【Keras-Inception-resnet v1】CIFAR-10

其他 2019-01-18 23:57:48 阅读次数: 0
版权声明：本文为博主原创文章，未经博主允许不得转载。 https://blog.csdn.net/bryant_meng/article/details/85640395

系列连载目录

  • 请查看博客 《Paper》 4.1 小节 【Keras】Classification in CIFAR-10 系列连载

学习借鉴

参考

代码下载

硬件

  • TITAN XP

文章目录

1 理论基础

参考【Inception-v4】《Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning》

2 Inception-resnet-v1 代码实现

[D] Why aren’t Inception-style networks successful on CIFAR-10/100?
在这里插入图片描述
在这里插入图片描述
图片来源 https://www.zhihu.com/question/50370954

2.1 Inception_resnet_v1

1）导入库，设置好超参数

import os  
os.environ["CUDA_DEVICE_ORDER"]="PCI_BUS_ID"   
os.environ["CUDA_VISIBLE_DEVICES"]="3" 

import keras
import numpy as np
import math

from keras.datasets import cifar10
from keras.layers import Conv2D, MaxPooling2D, AveragePooling2D, ZeroPadding2D, GlobalAveragePooling2D
from keras.layers import Flatten, Dense, Dropout,BatchNormalization,Activation, Convolution2D, add
from keras.models import Model
from keras.layers import Input, concatenate
from keras import optimizers, regularizers
from keras.preprocessing.image import ImageDataGenerator
from keras.initializers import he_normal
from keras.callbacks import LearningRateScheduler, TensorBoard, ModelCheckpoint

num_classes        = 10
batch_size         = 64         # 64 or 32 or other
epochs             = 300
iterations         = 782       
USE_BN=True
DROPOUT=0.2 # keep 80%
CONCAT_AXIS=3
weight_decay=1e-4
DATA_FORMAT='channels_last' # Theano:'channels_first' Tensorflow:'channels_last'

log_filepath  = './inception_resnet_v1'

2）数据预处理并设置 learning schedule

def color_preprocessing(x_train,x_test):
    x_train = x_train.astype('float32')
    x_test = x_test.astype('float32')
    mean = [125.307, 122.95, 113.865]
    std  = [62.9932, 62.0887, 66.7048]
    for i in range(3):
        x_train[:,:,:,i] = (x_train[:,:,:,i] - mean[i]) / std[i]
        x_test[:,:,:,i] = (x_test[:,:,:,i] - mean[i]) / std[i]
    return x_train, x_test

def scheduler(epoch):
    if epoch < 100:
        return 0.01
    if epoch < 200:
        return 0.001
    return 0.0001

# load data
(x_train, y_train), (x_test, y_test) = cifar10.load_data()
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test  = keras.utils.to_categorical(y_test, num_classes)
x_train, x_test = color_preprocessing(x_train, x_test)

3）定义网络结构

  • Factorization into smaller convolutions
    eg：3×3 → 3×1 + 1×3
def conv_block(x, nb_filter, nb_row, nb_col, border_mode='same', subsample=(1,1), bias=False):
    x = Convolution2D(nb_filter, nb_row, nb_col, subsample=subsample, border_mode=border_mode, bias=bias,
                     init="he_normal",dim_ordering='tf',W_regularizer=regularizers.l2(weight_decay))(x)
    x = BatchNormalization(momentum=0.9, epsilon=1e-5)(x)
    x = Activation('relu')(x)
    return x
  • stem（图6右边 in paper）
    在这里插入图片描述
    上图中，V 表示 padding 为 valid，没有标注的表示 same， 因为数据集分辨率的不同，如下代码全部用 same 表示：
def create_stem(img_input,concat_axis):
    x = Conv2D(32,kernel_size=(3,3),strides=(2,2),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(img_input)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(32,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(64,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = MaxPooling2D(pool_size=(3,3),strides=2,padding='same',data_format=DATA_FORMAT)(x)
    
    x = Conv2D(80,kernel_size=(1,1),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(192,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(256,kernel_size=(3,3),strides=(2,2),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    return x
  • inception module A（图4左边 in paper）
    在这里插入图片描述
def inception_A(x,params,concat_axis,padding='same',data_format=DATA_FORMAT,scale_residual=False,use_bias=True,kernel_initializer="he_normal",bias_initializer='zeros',kernel_regularizer=None,bias_regularizer=None,activity_regularizer=None,kernel_constraint=None,bias_constraint=None,weight_decay=weight_decay):
    (branch1,branch2,branch3,branch4)=params
    if weight_decay:
        kernel_regularizer=regularizers.l2(weight_decay)
        bias_regularizer=regularizers.l2(weight_decay)
    else:
        kernel_regularizer=None
        bias_regularizer=None
    #1x1
    pathway1=Conv2D(filters=branch1[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway1 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway1))
    #1x1->3x3
    pathway2=Conv2D(filters=branch2[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    pathway2=Conv2D(filters=branch2[1],kernel_size=(3,3),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway2)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    #1x1->3x3->3x3
    pathway3=Conv2D(filters=branch3[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway3 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway3))
    pathway3=Conv2D(filters=branch3[1],kernel_size=(3,3),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway3)
    pathway3 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway3))
    pathway3=Conv2D(filters=branch3[1],kernel_size=(3,3),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway3)
    pathway3 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway3))
    # concatenate
    pathway_123 = concatenate([pathway1,pathway2,pathway3],axis=concat_axis)
    pathway_123 = Conv2D(branch4[0],kernel_size=(1,1),strides=(1,1),padding='same',activation = 'linear',
                         kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(pathway_123)
    if scale_residual: 
        x = Lambda(lambda p: p * 0.1)(x)
    return add([x,pathway_123])

def reduce_A(x,params,concat_axis,padding='same',data_format=DATA_FORMAT,use_bias=True,kernel_initializer="he_normal",bias_initializer='zeros',kernel_regularizer=None,bias_regularizer=None,activity_regularizer=None,kernel_constraint=None,bias_constraint=None,weight_decay=weight_decay):
    (branch1,branch2)=params
    if weight_decay:
        kernel_regularizer=regularizers.l2(weight_decay)
        bias_regularizer=regularizers.l2(weight_decay)
    else:
        kernel_regularizer=None
        bias_regularizer=None
    #1x1
    pathway1 = Conv2D(filters=branch1[0],kernel_size=(3,3),strides=2,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway1 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway1))
    #1x1->3x3->3x3
    pathway2 = Conv2D(filters=branch2[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    pathway2 = Conv2D(filters=branch2[1],kernel_size=(3,3),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway2)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    pathway2 = Conv2D(filters=branch2[2],kernel_size=(3,3),strides=2,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway2)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    # max pooling
    pathway3 = MaxPooling2D(pool_size=(3,3),strides=2,padding=padding,data_format=DATA_FORMAT)(x)
    return concatenate([pathway1,pathway2,pathway3],axis=concat_axis)
  • inception module B（图4中间 in paper）
    在这里插入图片描述
def inception_B(x,params,concat_axis,padding='same',data_format=DATA_FORMAT,scale_residual=False,use_bias=True,kernel_initializer="he_normal",bias_initializer='zeros',kernel_regularizer=None,bias_regularizer=None,activity_regularizer=None,kernel_constraint=None,bias_constraint=None,weight_decay=weight_decay):
    (branch1,branch2,branch3)=params
    if weight_decay:
        kernel_regularizer=regularizers.l2(weight_decay)
        bias_regularizer=regularizers.l2(weight_decay)
    else:
        kernel_regularizer=None
        bias_regularizer=None
    #1x1
    pathway1=Conv2D(filters=branch1[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway1 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway1))
    #1x1->1x7->7x1
    pathway2=Conv2D(filters=branch2[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    pathway2 = conv_block(pathway2,branch2[1],1,7)
    pathway2 = conv_block(pathway2,branch2[2],7,1)
    # concatenate
    pathway_12 = concatenate([pathway1,pathway2],axis=concat_axis)
    pathway_12 = Conv2D(branch3[0],kernel_size=(1,1),strides=(1,1),padding='same',activation = 'linear',
                        kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(pathway_12)
    if scale_residual: 
        x = Lambda(lambda p: p * 0.1)(x)
    return add([x,pathway_12])
  • reduction B
    在这里插入图片描述
def reduce_B(x,params,concat_axis,padding='same',data_format=DATA_FORMAT,use_bias=True,kernel_initializer="he_normal",bias_initializer='zeros',kernel_regularizer=None,bias_regularizer=None,activity_regularizer=None,kernel_constraint=None,bias_constraint=None,weight_decay=weight_decay):
    (branch1,branch2,branch3)=params
    if weight_decay:
        kernel_regularizer=regularizers.l2(weight_decay)
        bias_regularizer=regularizers.l2(weight_decay)
    else:
        kernel_regularizer=None
        bias_regularizer=None
    #1x1->3x3
    pathway1 = Conv2D(filters=branch1[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway1 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway1)) 
    pathway1 = Conv2D(filters=branch1[1],kernel_size=(3,3),strides=2,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway1)
    pathway1 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway1))
    #1x1->3x3
    pathway2 = Conv2D(filters=branch2[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2)) 
    pathway2 = Conv2D(filters=branch2[1],kernel_size=(3,3),strides=2,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway2)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    #1x1->3x3->3x3
    pathway3 = Conv2D(filters=branch3[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway3 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway3))
    pathway3 = Conv2D(filters=branch3[1],kernel_size=(3,3),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway3)
    pathway3 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway3))
    pathway3 = Conv2D(filters=branch3[2],kernel_size=(3,3),strides=2,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(pathway3)
    pathway3 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway3))
    # max pooling
    pathway4 = MaxPooling2D(pool_size=(3,3),strides=2,padding=padding,data_format=DATA_FORMAT)(x)
    return concatenate([pathway1,pathway2,pathway3,pathway4],axis=concat_axis)
  • inception module C（图4右边 in paper）
    在这里插入图片描述
def inception_C(x,params,concat_axis,padding='same',data_format=DATA_FORMAT,scale_residual=False,use_bias=True,kernel_initializer="he_normal",bias_initializer='zeros',kernel_regularizer=None,bias_regularizer=None,activity_regularizer=None,kernel_constraint=None,bias_constraint=None,weight_decay=weight_decay):
    (branch1,branch2,branch3)=params
    if weight_decay:
        kernel_regularizer=regularizers.l2(weight_decay)
        bias_regularizer=regularizers.l2(weight_decay)
    else:
        kernel_regularizer=None
        bias_regularizer=None
    #1x1
    pathway1=Conv2D(filters=branch1[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway1 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway1))
    #1x1->1x3->3x1
    pathway2=Conv2D(filters=branch2[0],kernel_size=(1,1),strides=1,padding=padding,data_format=data_format,use_bias=use_bias,kernel_initializer=kernel_initializer,bias_initializer=bias_initializer,kernel_regularizer=kernel_regularizer,bias_regularizer=bias_regularizer,activity_regularizer=activity_regularizer,kernel_constraint=kernel_constraint,bias_constraint=bias_constraint)(x)
    pathway2 = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(pathway2))
    pathway2 = conv_block(pathway2,branch2[1],1,3)
    pathway2 = conv_block(pathway2,branch2[2],3,1)
    # concatenate
    pathway_12 = concatenate([pathway1,pathway2],axis=concat_axis)
    pathway_12 = Conv2D(branch3[0],kernel_size=(1,1),strides=(1,1),padding='same',activation = 'linear',
                        kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(pathway_12)
    if scale_residual: 
        x = Lambda(lambda p: p * 0.1)(x)
    return add([x,pathway_12])

4）搭建网络
用 3）中设计好的模块来搭建网络
整体结构（图6左边 in paper）
在这里插入图片描述

def create_model(img_input):
    # stem
    x = create_stem(img_input,concat_axis=CONCAT_AXIS)
    # 5 x inception_A
    for _ in range(5):
        x=inception_A(x,params=[(32,),(32,32),(32,32,32),(256,)],concat_axis=CONCAT_AXIS)
    # reduce A
    x=reduce_A(x,params=[(384,),(192,224,256)],concat_axis=CONCAT_AXIS) # 768
    # 10 x inception_B
    for _ in range(10):
        x=inception_B(x,params=[(128,),(128,128,128),(896,)],concat_axis=CONCAT_AXIS)
    # reduce B
    x=reduce_B(x,params=[(256,384),(256,256),(256,256,256)],concat_axis=CONCAT_AXIS) # 1280
    # 5 x inception_C
    for _ in range(5):
        x=inception_C(x,params=[(192,),(192,192,192),(1792,)],concat_axis=CONCAT_AXIS)
    x=GlobalAveragePooling2D()(x)
    x=Dropout(DROPOUT)(x)
    x = Dense(num_classes,activation='softmax',kernel_initializer="he_normal",
              kernel_regularizer=regularizers.l2(weight_decay))(x)
    return x

5）生成模型

img_input=Input(shape=(32,32,3))
output = create_model(img_input)
model=Model(img_input,output)
model.summary()

参数量如下：

Total params: 21,146,522
Trainable params: 21,119,418
Non-trainable params: 27,104

和 inception_v3 参数量差不多

Total params: 23,397,866
Trainable params: 23,359,978
Non-trainable params: 37,888

6）开始训练

# set optimizer
sgd = optimizers.SGD(lr=.1, momentum=0.9, nesterov=True)
model.compile(loss='categorical_crossentropy', optimizer=sgd, metrics=['accuracy'])

# set callback
tb_cb = TensorBoard(log_dir=log_filepath, histogram_freq=0)
change_lr = LearningRateScheduler(scheduler)
cbks = [change_lr,tb_cb]

# set data augmentation
datagen = ImageDataGenerator(horizontal_flip=True,
                             width_shift_range=0.125,
                             height_shift_range=0.125,
                             fill_mode='constant',cval=0.)
datagen.fit(x_train)

# start training
model.fit_generator(datagen.flow(x_train, y_train,batch_size=batch_size),
                    steps_per_epoch=iterations,
                    epochs=epochs,
                    callbacks=cbks,
                    validation_data=(x_test, y_test))
model.save('inception_resnet_v1.h5')

7）结果分析
training accuracy 和 training loss
在这里插入图片描述 在这里插入图片描述 在这里插入图片描述
在这里插入图片描述 在这里插入图片描述

  • accuracy
    在这里插入图片描述
  • loss
    在这里插入图片描述
    在训练集上效果还行，比 inception-v3 好

test accuracy 和 test loss
在这里插入图片描述 在这里插入图片描述 在这里插入图片描述
在这里插入图片描述 在这里插入图片描述

  • accuracy
    在这里插入图片描述
  • loss
    在这里插入图片描述
    …………
    结果意料之中，因为前面的 stem 把 feature map 的分辨率降到了 4（ 32 2 3 = 4 \frac{32}{2^3} =4 ），后面 inception module 中 1x7 与 7x1 的效果可想而知，都是在处理 padding， 差，相当的差。
    不要慌，问题不大，有了 inception_v1 to inception_v4 摸爬滚打的经验，Everything is under control.

2.2 Inception_resnet_v1_stem_slim

对2.1 小节中的 inception 模块函数 def create_stem（……）进行如下修改：
把 stride 为 2 的 convolution 和 pooling 都替换成 1，保持 feature map 的分辨率，让 1×7，7×1 convolution 有的放矢

def create_stem(img_input,concat_axis):
    x = Conv2D(32,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(img_input)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(32,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(64,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = MaxPooling2D(pool_size=(3,3),strides=1,padding='same',data_format=DATA_FORMAT)(x)
    
    x = Conv2D(80,kernel_size=(1,1),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(192,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    
    x = Conv2D(256,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(x)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))
    return x

其它代码同 inception_resnet_v1 参数量如下（不变）：

Total params: 21,146,522
Trainable params: 21,119,418
Non-trainable params: 27,104

  • Inception_resnet_v1
    Total params: 21,146,522
  • inception_v3
    Total params: 23,397,866

结果分析如下：
training accuracy 和 training loss
在这里插入图片描述 在这里插入图片描述 在这里插入图片描述
在这里插入图片描述在这里插入图片描述

  • accuracy
    在这里插入图片描述
  • loss
    在这里插入图片描述

test accuracy 和 test loss
在这里插入图片描述 在这里插入图片描述 在这里插入图片描述
在这里插入图片描述在这里插入图片描述

  • accuracy
    在这里插入图片描述
  • loss
    在这里插入图片描述
    结果还行，上到了 93%+

2.3 Inception_resnet_v1_slim

把 Inception_resnet_v1 中 stern 结构直接替换成一个卷积，inception 结构不变，因为stern结果会把原图降到1/8的分辨率，对于 ImageNet 还行，CIFRA-10的话有些吃不消了

def create_stem(img_input,concat_axis):
    x = Conv2D(256,kernel_size=(3,3),strides=(1,1),padding='same',
               kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(img_input)
    x = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(x))    
    return x

其它代码同 inception_resnet_v1
参数量如下，降低了一丢丢：

Total params: 20,537,194
Trainable params: 20,510,890
Non-trainable params: 26,304

  • Inception_resnet_v1
    Total params: 21,146,522
  • inception_resnet_v1_stem_slim
    Total params: 21,146,522
  • inception_v3
    Total params: 23,397,866

结果分析如下：
test accuracy 和 test loss
在这里插入图片描述
在这里插入图片描述 在这里插入图片描述

  • accuracy
    在这里插入图片描述
  • loss
    在这里插入图片描述

提升了一点点，没有突破94%

2.4 Inception_resnet_v1_slim_0.8

之前代码有写错过，所以对这个 dropout 参数不是很自信，做了一个dropout 的实验，把一开始超参数的设定中DROPOUT=0.2 # keep 80% 改为 0.8，tensorflow中一般设置的是 keep_drop = 0.8

结果分析如下：
test accuracy 和 test loss
在这里插入图片描述
在这里插入图片描述在这里插入图片描述
哈哈，这是马后炮耶，刚开始 inception_resnet_v1_slim 效果很差的， 所以才会做这个实验，后来发现代码写错了，更正后，差距就很明显了。好拉，确认 keras 中的 dropout 是丢弃的百分比了。

2.5 Inception_resnet_v1_slim_scale

在这里插入图片描述
论文中图7，之前 inception_resnet_v1 代码中也有这种结构

也即如下部分代码：

 if scale_residual: 
     x = Lambda(lambda p: p * 0.1)(x)

现在把这个功能激活，修改代码如下 scale_residual=True

def create_model(img_input):
    # stem
    x = create_stem(img_input,concat_axis=CONCAT_AXIS)
    # 5 x inception_A
    for _ in range(5):
        x=inception_A(x,params=[(32,),(32,32),(32,32,32),(256,)],concat_axis=CONCAT_AXIS,scale_residual=True)
    # reduce A
    x=reduce_A(x,params=[(384,),(192,224,256)],concat_axis=CONCAT_AXIS) # 768
    # 10 x inception_B
    for _ in range(10):
        x=inception_B(x,params=[(128,),(128,128,128),(896,)],concat_axis=CONCAT_AXIS,scale_residual=True)
    # reduce B
    x=reduce_B(x,params=[(256,384),(256,256),(256,256,256)],concat_axis=CONCAT_AXIS) # 1280
    # 5 x inception_C
    for _ in range(5):
        x=inception_C(x,params=[(192,),(192,192,192),(1792,)],concat_axis=CONCAT_AXIS,scale_residual=True)
    x=GlobalAveragePooling2D()(x)
    x=Dropout(DROPOUT)(x)
    x = Dense(num_classes,activation='softmax',kernel_initializer="he_normal",
              kernel_regularizer=regularizers.l2(weight_decay))(x)
    return x

参数量不变，如下所示：

Total params: 20,537,194
Trainable params: 20,510,890
Non-trainable params: 26,304

结果分析如下：
test accuracy 和 test loss
在这里插入图片描述
在这里插入图片描述在这里插入图片描述

  • accuracy
    在这里插入图片描述

哎呀，没有之前的好！可能需要调整下 hyper-parameters

2.6 Inception_resnet_v1_slim_ac

加辅助分类器，auxiliary classifier，具体参考 【Keras-Inception v1】CIFAR-10

在 inception module B 和 reduction B 之间，加入了 auxiliary classifier ，代码修改如下：

def create_model(img_input):
    # stem
    x = create_stem(img_input,concat_axis=CONCAT_AXIS)
    # 5 x inception_A
    for _ in range(5):
        x=inception_A(x,params=[(32,),(32,32),(32,32,32),(256,)],concat_axis=CONCAT_AXIS)
    # reduce A
    x=reduce_A(x,params=[(384,),(192,224,256)],concat_axis=CONCAT_AXIS) # 768
    
    # 10 x inception_B
    for _ in range(10):
        x=inception_B(x,params=[(128,),(128,128,128),(896,)],concat_axis=CONCAT_AXIS)
        
    # ac
    aux_out = AveragePooling2D(pool_size=(5,5),strides=3,padding='same',data_format=DATA_FORMAT)(x)
    aux_out = Conv2D(128,kernel_size=(1,1),strides=(1,1),padding='same',
                     kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(aux_out)
    aux_out = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(aux_out))   
    aux_out = Conv2D(768,kernel_size=(5,5),strides=(1,1),padding='same',
                     kernel_initializer="he_normal",kernel_regularizer=regularizers.l2(weight_decay))(aux_out)
    aux_out = Activation('relu')(BatchNormalization(momentum=0.9, epsilon=1e-5)(aux_out))       
    aux_out = Flatten()(aux_out)
    aux_out = Dense(num_classes,activation='softmax',kernel_initializer="he_normal",
                    kernel_regularizer=regularizers.l2(weight_decay))(aux_out)
    
    # reduce B
    x=reduce_B(x,params=[(256,384),(256,256),(256,256,256)],concat_axis=CONCAT_AXIS) # 1280
    # 5 x inception_C
    for _ in range(5):
        x=inception_C(x,params=[(192,),(192,192,192),(1792,)],concat_axis=CONCAT_AXIS)
    x=GlobalAveragePooling2D()(x)
    x=Dropout(DROPOUT)(x)
    x = Dense(num_classes,activation='softmax',kernel_initializer="he_normal",
              kernel_regularizer=regularizers.l2(weight_decay))(x)
    
    output = Add()([x,aux_out])
    return output

参数量如下所示：

Total params: 23,390,452
Trainable params: 23,362,356
Non-trainable params: 28,096

  • Inception_resnet_v1
    Total params: 21,146,522
  • inception_resnet_v1_stem_slim
    Total params: 21,146,522
  • inception_resnet_v1_slim_scale
    Total params: 20,537,194
  • inception_resnet_v1_slim
    Total params: 20,537,194
  • inception_v3
    Total params: 23,397,866

结果分析如下：
test accuracy 和 test loss
在这里插入图片描述
在这里插入图片描述在这里插入图片描述

  • accuracy
    在这里插入图片描述

对不起，打扰了，好像直接崩了！
在这里插入图片描述

3 总结

模型大小
在这里插入图片描述

猜你喜欢

转载自blog.csdn.net/bryant_meng/article/details/85640395
今日推荐
开源日报 | Chrome内置Gemini的意义不在于Gemini；中国AI追随之路的五大误区；ECharts创始人“下海”养鱼；谷歌I/O开发者大会什么都有，只是没有惊喜
微软回应中国区AI团队“打包赴美”传闻
基于大语言模型的开源知识库问答系统 MaxKB GitHub Star 数量突破 5,000 个！
美国拟限制 AI 大模型出口中国和俄罗斯
苹果将与 OpenAI 达成协议，将 ChatGPT 应用于 iPhone
openKylin 社区生态委员会第六次会议圆满召开
阿里云正式发布通义千问 2.5
Python 3.13 发布首个 Beta：实验性自由线程模式和 JIT、改进交互式解释器
Stack Overflow 拿我的代码去训练 AI 大模型，还封了我的账号
Pop!_OS 的 COSMIC 桌面完成 App Store 上架工作
《2024 年一季度互联网投融资运行情况》研究报告
报告：Django 仍然是 74% 开发者的首选
周排行
返回指定时间格式
fopen函数中的mode参数
Java 单例模式探讨
Flex remoteobject工作原理探讨
寻找mplayer的便捷安装方法
30天了解30种技术系列---(26)MySQL自动化运维工具Inception
关于Jboss/Tomcat/Jetty的JNDI定义123
程序减肥，strip，eu-strip 及其符号表
AsyncTask、View.post(Runnable)、ViewTreeObserver三种方式总结frame animation自动启动
Json和Bean的互相转换
每日归档
更多
2024-05-15(24)
2024-05-14(0)
2024-05-13(18)
2024-05-12(0)
2024-05-11(38)
2024-05-10(38)
2024-05-09(35)
2024-05-08(42)
2024-05-07(14)
2024-05-06(40)

代码天地 © 2018 codetd.com

境外服务器   最新电视剧2022