o #jgC@sddlmZddlZddlmZddlmZddlmZddlm Z gZ d d d Z Gd d d ej Z GdddZGdddej ZGdddej ZGdddeZGdddeZd ddZd ddZd ddZdS)!)partialN)nn)get_weights_path_from_url)ConvNormActivation_make_divisible)zIhttps://paddle-hapi.bj.bcebos.com/models/mobilenet_v3_small_x1.0.pdparamsZ 34fe0e7c1f8b00b2b056ad6788d0590c)zIhttps://paddle-hapi.bj.bcebos.com/models/mobilenet_v3_large_x1.0.pdparamsZ 118db5792b4e183b925d8e8e334db3df)zmobilenet_v3_small_x1.0zmobilenet_v3_large_x1.0cs:eZdZdZejejffdd ZddZddZ Z S)SqueezeExcitationa This block implements the Squeeze-and-Excitation block from https://arxiv.org/abs/1709.01507 (see Fig. 1). Parameters ``activation``, and ``scale_activation`` correspond to ``delta`` and ``sigma`` in eq. 3. This code is based on the torchvision code with modifications. You can also see at https://github.com/pytorch/vision/blob/main/torchvision/ops/misc.py#L127 Args: input_channels (int): Number of channels in the input image. squeeze_channels (int): Number of squeeze channels. activation (Callable[..., paddle.nn.Layer], optional): ``delta`` activation. Default: ``paddle.nn.ReLU``. scale_activation (Callable[..., paddle.nn.Layer]): ``sigma`` activation. Default: ``paddle.nn.Sigmoid``. csJttd|_t||d|_t||d|_||_||_ dS)Nr) super__init__rAdaptiveAvgPool2DavgpoolZConv2Dfc1fc2 activationscale_activation)selfZinput_channelsZsqueeze_channelsrr __class__a/var/www/html/Deteccion_Ine/venv/lib/python3.10/site-packages/paddle/vision/models/mobilenetv3.pyr 4s   zSqueezeExcitation.__init__cCs2||}||}||}||}||SN)rrrrrrinputscalerrr_scaleBs     zSqueezeExcitation._scalecCs||}||Sr)rrrrrforwardIs zSqueezeExcitation.forward) __name__ __module__ __qualname____doc__rReLUZSigmoidr rr __classcell__rrrrr &sr c@s&eZdZ dddZedddZdS)InvertedResidualConfig?c Cs|j||d|_||_|j||d|_|j||d|_||_|dur&d|_n|dkr/tj|_n|dkr8tj |_nt d|||_ dS)N)rrelu hardswishz*The activation function is not supported: ) adjust_channels in_channelskernelexpanded_channels out_channelsuse_seactivation_layerrr" Hardswish RuntimeErrorstride) rr)r*r+r,r-rr1rrrrr Os"    zInvertedResidualConfig.__init__cCst||dS)Nr)Zchannelsrrrrr(msz&InvertedResidualConfig.adjust_channelsN)r%)rrr r staticmethodr(rrrrr$Ns  r$cs$eZdZfddZddZZS)InvertedResidualc st|dko ||k|_||_||k|_|jr%t||ddd||d|_t||||t|dd|||d|_|jrHt |t |dt j d|_ t||ddd|dd|_dS)Nrr)r)r, kernel_sizer1padding norm_layerr.rr)r,r5r1r6groupsr7r.)r)r r use_res_connectr-expandr expand_convintbottleneck_convr r rZ Hardsigmoidmid_se linear_conv) rr)r+r, filter_sizer1r-r.r7rrrr ssN      zInvertedResidual.__init__cCsN|}|jr ||}||}|jr||}||}|jr%t||}|Sr) r<r=r?r-r@rAr;paddleadd)rxidentityrrrrs     zInvertedResidual.forward)rrr r rr#rrrrr4rs 7r4cs,eZdZdZ d fdd ZddZZS) MobileNetV3aOMobileNetV3 model from `"Searching for MobileNetV3" `_. Args: config (list[InvertedResidualConfig]): MobileNetV3 depthwise blocks config. last_channel (int): The number of channels on the penultimate layer. scale (float, optional): Scale of channels in each layer. Default: 1.0. num_classes (int, optional): Output dim of last fc layer. If num_classes <=0, last fc layer will not be defined. Default: 1000. with_pool (bool, optional): Use pool before the last fc layer or not. Default: True. r%Tc s t||_||_||_||_||_|dj|_|dj|_ |j d|_ t t j dddtd|jddd d t jd |_t jfd d |jD|_t|j |j d d dd t jd |_|rct d |_|dkrt t |j |jt t jddt |j||_dSdS)NrgMbP?gGz?)epsilonZmomentumrr)r)r,r5r1r6r9r.r7c s2g|]}t|j|j|j|j|j|j|jdqS))r)r+r,rBr1r-r.r7)r4r)r+r,r*r1r-r.).0cfgr7rr s z(MobileNetV3.__init__..r8g?)p)r r configr last_channel num_classes with_poolr)Zfirstconv_in_channelsZlastconv_in_channelsZlastconv_out_channelsrrZ BatchNorm2Drr/convZ Sequentialblockslastconvr rZLinearZDropout classifier)rrRrSrrTrUrrOrr sZ          zMobileNetV3.__init__cCsR||}||}||}|jr||}|jdkr't|d}||}|S)Nrr) rVrWrXrUrrTrCflattenrY)rrErrrrs       zMobileNetV3.forwardr%rHT)rrr r!r rr#rrrrrGs  @rGc"eZdZdZdfdd ZZS)MobileNetV3SmallaMobileNetV3 Small architecture model from `"Searching for MobileNetV3" `_. Args: scale (float, optional): Scale of channels in each layer. Default: 1.0. num_classes (int, optional): Output dim of last fc layer. If num_classes <= 0, last fc layer will not be defined. Default: 1000. with_pool (bool, optional): Use pool before the last fc layer or not. Default: True. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Small architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import MobileNetV3Small >>> # Build model >>> model = MobileNetV3Small(scale=1.0) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] r%rHTcstddddddd|tddddddd|tddd dddd |tdd d d ddd|td d dd ddd |td d dd ddd |td d ddddd |tdd ddddd |tdd dd ddd|td d dd ddd |td d dd ddd |g }td|d}tj|||||ddS)NrLTr&rHFXr`(r'x0i i@ir2rSrrUrTr$r r r rrrTrUrRrSrrrr 0s(  zMobileNetV3Small.__init__r[rrr r!r r#rrrrr]r]cr\)MobileNetV3LargeaMobileNetV3 Large architecture model from `"Searching for MobileNetV3" `_. Args: scale (float, optional): Scale of channels in each layer. Default: 1.0. num_classes (int, optional): Output dim of last fc layer. If num_classes <= 0, last fc layer will not be defined. Default: 1000. with_pool (bool, optional): Use pool before the last fc layer or not. Default: True. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Large architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import MobileNetV3Large >>> # Build model >>> model = MobileNetV3Large(scale=1.0) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] r%rHTcsXtddddddd|tddddddd|tddd dddd|tdd d d d dd|td d d d d dd|td d d d d dd|td dddddd|tddddddd|tddddddd|tddddddd|tddddd dd|tddddd dd|tdd ddd dd|tdd ddd dd|tdd ddd dd|g}td|d}tj|||||ddS)Nr^rLFr&r@r`rr_rbrdTrfrePr'ipiiir2rirjrkrrrr esT# zMobileNetV3Large.__init__r[rlrrrrrnHrmrnFr%cKs|dkrtdd|i|}n tdd|i|}|rB|d|}|tvs+J|dtt|dt|d}t|}|||S)Nmobilenet_v3_largerZ_xzJ model do not have a pretrained model now, you should set pretrained=Falserrr)rnr] model_urlsrrCloadZset_dict)arch pretrainedrkwargsmodelZ weight_pathparamrrr _mobilenet_v3s   r}cKt d||d|}|S)aMobileNetV3 Small architecture model from `"Searching for MobileNetV3" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. scale (float, optional): Scale of channels in each layer. Default: 1.0. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`MobileNetV3Small `. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Small architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import mobilenet_v3_small >>> # Build model >>> model = mobilenet_v3_small() >>> # Build model and load imagenet pretrained weight >>> # model = mobilenet_v3_small(pretrained=True) >>> # Build mobilenet v3 small model with scale=0.5 >>> model = mobilenet_v3_small(scale=0.5) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] mobilenet_v3_smallrryN)rr}ryrrzr{rrrr!rcKr~)aMobileNetV3 Large architecture model from `"Searching for MobileNetV3" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. scale (float, optional): Scale of channels in each layer. Default: 1.0. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`MobileNetV3Large `. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Large architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import mobilenet_v3_large >>> # Build model >>> model = mobilenet_v3_large() >>> # Build model and load imagenet pretrained weight >>> # model = mobilenet_v3_large(pretrained=True) >>> # Build mobilenet v3 large model with scale=0.5 >>> model = mobilenet_v3_large(scale=0.5) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] rurN)rurrrrrrurru)Fr%) functoolsrrCrZpaddle.utils.downloadropsrZ_utilsr __all__rvZLayerr r$r4rGr]rnr}rrurrrrs&      ($E\5 K '