drcn.py 文件源码

def create_convnet(self, _input, dense_dim=1000, dy=10, nb_filters=[64, 128], kernel_size=(3, 3), pool_size=(2, 2), 
        dropout=0.5, bn=True, output_activation='softmax', opt='adam'):

        """
        Create convnet model / encoder of DRCN

        Args:
            _input (Tensor)        : input layer
            dense_dim (int)            : dimensionality of the final dense layers 
            dy (int)           : output dimensionality
            nb_filter (list)       : list of #Conv2D filters
            kernel_size (tuple)    : Conv2D kernel size
            pool_size (tuple)      : MaxPool kernel size
            dropout (float)        : dropout rate
            bn (boolean)           : batch normalization mode
            output_activation (string) : act. function for output layer
            opt (string)           : optimizer

        Store the shared layers into self.enc_functions list
        """

        _h = _input

        self.enc_functions = [] # to store the shared layers, will be used later for constructing conv. autoencoder
        for i, nf in enumerate(nb_filters):
            enc_f = Conv2D(nf, kernel_size, padding='same')
            _h = enc_f(_h)
            self.enc_functions.append(enc_f)

            _h = Activation('relu')(_h)

            if i < 2:
                _h = MaxPooling2D(pool_size=pool_size, padding='same')(_h)

        _h = Flatten()(_h)      

        enc_f = Dense(dense_dim)
        _h = enc_f(_h) 
        self.enc_functions.append(enc_f)
        if bn:
            _h = BatchNormalization()(_h)
        _h = Activation('relu')(_h)
        _h = Dropout(dropout)(_h)

        enc_f = Dense(dense_dim)
        _h = enc_f(_h)
        self.enc_functions.append(enc_f)
        if bn:
            _h = BatchNormalization()(_h)
        _feat = Activation('relu')(_h)
        _h = Dropout(dropout)(_feat)

        _y = Dense(dy, activation=output_activation)(_h)

        # convnet
        self.convnet_model = Model(input=_input, output=_y)
        self.convnet_model.compile(loss='categorical_crossentropy', optimizer=opt)
        print(self.convnet_model.summary())

        self.feat_model = Model(input=_input, output=_feat)