelsanchorsrBr*r+r<as z"SSDLiteClassificationHead.__init__ rHrIrJrKrLrrrMr<rPr*r*rBr+r=`sr=cs8eZdZeeeeedejfdfdd ZZ S)r?.)r r8r#csNt}t||D] \}}|t|d|d|qt|t|ddS)Nr rQ)rAr r8r#Zbbox_regrUrVrBr*r+r<ls zSSDLiteRegressionHead.__init__rWr*r*rBr+r?ksr?csPeZdZd ejeedejfeedfdd Ze e e e fddd Z Z S) SSDLiteFeatureExtractorMobileNet?.)backbonec4_posr# width_mult min_depthc stt|||jr$tdttjg|d|||jdRtj||jddg||ddR|_fdd}t t |dj |d|t |d|d|t |d|d|t |d|d |g}t |||_ dS) Nz0backbone[c4_pos].use_res_connect should be Falserrcstt|Sr:)maxrL)dr_r^r*r+z;SSDLiteFeatureExtractorMobileNet.__init__..i)r;r<r Zuse_res_connect ValueErrorrr'blockfeaturesrRr-r!r5extra)rAr\r]r#r^r_Z get_depthrkrBrbr+r<us$  $*z)SSDLiteFeatureExtractorMobileNet.__init__rDcCsVg}|jD]}||}||q |jD]}||}||q(tddt|DS)NcSsg|]\}}t||fqSr*)rO).0ivr*r*r+ rdz.)rjrTrkr enumerate)rArEoutputrir*r*r+rGs    z(SSDLiteFeatureExtractorMobileNet.forward)rZr[)rHrIrJrrMrLrfloatr<r rNrOrGrPr*r*rBr+rYts !rY)r\trainable_layersr#cCs|j}dgddt|Dt|dg}t|}d|krH|ksRntd|dkrbt|n |||}|d|D]}|D]}|dqqzt||d|S)NrcSs g|]\}}t|ddr|qS)Z_is_cnF)getattr)rlrmbr*r*r+rordz(_mobilenet_extractor..rzYtrainable_layers should be in the range [0, {num_stages}], instead got {trainable_layers}F)rjrplenrh parametersZrequires_grad_rY)r\rsr#Z stage_indicesZ num_stagesZ freeze_beforeruZ parameterr*r*r+_mobilenet_extractors& ryc @s8eZdZedededddddiidd d d d ZeZd S)rzShttps://download.pytorch.org/models/ssdlite320_mobilenet_v3_large_coco-a79551df.pthi}4)rrz]https://github.com/pytorch/vision/tree/main/references/detection#ssdlite320-mobilenetv3-largez COCO-val2017Zbox_mapgL5@g-?gt*@zSThese weights were produced by following a similar training recipe as on the paper.)Z num_params categoriesZmin_sizeZrecipeZ_metricsZ_ops _file_sizeZ_docs)urlZ transformsmetaN)rHrIrJrr rCOCO_V1DEFAULTr*r*r*r+rs"Z pretrainedZpretrained_backbone)weightsweights_backboneT)rprogressr9rtrainable_backbone_layersr#)rrr9rrr#kwargsr$cKst|}t|}d|vr&td|durJd}td|t|jd}n |durVd}t|dupf|du|dd}|du}|durt t j dd d }t f||||d |}|durt |t|||}d } td dtdDddd} t|| } | } t| t| jkr*tdt| dt| jddddgdgdd} i| |}t|| | |fdt| | ||i|}|dur||j|dd|S)aSSDlite model architecture with input size 320x320 and a MobileNetV3 Large backbone, as described at `Searching for MobileNetV3 `__ and `MobileNetV2: Inverted Residuals and Linear Bottlenecks `__. .. betastatus:: detection module See :func:`~torchvision.models.detection.ssd300_vgg16` for more details. Example: >>> model = torchvision.models.detection.ssdlite320_mobilenet_v3_large(weights=SSDLite320_MobileNet_V3_Large_Weights.DEFAULT) >>> model.eval() >>> x = [torch.rand(3, 320, 320), torch.rand(3, 500, 400)] >>> predictions = model(x) Args: weights (:class:`~torchvision.models.detection.SSDLite320_MobileNet_V3_Large_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.detection.SSDLite320_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). weights_backbone (:class:`~torchvision.models.MobileNet_V3_Large_Weights`, optional): The pretrained weights for the backbone. trainable_backbone_layers (int, optional): number of trainable (not frozen) layers starting from final block. Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable. If ``None`` is passed (the default) this value is set to 6. norm_layer (callable, optional): Module specifying the normalization layer to use. **kwargs: parameters passed to the ``torchvision.models.detection.ssd.SSD`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.detection.SSDLite320_MobileNet_V3_Large_Weights :members: sizez?The size of the model is already fixed; ignoring the parameter.Nr9rz[gMbP?r/)ZepsZmomentum)rrr#Z reduced_tail)@rcSsg|] }ddgqS)rr r*)rl_r*r*r+ro,rdz1ssdlite320_mobilenet_v3_large..g?gffffff?)Z min_ratioZ max_ratioz4The length of the output channels from the backbone z? do not match the length of the anchor generator aspect ratios g?i,)?rr)Z score_threshZ nms_threshZdetections_per_imgZtopk_candidatesZ image_meanZ image_stdheadT)rZ check_hash)rverifyrwarningswarnrrwr}rrrZ BatchNorm2drr5ryrrange det_utilsZretrieve_out_channelsZnum_anchors_per_locationZ aspect_ratiosrhrr6Zload_state_dictZget_state_dict)rrr9rrr#rZ reduce_tailr\rZanchor_generatorr!r8defaultsmodelr*r*r+rsp8       )sn              1