ó ¾÷Xc@@sŽdZddlmZddlmZddlZddlmZddlmZddlm Z dd lm Z dd lm Z dd lm Z dd lm Z dd lmZddlmZddlmZddlmZddlmZddlmZddlmZddlmZddlmZddlmZdZdZdddfed„Z e!deeedd„Z"d„Z#dS( s]Inception V3 model for Keras. Note that the input image format for this model is different than for the VGG16 and ResNet models (299x299 instead of 224x224), and that the input preprocessing function is also different (same as Xception). # Reference - [Rethinking the Inception Architecture for Computer Vision](http://arxiv.org/abs/1512.00567) i(tprint_function(tabsolute_importNi(tModel(tlayers(t Activation(tDense(tInput(tBatchNormalization(tConv2D(t MaxPooling2D(tAveragePooling2D(tGlobalAveragePooling2D(tGlobalMaxPooling2D(tget_source_inputs(tconvert_all_kernels_in_model(tget_file(tbackendi(tdecode_predictions(t_obtain_input_shapeszhttps://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels.h5s€https://github.com/fchollet/deep-learning-models/releases/download/v0.5/inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5tsamec C@sÀ|d k r#|d}|d}n d }d }tjƒdkrJd} nd} t|||fd|d|dtd |ƒ|ƒ}td | d td |ƒ|ƒ}td d |ƒ|ƒ}|S(sUtility function to apply conv + BN. # Arguments x: input tensor. filters: filters in `Conv2D`. num_row: height of the convolution kernel. num_col: width of the convolution kernel. padding: padding mode in `Conv2D`. strides: strides in `Conv2D`. name: name of the ops; will become `name + '_conv'` for the convolution and `name + '_bn'` for the batch norm layer. # Returns Output tensor after applying `Conv2D` and `BatchNormalization`. t_bnt_convtchannels_firstiitstridestpaddingtuse_biastnametaxistscaletreluN(tNonetKtimage_data_formatRtFalseRR( txtfilterstnum_rowtnum_colRRRtbn_namet conv_nametbn_axis((s>/tmp/pip-build-isqEY4/keras/keras/applications/inception_v3.pyt conv2d_bn)s"    !timagenetièc C@s¿ |dd?hkr!tdƒ‚n|dkrN|rN|dkrNtdƒ‚nt|ddddd tjƒd |ƒ}|d?kr–td |ƒ}ntd |d |ƒ}tjƒd krÆd}nd}t|ddddd@ddƒ}t|dddddƒ}t|dddƒ}tdAddBƒ|ƒ}t|dddddƒ}t|dddddƒ}tdCddDƒ|ƒ}t|dddƒ} t|dddƒ} t| dddƒ} t|dddƒ} t| dddƒ} t| dddƒ} tdEddFddƒ|ƒ} t| dddƒ} t j | | | | gd|ddƒ}t|dddƒ} t|dddƒ} t| dddƒ} t|dddƒ} t| dddƒ} t| dddƒ} tdGddHddƒ|ƒ} t| dddƒ} t j | | | | gd|ddƒ}t|dddƒ} t|dddƒ} t| dddƒ} t|dddƒ} t| dddƒ} t| dddƒ} tdIddJddƒ|ƒ} t| dddƒ} t j | | | | gd|dd ƒ}t|d!ddddKddƒ} t|dddƒ} t| dddƒ} t| dddddLddƒ} tdMddNƒ|ƒ} t j | | | gd|dd"ƒ}t|dddƒ} t|d#ddƒ}t|d#dd$ƒ}t|dd$dƒ}t|d#ddƒ}t|d#d$dƒ}t|d#dd$ƒ}t|d#d$dƒ}t|ddd$ƒ}tdOddPddƒ|ƒ} t| dddƒ} t j | ||| gd|dd%ƒ}x9t dƒD]+}t|dddƒ} t|d&ddƒ}t|d&dd$ƒ}t|dd$dƒ}t|d&ddƒ}t|d&d$dƒ}t|d&dd$ƒ}t|d&d$dƒ}t|ddd$ƒ}tdQddRddƒ|ƒ} t| dddƒ} t j | ||| gd|dd't d|ƒƒ}q×Wt|dddƒ} t|dddƒ}t|ddd$ƒ}t|dd$dƒ}t|dddƒ}t|dd$dƒ}t|ddd$ƒ}t|dd$dƒ}t|ddd$ƒ}tdSddTddƒ|ƒ} t| dddƒ} t j | ||| gd|dd(ƒ}t|dddƒ} t| d)ddddUddƒ} t|dddƒ}t|ddd$ƒ}t|dd$dƒ}t|dddddVddƒ}tdWddXƒ|ƒ} t j | || gd|dd*ƒ}xjt dƒD]\}t|d)ddƒ} t|d!ddƒ} t| d!ddƒ}t| d!ddƒ}t j ||gd|dd+t |ƒƒ} t|d,ddƒ} t| d!ddƒ} t| d!ddƒ}t| d!ddƒ}t j ||gd|ƒ} tdYddZddƒ|ƒ} t| dddƒ} t j | | | | gd|dd't d-|ƒƒ}qüW|r˜ t dd.ƒ|ƒ}t|d/d0dd1ƒ|ƒ}n<|d2kr¶ t ƒ|ƒ}n|d3krÔ tƒ|ƒ}n|d?k rï t|ƒ}n|}t||dd4ƒ}|dkr» tjƒd krM tjƒd5krM tjd6ƒqM n|rq td7td8d9d:d;ƒ}ntd<td8d9d:d=ƒ}|j|ƒtjƒd>kr» t|ƒq» n|S([s³Instantiates the Inception v3 architecture. Optionally loads weights pre-trained on ImageNet. Note that when using TensorFlow, for best performance you should set `image_data_format="channels_last"` in your Keras config at ~/.keras/keras.json. The model and the weights are compatible with both TensorFlow and Theano. The data format convention used by the model is the one specified in your Keras config file. Note that the default input image size for this model is 299x299. # Arguments include_top: whether to include the fully-connected layer at the top of the network. weights: one of `None` (random initialization) or "imagenet" (pre-training on ImageNet). input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. input_shape: optional shape tuple, only to be specified if `include_top` is False (otherwise the input shape has to be `(299, 299, 3)` (with `channels_last` data format) or `(3, 299, 299)` (with `channels_first` data format). It should have exactly 3 inputs channels, and width and height should be no smaller than 139. E.g. `(150, 150, 3)` would be one valid value. pooling: Optional pooling mode for feature extraction when `include_top` is `False`. - `None` means that the output of the model will be the 4D tensor output of the last convolutional layer. - `avg` means that global average pooling will be applied to the output of the last convolutional layer, and thus the output of the model will be a 2D tensor. - `max` means that global max pooling will be applied. classes: optional number of classes to classify images into, only to be specified if `include_top` is True, and if no `weights` argument is specified. # Returns A Keras model instance. # Raises ValueError: in case of invalid argument for `weights`, or invalid input shape. R*spThe `weights` argument should be either `None` (random initialization) or `imagenet` (pre-training on ImageNet).ièsSIf using `weights` as imagenet with `include_top` as true, `classes` should be 1000t default_sizei+tmin_sizei‹t data_formatt include_toptshapettensorRiii RiRtvalidi@iPiÀi0ii`RRRtmixed0tmixed1tmixed2i€tmixed3i€itmixed4i tmixedtmixed7i@tmixed8tmixed9_iÀi tavg_poolt activationtsoftmaxt predictionstavgtmaxt inception_v3t tensorflowsòYou are using the TensorFlow backend, yet you are using the Theano image data format convention (`image_data_format="channels_first"`). For best performance, set `image_data_format="channels_last"` in your Keras config at ~/.keras/keras.json.s2inception_v3_weights_tf_dim_ordering_tf_kernels.h5t cache_subdirtmodelstmd5_hasht 9a0d58056eeedaa3f26cb7ebd46da564s8inception_v3_weights_tf_dim_ordering_tf_kernels_notop.h5t bcbd6486424b2319ff4ef7d526e38f63ttheanoN(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(ii(Rt ValueErrorRRR RR)R R Rt concatenatetrangetstrR RR R RRtwarningstwarnRt WEIGHTS_PATHtWEIGHTS_PATH_NO_TOPt load_weightsR(R.tweightst input_tensort input_shapetpoolingtclassest img_inputt channel_axisR"t branch1x1t branch5x5t branch3x3dblt branch_poolt branch3x3t branch7x7t branch7x7dbltit branch7x7x3t branch3x3_1t branch3x3_2tbranch3x3dbl_1tbranch3x3dbl_2tinputstmodelt weights_path((s>/tmp/pip-build-isqEY4/keras/keras/applications/inception_v3.pyt InceptionV3Us^7    !   !  %!       cC@s"|d:}|d8}|d9}|S(Ngào@gà?g@((R"((s>/tmp/pip-build-isqEY4/keras/keras/applications/inception_v3.pytpreprocess_input…s   ($t__doc__t __future__RRRMRDRtRRRRRRR R R R tengine.topologyR tutils.layer_utilsRtutils.data_utilsRRRtimagenet_utilsRRRORPRR)tTrueRiRj(((s>/tmp/pip-build-isqEY4/keras/keras/applications/inception_v3.pyt sB   & ÿ,