\;j@ddlZddlZddlmZddlmZddlmZmZmZm Z m Z m Z m Z ddl mZddlmZgZddd d d d ZGd dej&ZGddej&ZGddej&ZGddej&ZGddej&ZGddej&ZdZddZddZddZddZddZy) N)nn) ParamAttr)AdaptiveAvgPool2D AvgPool2D BatchNormConv2DDropoutLinear MaxPool2D)Uniform)get_weights_path_from_url)zYhttps://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet121_pretrained.pdparams db1b239ed80a905290fd8b01d3af08e4)zYhttps://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet161_pretrained.pdparams 62158869cb315098bd25ddbfd308a853)zYhttps://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet169_pretrained.pdparams 82cc7c635c3f19098c748850efb2d796)zYhttps://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet201_pretrained.pdparams 16ca29565a7712329cf9e36e02caaf58)zYhttps://paddle-imagenet-models-name.bj.bcebos.com/dygraph/DenseNet264_pretrained.pdparams 3270ce516b85370bba88cfdd9f60bff4) densenet121 densenet161 densenet169 densenet201 densenet264c.eZdZ dfd ZdZxZS) BNACConvLayerc t|t|||_t ||||||t d|_y)NactF in_channels out_channels kernel_sizestridepaddinggroups weight_attr bias_attr)super__init__r _batch_normrr_conv self num_channels num_filters filter_sizer!padr#r __class__s fG:\00. PROJECTS\API\Inventory\templateJSON\kerjaOCR\Lib\site-packages\paddle/vision/models/densenet.pyr'zBNACConvLayer.__init__;sE $\s;$$#!    cJ|j|}|j|}|SN)r(r)r+inputys r1forwardzBNACConvLayer.forwardSs$   U # JJqMr2rr:relu__name__ __module__ __qualname__r'r8 __classcell__r0s@r1rr:s    0r2rc$eZdZfdZdZxZS) DenseLayerct|||_t|||zddd|_t||z|ddd|_|rt |d|_yy)Nr:rr,r-r.r/r!downscale_in_infer)pmode)r&r'dropoutr bn_ac_func1 bn_ac_func2r dropout_func)r+r, growth_ratebn_sizerJr0s r1r'zDenseLayer.__init__Zss  (%+-  ) ;.#    ''8L MD  r2c|j|}|j|}|jr|j|}t j ||gd}|S)Nr:)axis)rKrLrJrMpaddleconcat)r+r6convs r1r8zDenseLayer.forwardqsR&% <<$$T*D}}eT]3 r2r<rAs@r1rCrCYsN.r2rCc(eZdZ dfd ZdZxZS) DenseBlockc t |||_g|_|}t |D]G}|jj |j |d|dzt||||||z}Iy)N_r:)r,rNrOrJ)r&r'rJdense_layer_funcrangeappend add_sublayerrC) r+r, num_layersrOrNrJname pre_channellayerr0s r1r'zDenseBlock.__init__{s  "" :&E  ! ! ( (!!fAeai[)%0$/ ' '  & 3K'r2c<|}|jD] }||} |Sr4)rY)r+r6rTfuncs r1r8zDenseBlock.forwards%))D:D* r2r4r<rAs@r1rVrVzsLP4,r2rVc$eZdZfdZdZxZS)TransitionLayercrt|t||ddd|_t ddd|_y)Nr:rrEr r!r")r&r'r conv_ac_funcr pool2d_avg)r+r,num_output_featuresr0s r1r'zTransitionLayer.__init__s< )%+  $!QGr2cJ|j|}|j|}|Sr4)rhrir5s r1r8zTransitionLayer.forwards%   e $ OOA r2r<rAs@r1rdrds Hr2rdc.eZdZ dfd ZdZxZS) ConvBNLayerc t|t||||||td|_t |||_y)NFrr)r&r'rrr)rr(r*s r1r'zConvBNLayer.__init__sG $$#!    %[c:r2cJ|j|}|j|}|Sr4)r)r(r5s r1r8zConvBNLayer.forwards% JJu    Q r2r9r<rAs@r1rmrms   ;0r2rmc4eZdZdZ dfd ZdZxZS)DenseNetaDenseNet model from `"Densely Connected Convolutional Networks" `_. Args: layers (int, optional): Layers of DenseNet. Default: 121. bn_size (int, optional): Expansion of growth rate in the middle layer. Default: 4. dropout (float, optional): Dropout rate. Default: :math:`0.0`. 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 DenseNet model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import DenseNet >>> # Build model >>> densenet = DenseNet() >>> x = paddle.rand([1, 3, 224, 224]) >>> out = densenet(x) >>> print(out.shape) [1, 1000] ct|||_||_gd}||vs Jd|d|ddgdfddgd fddgd fddgd fddgd fd}||\}} } t d |ddd d|_t d dd|_| |_g|_ g|_ |} |} t| D]\} }|jj|jd| dzt| ||| |dt| dzz| || zz} | } | t!| dz k7sq|jj|jd| dzdt#| | dz| dz} | dz} t%| d|_|jrt)d|_|jdkDrMdt-j.| dzz }t1| |t3t5| |t3|_yy)N)yzsupported layers are z but input layer is @ ) `0)rzr{$r|)rzr{ryry)rzr{rry)rzr{rxrrFrfr;)r,r-r.r!r/rr:rgdb_conv_rT)r,r]rOrNrJr^tr_conv_blk)r,rjrrg?) initializer)r$r%)r&r' num_classes with_poolrm conv1_funcr pool2d_max block_configdense_block_func_listtransition_func_list enumerater[r\rVstrlenrdr batch_normrrimathsqrtr rr out)r+layersrOrJrrsupported_layers densenet_specnum_init_featuresrNrpre_num_channels num_featuresir]stdvr0s r1r'zDenseNet.__init__sX &"4 & & R "#3"44H Q R &b/*b/*b/*b/*b/*  8EV7L4; %)  $!QG(%'"$&!,( &|4MAz  & & - -!!q1ug&%5#- '$/ '#c!a%j0   (*{*BBL+ C %))))00%%!!a%-')90<0A$01#4 +q0 ;5>$Lf= >>/2DO   a <##566D%'4%2FG#+ DH r2c|j|}|j|}t|jD]K\}}|j||}|t |jdz k7s7|j ||}M|j|}|jr|j|}|jdkDr)tjdd}|j|}S)Nr:r) start_axis stop_axis)rrrrrrrrrrirrRflattenr)r+r6rTrr]r7s r1r8zDenseNet.forward<su%t$&t'8'89MAz04--a06DC))*Q..3t003D9: t$ >>%A   a qQ"=A Ar2)rsgiT)r=r>r?__doc__r'r8r@rAs@r1rqrqs&@ Pdr2rqc tdd|i|}|rX|tvs J|dtt|dt|d}tj|}|j ||S)NrzJ model do not have a pretrained model now, you should set pretrained=Falserr:)rq model_urlsr rRloadset_dict)archr pretrainedkwargsmodel weight_pathparams r1 _densenetrQs}  -F -f -E J  _V] ^ _ / t Q D!1!!4  K( u Lr2c tdd|fi|S)aDenseNet 121-layer model from `"Densely Connected Convolutional Networks" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`DenseNet `. Returns: :ref:`api_paddle_nn_Layer`. An instance of DenseNet 121-layer model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import densenet121 >>> # Build model >>> model = densenet121() >>> # Build model and load imagenet pretrained weight >>> # model = densenet121(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] rrsrrrs r1rra< ]C >v >>r2c tdd|fi|S)aDenseNet 161-layer model from `"Densely Connected Convolutional Networks" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`DenseNet `. Returns: :ref:`api_paddle_nn_Layer`. An instance of DenseNet 161-layer model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import densenet161 >>> # Build model >>> model = densenet161() >>> # Build model and load imagenet pretrained weight >>> # model = densenet161(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] rrtrrs r1rrrr2c tdd|fi|S)aDenseNet 169-layer model from `"Densely Connected Convolutional Networks" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`DenseNet `. Returns: :ref:`api_paddle_nn_Layer`. An instance of DenseNet 169-layer model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import densenet169 >>> # Build model >>> model = densenet169() >>> # Build model and load imagenet pretrained weight >>> # model = densenet169(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] rrurrs r1rrrr2c tdd|fi|S)aDenseNet 201-layer model from `"Densely Connected Convolutional Networks" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`DenseNet `. Returns: :ref:`api_paddle_nn_Layer`. An instance of DenseNet 201-layer model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import densenet201 >>> # Build model >>> model = densenet201() >>> # Build model and load imagenet pretrained weight >>> # model = densenet201(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] rrvrrs r1rrs: ]C >v >>r2c tdd|fi|S)aDenseNet 264-layer model from `"Densely Connected Convolutional Networks" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`DenseNet `. Returns: :ref:`api_paddle_nn_Layer`. An instance of DenseNet 264-layer model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import densenet264 >>> # Build model >>> model = densenet264() >>> # Build model and load imagenet pretrained weight >>> # model = densenet264(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] rrwrrs r1rrrr2)F) rrRrpaddle.base.param_attrr paddle.nnrrrrr r r paddle.nn.initializerr paddle.utils.downloadr __all__rLayerrrCrVrdrmrqrrrrrrrr2r1rs ,*; # 0BHH>B<bhh("((>CrxxCL  ?B?B?B?@?r2