\;jC#ddlZddlmZddlmZddlmZgZddiZGdd ejZ Gd d ejZ dd Z dd Z y)N)nn)get_weights_path_from_url)ConvNormActivationzmobilenetv1_1.0)zAhttps://paddle-hapi.bj.bcebos.com/models/mobilenetv1_1.0.pdparams 3033ab1975b1670bef51545feb65fc45c$eZdZfdZdZxZS)DepthwiseSeparablec t|t|t||zd|dt||z|_tt||zt||zddd|_y)N) kernel_sizestridepaddinggroupsr)r rr)super__init__rint_depthwise_conv_pointwise_conv)self in_channels out_channels1 out_channels2 num_groupsrscale __class__s iG:\00. PROJECTS\API\Inventory\templateJSON\kerjaOCR\Lib\site-packages\paddle/vision/models/mobilenetv1.pyrzDepthwiseSeparable.__init__ su 1   % &zE)*   2  % &  % &  cJ|j|}|j|}|S)N)rr)rxs rforwardzDepthwiseSeparable.forward<s'   #   #r)__name__ __module__ __qualname__rr! __classcell__rs@rr r s  8rr c*eZdZdZdfd ZdZxZS) MobileNetV1awMobileNetV1 model from `"MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications" `_. Args: scale (float, optional): Scale of channels in each layer. Default: 1.0. num_classes (int, optional): Output dim of last fc layer. If num_classes <= 0, last fc layer will not be defined. Default: 1000. with_pool (bool, optional): Use pool before the last fc layer or not. Default: True. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV1 model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import MobileNetV1 >>> model = MobileNetV1() >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] c t|||_g|_||_||_t dtd|zddd|_|jttd|zdddd|d }|jj||jttd|zdd dd|d }|jj||jttd |zd d d d|d }|jj||jttd |zd d d d|d }|jj||jttd |zd d d d|d }|jj||jttd |zd dd d|d } |jj| tdD]Z} |jttd|zdddd|dt| dzz } |jj| \|jttd|zdddd|d } |jj| |jttd|zdddd|d } |jj| |rtjd|_|dkDr(tj"td|z||_yy)Nr rr )r out_channelsr rr@)rrrrrrconv2_1)sublayernameconv2_2conv3_1conv3_2conv4_1iconv4_2conv5_iconv5_6conv6r)rrrdwsl num_classes with_poolrrconv1 add_sublayerr appendrangestrrAdaptiveAvgPool2D pool2d_avgLinearfc)rrr<r=dws21dws22dws31dws32dws41dws42itmpdws56dws6rs rrzMobileNetV1.__init__^s   &"'R%Z   !!'U O   "   !!'U O ! "   !!'e ,!! "   !!'e ,!! "   !!'e ,!! "   !!'e ,!! "   qA##+ #C%K 0"%"%" AE *$ C II  S !!!'e ,!" "     'u -"" !     2215DO ?iiD5L 1;?DG rc|j|}|jD] }||} |jr|j|}|jdkDr't j |d}|j|}|S)Nrr )r>r;r=rDr<paddleflattenrF)rr dwss rr!zMobileNetV1.forwardsm JJqM99CAA >>"A   a q!$A Ar)?iT)r"r#r$__doc__rr!r%r&s@rr(r(Bs6I@V rr(c tdi|}|rX|tvs J|dtt|dt|d}tj|}|j ||S)NzJ model do not have a pretrained model now, you should set pretrained=Falserr )r( model_urlsrrRload load_dict)arch pretrainedkwargsmodel weight_pathparams r _mobilenetrbsx  !& !E J  _V] ^ _ / t Q D!1!!4  K(  Lrc <tdt|z|fd|i|}|S)aMobileNetV1 from `"MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications" `_. 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:`MobileNetV1 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV1 model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import mobilenet_v1 >>> # Build model >>> model = mobilenet_v1() >>> # Build model and load imagenet pretrained weight >>> # model = mobilenet_v1(pretrained=True) >>> # build mobilenet v1 with scale=0.5 >>> model_scale = mobilenet_v1(scale=0.5) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] mobilenetv1_r)rbrB)r]rr^r_s r mobilenet_v1res4D U#Z 7< @F E Lr)F)FrU) rRrpaddle.utils.downloadropsr__all__rYLayerr r(rbrerXrrrjsW;$    Fr"((rj  %r