keras densenet设计

最新推荐文章于 2023-08-29 15:54:36 发布

zoot

最新推荐文章于 2023-08-29 15:54:36 发布

阅读量1.1k

点赞数

分类专栏： keras

本文链接：https://blog.csdn.net/zz_hh_uu_/article/details/84952731

版权

keras 专栏收录该内容

2 篇文章 0 订阅

订阅专栏

针对定长文字图片的设计

from keras.models import Model
from keras.layers.core import Dense, Dropout, Activation, Reshape, Permute
from keras.layers.convolutional import Conv2D, Conv2DTranspose, ZeroPadding2D
from keras.layers.pooling import AveragePooling2D, GlobalAveragePooling2D
from keras.layers import Input, Flatten
from keras.layers.merge import concatenate
from keras.layers.normalization import BatchNormalization
from keras.regularizers import l2
from keras.layers.wrappers import TimeDistributed


def conv_block(input, growth_rate, dropout_rate=None, weight_decay=1e-4):
    x = BatchNormalization(axis=-1, epsilon=1.1e-5)(input)
    x = Activation('relu')(x)
    x = Conv2D(growth_rate, (3,3), kernel_initializer='he_normal', padding='same')(x)
    if(dropout_rate):
        x = Dropout(dropout_rate)(x)
    return x

def dense_block(x, nb_layers, nb_filter, growth_rate, droput_rate=0.2, weight_decay=1e-4):
    for i in range(nb_layers):
        cb = conv_block(x, growth_rate, droput_rate, weight_decay)
        x = concatenate([x, cb], axis=-1)
        nb_filter += growth_rate
    return x, nb_filter

def transition_block(input, nb_filter, dropout_rate=None, pooltype=1, weight_decay=1e-4):
    x = BatchNormalization(axis=-1, epsilon=1.1e-5)(input)
    x = Activation('relu')(x)
    x = Conv2D(nb_filter, (1, 1), kernel_initializer='he_normal', padding='same', use_bias=False,
               kernel_regularizer=l2(weight_decay))(x)

    if(dropout_rate):
        x = Dropout(dropout_rate)(x)

    if(pooltype == 2):
        x = AveragePooling2D((2, 2), strides=(2, 2))(x)
    elif(pooltype == 1):
        x = ZeroPadding2D(padding = (0, 1))(x)
        x = AveragePooling2D((2, 2), strides=(2, 1))(x)
    elif(pooltype == 3):
        x = AveragePooling2D((2, 2), strides=(2, 1))(x)
    return x, nb_filter

def dense_cnn(input, nclass):

    _dropout_rate = 0.2 
    _weight_decay = 1e-4

    _nb_filter = 64
    # conv 64 5*5 s=2
    x = Conv2D(_nb_filter, (5, 5), strides=(2, 2), kernel_initializer='he_normal', padding='same',
               use_bias=False, kernel_regularizer=l2(_weight_decay))(input)
   
    # 64 + 8 * 8 = 128
    x, _nb_filter = dense_block(x, 8, _nb_filter, 8, None, _weight_decay)
    # 128
    x, _nb_filter = transition_block(x, 128, _dropout_rate, 2, _weight_decay)

    # 128 + 8 * 8 = 192
    x, _nb_filter = dense_block(x, 8, _nb_filter, 8, None, _weight_decay)
    # 192 -> 128
    x, _nb_filter = transition_block(x, 128, _dropout_rate, 2, _weight_decay)

    # 128 + 8 * 8 = 192
    x, _nb_filter = dense_block(x, 8, _nb_filter, 8, None, _weight_decay)

    x = BatchNormalization(axis=-1, epsilon=1.1e-5)(x)
    x = Activation('relu')(x)
    x = Permute((2, 1, 3), name='permute')(x)
    x = TimeDistributed(Flatten(), name='flatten')(x)
    y_pred = Dense(nclass, name='out', activation='softmax')(x)

    # basemodel = Model(inputs=input, outputs=y_pred)
    # basemodel.summary()

    return y_pred

def dense_blstm(input):
    pass

input = Input(shape=(32, 280, 1), name='the_input')
dense_cnn(input, 7000)

说明

x = Permute((2, 1, 3), name=‘permute’)(x) ，操作前样本维度(None,4,35,192),操作后样本维度(None,35,4,192)，其中样本None理解为样本个数或者批次，192是图片通道数，中间两个4，35是图片高和宽
x = TimeDistributed(Flatten(), name=‘flatten’)(x)，将(None,35,4,192)作为输入，操作后变为(None,35,768),其中768=192*4
y_pred = Dense(nclass, name=‘out’, activation=‘softmax’)(x) ，操作后维度是(?, 35, 7000)

网络结构

第一个层是卷积操作，卷积输出通道64，卷积模版为5*5大小，步长沿x方向和y方向分别为2和2，随着接着第二个dense block层。
第二大层dense block，分为8小层，每层又进行BN->Activation->Conv2D->Dropout操作和沿着通道拼接操作，其中保持通道数等于growth_rate，_nb_filter = nb_filter + growth_rate8 = 64 + 88 = 128，其它参数也不变。
第三层是transition block层，里面有4个操作，即BN->Activation->Conv2D->AveragePooling2D，卷积输出通道数设置为_nb_filter = 128。
第四层是dense block，_nb_filter = nb_filter + growth_rate8 = 128 + 88 = 192。每层又进行BN->Activation->Conv2D->Dropout操作和沿着通道拼接操作，输出数据通道是192。
第五层transition block层，里面有4个操作，即BN->Activation->Conv2D->AveragePooling2D，卷积输出通道数依然设置为_nb_filter = 128。
第六层是dense block层，每层又进行BN->Activation->Conv2D->Dropout操作和沿着通道拼接操作，_nb_filter = nb_filter + growth_rate8 = 128 + 88 = 192。输出数据通道是192，_nb_filter 增长为192。
第七层是rnn层，分别是BN->Activation->Permute->TimeDistributed->Dense操作。