DenseNet学习与tensorflow2实现

《Densely Connected Convolutional Networks》提出了DenseNet，它用前馈的方式连接每一层与所有其他层，网络的核心结构为如下所示的Dense块，在每一个Dense块中，存在多个Dense层，即下图所示的H1～H4。各Dense层之间彼此均相互连接。

不一定需要在Dense块中对每一个Dense层均直接进行相互连接，来缩小网络的结构；也可能可以在不相邻的Dense块之间通过简单的下采样操作进行连接，进一步提升网络对不同尺度的特征的利用效率。
基本架构如下：

下面我将使用tensorflow2实现：

# -*- coding: utf-8 -*-
"""
Created on Fri Feb 28 15:12:26 2020

@author: asus
"""

import tensorflow as tf
from tensorflow.keras.layers import Dense, Conv2D, GlobalAveragePooling2D,\
 BatchNormalization, AveragePooling2D, MaxPooling2D, Dropout, Concatenate
from tensorflow.keras import Model, Input

growth_rate = 12
inpt = Input(shape=(32,32,3))

def DenseLayer(x, nb_filter, bn_size=4, alpha=0.0, drop_rate=0.2):
    
    # Bottleneck layers
    x = BatchNormalization(axis=3)(x)
    x = tf.nn.leaky_relu(x, alpha=0.2)
    x = Conv2D(bn_size*nb_filter, (1, 1), strides=(1,1), padding='same')(x)
    
    # Composite function
    x = BatchNormalization(axis=3)(x)
    x = tf.nn.leaky_relu(x, alpha=0.2)
    x = Conv2D(nb_filter, (3, 3), strides=(1,1), padding='same')(x)
    
    if drop_rate: x = Dropout(drop_rate)(x)
    
    return x

def DenseBlock(x, nb_layers, growth_rate, drop_rate=0.2):
    
    for ii in range(nb_layers):
        conv = DenseLayer(x, nb_filter=growth_rate, drop_rate=drop_rate)
        x = Concatenate()([x, conv])
        
    return x


def TransitionLayer(x, compression=0.5, alpha=0.0, is_max=0):
    
    nb_filter = int(x.shape.as_list()[-1]*compression)
    x = BatchNormalization(axis=3)(x)
    x = tf.nn.leaky_relu(x, alpha=0.2)
    x = Conv2D(nb_filter, (1, 1), strides=(1,1), padding='same')(x)
    if is_max != 0: x = MaxPooling2D(pool_size=(2, 2), strides=2)(x)
    else: x = AveragePooling2D(pool_size=(2, 2), strides=2)(x)
    
    return x



#growth_rate = 12
#inpt = Input(shape=(32,32,3))
 
x = Conv2D(growth_rate*2, (3, 3), strides=1, padding='same')(inpt)
x = BatchNormalization(axis=3)(x)
x = tf.nn.leaky_relu(x, alpha=0.1)
 
x = DenseBlock(x, 12, growth_rate, drop_rate=0.2)
 
x = TransitionLayer(x)
 
x = DenseBlock(x, 12, growth_rate, drop_rate=0.2)
 
x = TransitionLayer(x)
 
x = DenseBlock(x, 12, growth_rate, drop_rate=0.2)
 
x = BatchNormalization(axis=3)(x)
x = GlobalAveragePooling2D()(x)
 
x = Dense(10, activation='softmax')(x)
 
model = Model(inpt, x)
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
 
model.summary()

参考文献：
https://blog.csdn.net/zhongqianli/article/details/86651797
https://blog.csdn.net/shi2xian2wei2/article/details/84425777

• 点赞
• 收藏
• 分享
• • 文章举报

到达起点 发布了34 篇原创文章 · 获赞 2 · 访问量 1267 私信 关注

标签：tensorflow2,0.2,Dense,nb,drop,学习,rate,growth,DenseNet
来源： https://blog.csdn.net/m0_37663944/article/details/104556663