Assessment algorithm of scene complexity based on X-CENet

Fanshu Shen, Yan Wen, Z. Zuo
{"title":"Assessment algorithm of scene complexity based on X-CENet","authors":"Fanshu Shen, Yan Wen, Z. Zuo","doi":"10.1117/12.2653877","DOIUrl":null,"url":null,"abstract":"In order to realize the rapid perception of complex scenes, the traditional scene complexity assessment algorithm has strong limitations in feature representation and scope of application, and it is difficult to deal with complex scenes. However, the existing deep network methods are lack of the consideration of the correlation between the underlying features of gray image and the complexity level, and the amount of parameters is too high to meet the needs of rapid response in practical applications. Based on the deep separable convolution module and residual connection structure, this paper designs a lightweight complexity assessment network X-CENet with stronger feature expression ability. A dense connection module which makes full use of multi-level features is introduced to improve the feature expression ability of the network for scene images. The underlying information such as image texture is particularly important for the assessment of complexity, so the feature cascade layer of the head and tail of the main modules is added to strengthen the utilization of the underlying feature information in the network. Experiments show that compared with other deep networks, this method can obtain higher assessment accuracy in the dimensions of image characteristics and detection performance with smaller parameters. Compared with the Inception V3 with similar parameter amount, this method improves the LCC index by 2.849% and the SRCC index by 3.338%.","PeriodicalId":253792,"journal":{"name":"Conference on Optics and Communication Technology","volume":"48 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Conference on Optics and Communication Technology","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1117/12.2653877","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

Abstract

In order to realize the rapid perception of complex scenes, the traditional scene complexity assessment algorithm has strong limitations in feature representation and scope of application, and it is difficult to deal with complex scenes. However, the existing deep network methods are lack of the consideration of the correlation between the underlying features of gray image and the complexity level, and the amount of parameters is too high to meet the needs of rapid response in practical applications. Based on the deep separable convolution module and residual connection structure, this paper designs a lightweight complexity assessment network X-CENet with stronger feature expression ability. A dense connection module which makes full use of multi-level features is introduced to improve the feature expression ability of the network for scene images. The underlying information such as image texture is particularly important for the assessment of complexity, so the feature cascade layer of the head and tail of the main modules is added to strengthen the utilization of the underlying feature information in the network. Experiments show that compared with other deep networks, this method can obtain higher assessment accuracy in the dimensions of image characteristics and detection performance with smaller parameters. Compared with the Inception V3 with similar parameter amount, this method improves the LCC index by 2.849% and the SRCC index by 3.338%.
基于X-CENet的场景复杂度评估算法
为了实现对复杂场景的快速感知,传统的场景复杂性评估算法在特征表示和应用范围上存在较强的局限性,难以处理复杂场景。然而,现有的深度网络方法缺乏对灰度图像底层特征与复杂度之间的相关性的考虑,且参数量过高,无法满足实际应用中快速响应的需要。基于深度可分离卷积模块和残差连接结构,设计了一个具有更强特征表达能力的轻量级复杂性评估网络X-CENet。为了提高网络对场景图像的特征表达能力,引入了充分利用多层次特征的密集连接模块。图像纹理等底层信息对于复杂度的评估尤为重要,因此增加了主模块的头部和尾部的特征级联层,加强了网络中底层特征信息的利用。实验表明,与其他深度网络相比,该方法可以在较小的参数下获得更高的图像特征维度评估精度和检测性能。与参数量相近的Inception V3相比,该方法的LCC指数提高了2.849%,SRCC指数提高了3.338%。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信