block29-CSDN博客

本文链接：https://blog.csdn.net/luoganttcc/article/details/89383835

本文详细介绍使用Keras构建深度学习模型的过程，包括多GPU配置、数据集读取及预处理，通过定制化的block unit模块实现卷积神经网络，最终完成模型训练与评估。文章深入探讨了模型各层的配置与优化策略。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Mon Dec  3 16:35:49 2018

@author: lg
"""

from keras.layers import Reshape,Convolution2D,PReLU,ZeroPadding2D,BatchNormalization
from keras.layers import AveragePooling2D,Dropout,Dense,GlobalAveragePooling2D
from keras.models import load_model
from keras.models import Model
from keras.layers import Input, Add, Activation, ZeroPadding2D,concatenate
from keras.layers import TimeDistributed,LSTM
import numpy as np
import pandas as pd
import tensorflow as tf
from keras.models import load_model

from mongo_data import data_set


from keras.utils import multi_gpu_model
import config
time_span=config.time_span()
feature=config.feature_number()
gpu_num=config.gpu_number()

input_shape = (time_span, feature, 1)
X_input = Input(input_shape)

def block_unit(filt,x):
    
    conv3=ZeroPadding2D(padding=(1, 1))(x)
    conv3=Convolution2D(filt, 3, 3, border_mode='same',init='glorot_uniform')(conv3)
    #model.add(Activation('relu'))
    l3=PReLU()(conv3)
    l3=BatchNormalization()(l3)

    conv4=ZeroPadding2D(padding=(1, 1))(l3)
    conv4=Convolution2D(filt, 3, 3, border_mode='same',init='glorot_uniform')(conv4)
    #model.add(Activation('relu'))
    l4=PReLU()(conv4)
    l4=BatchNormalization()(l4)

    m4=AveragePooling2D((3, 3), strides=(3, 3))(l4)
    d4=Dropout(0.25)(m4)
    
    return d4


def get_model(X_input):#重新建立模型，与原来不一样的是这里inp是传入
    n_classes = 5
    
    x=block_unit(32,X_input)
    
    x=block_unit(64,x)
    x=block_unit(128,x)
#    x=block_unit(256,x)
    x=GlobalAveragePooling2D()(x)

    x=Dense(1024)(x)
    x=PReLU()(x)
    x=Dropout(0.5)(x)
    x=Dense(1024)(x)
    x=PReLU()(x)
    x=Dropout(0.5)(x)
    x=Dropout(0.5)(x)
    x=Dense(512)(x)
    x=PReLU()(x)
    x=Dropout(0.5)(x)
    result=Dense(n_classes, activation='softmax')(x)


#    result=g
    model=Model(inputs=X_input,outputs=result)
    return model