论文Automatic Multi-organ Segmentation on Abdominal CT with Dense V-networks架构代码

论文地址：https://ieeexplore.ieee.org/document/8291609/
论文模型架构如下：
总体和U-net架构很像，不同之处是使用了卷积单元，里面各卷积层使用全连接的方式。

DenseVnet.png

卷积单元代码如下：

    def Inception_dilation(self, inputs, f):        
        conv3 = Conv2D(f , (3, 3), padding='same', activation= 'selu', kernel_initializer = 'lecun_normal')(inputs)

        conv5 = Conv2D(f , (3, 3), padding='same', dilation_rate = (3, 3), activation='selu', kernel_initializer='lecun_normal')(inputs)

        conv7 = Conv2D(f, (3, 3), padding='same', dilation_rate = (5, 5), activation= 'selu', kernel_initializer = 'lecun_normal')(inputs)

        conv9 = Conv2D(f, (3, 3), padding='same', dilation_rate = (7, 7), activation= 'selu', kernel_initializer = 'lecun_normal')(inputs)
      
        merge2 = concatenate([conv3, conv5, conv7, conv9], concat_axis=3)
        return merge2
    def densevnet(self, inputs, f):
        conv3 = self.Inception_dilation(inputs, f)

        conv5 = self.Inception_dilation(conv3, f)
        
        merge1 = concatenate([conv3, conv5], concat_axis=3)
        
        conv7 = self.Inception_dilation(merge1, f)
        
        merge2 = concatenate([conv3, conv5, merge1], concat_axis=3)
        
        conv9 = self.Inception_dilation(merge2, f)
        
        merge3 = concatenate([conv3, conv5, conv7, conv9], concat_axis=3)
        return merge3    

总体架构代码可以参考U-net.