基于网络重参数化和特征自适应加权的交通标志检测器

IF 1.8 4区 计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE Journal of Ambient Intelligence and Smart Environments Pub Date : 2022-07-26 DOI:10.3233/ais-220038
Jianming Zhang, Zhuofan Zheng, Xianding Xie, Yan Gui, Gwang-Jun Kim
{"title":"基于网络重参数化和特征自适应加权的交通标志检测器","authors":"Jianming Zhang, Zhuofan Zheng, Xianding Xie, Yan Gui, Gwang-Jun Kim","doi":"10.3233/ais-220038","DOIUrl":null,"url":null,"abstract":"Traffic sign detection is a challenging task. Although existing deep learning techniques have made great progress in detecting traffic signs, there are still many unsolved challenges. We propose a novel traffic sign detection network named ReYOLO that learns rich contextual information and senses scale variations to efficiently detect small and ambiguous traffic signs in the wild. Specifically, we first replace the conventional convolutional block with modules that are built by structural reparameterization methods and are embedded into bigger structures, thus decoupling the training structures and the inference structures using parameter transformation, and allowing the model to learn more effective features. We then design a novel weighting mechanism which can be embedded into a feature pyramid to exploit foreground features at different scales to narrow the semantic gap between multiple scales. To fully evaluate the proposed method, we conduct experiments on a traditional traffic sign dataset GTSDB as well as two new traffic sign datasets TT100K and CCTSDB2021, achieving 97.2%, 68.3% and 83.9% mAP (Mean Average Precision) for the three-class detection challenge in these three datasets.","PeriodicalId":49316,"journal":{"name":"Journal of Ambient Intelligence and Smart Environments","volume":"19 7 1","pages":"317-334"},"PeriodicalIF":1.8000,"publicationDate":"2022-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"13","resultStr":"{\"title\":\"ReYOLO: A traffic sign detector based on network reparameterization and features adaptive weighting\",\"authors\":\"Jianming Zhang, Zhuofan Zheng, Xianding Xie, Yan Gui, Gwang-Jun Kim\",\"doi\":\"10.3233/ais-220038\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Traffic sign detection is a challenging task. Although existing deep learning techniques have made great progress in detecting traffic signs, there are still many unsolved challenges. We propose a novel traffic sign detection network named ReYOLO that learns rich contextual information and senses scale variations to efficiently detect small and ambiguous traffic signs in the wild. Specifically, we first replace the conventional convolutional block with modules that are built by structural reparameterization methods and are embedded into bigger structures, thus decoupling the training structures and the inference structures using parameter transformation, and allowing the model to learn more effective features. We then design a novel weighting mechanism which can be embedded into a feature pyramid to exploit foreground features at different scales to narrow the semantic gap between multiple scales. To fully evaluate the proposed method, we conduct experiments on a traditional traffic sign dataset GTSDB as well as two new traffic sign datasets TT100K and CCTSDB2021, achieving 97.2%, 68.3% and 83.9% mAP (Mean Average Precision) for the three-class detection challenge in these three datasets.\",\"PeriodicalId\":49316,\"journal\":{\"name\":\"Journal of Ambient Intelligence and Smart Environments\",\"volume\":\"19 7 1\",\"pages\":\"317-334\"},\"PeriodicalIF\":1.8000,\"publicationDate\":\"2022-07-26\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"13\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Ambient Intelligence and Smart Environments\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://doi.org/10.3233/ais-220038\",\"RegionNum\":4,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Ambient Intelligence and Smart Environments","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.3233/ais-220038","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 13

摘要

交通标志检测是一项具有挑战性的任务。尽管现有的深度学习技术在检测交通标志方面取得了很大进展,但仍有许多未解决的挑战。我们提出了一种新的交通标志检测网络ReYOLO,它可以学习丰富的上下文信息并感知尺度变化,从而有效地检测出小型和模糊的交通标志。具体来说,我们首先用结构重参数化方法构建的模块取代传统的卷积块,并嵌入到更大的结构中,从而使用参数变换将训练结构和推理结构解耦,并允许模型学习更有效的特征。然后,我们设计了一种新的加权机制,该机制可以嵌入到特征金字塔中,以利用不同尺度的前景特征来缩小多尺度之间的语义差距。为了充分评估所提出的方法,我们在传统的交通标志数据集GTSDB以及两个新的交通标志数据集TT100K和CCTSDB2021上进行了实验,对这三个数据集的三类检测挑战实现了97.2%,68.3%和83.9%的mAP (Mean Average Precision)。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
查看原文
分享 分享
微信好友 朋友圈 QQ好友 复制链接
本刊更多论文
ReYOLO: A traffic sign detector based on network reparameterization and features adaptive weighting
Traffic sign detection is a challenging task. Although existing deep learning techniques have made great progress in detecting traffic signs, there are still many unsolved challenges. We propose a novel traffic sign detection network named ReYOLO that learns rich contextual information and senses scale variations to efficiently detect small and ambiguous traffic signs in the wild. Specifically, we first replace the conventional convolutional block with modules that are built by structural reparameterization methods and are embedded into bigger structures, thus decoupling the training structures and the inference structures using parameter transformation, and allowing the model to learn more effective features. We then design a novel weighting mechanism which can be embedded into a feature pyramid to exploit foreground features at different scales to narrow the semantic gap between multiple scales. To fully evaluate the proposed method, we conduct experiments on a traditional traffic sign dataset GTSDB as well as two new traffic sign datasets TT100K and CCTSDB2021, achieving 97.2%, 68.3% and 83.9% mAP (Mean Average Precision) for the three-class detection challenge in these three datasets.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
Journal of Ambient Intelligence and Smart Environments
Journal of Ambient Intelligence and Smart Environments COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE-COMPUTER SCIENCE, INFORMATION SYSTEMS
CiteScore
4.30
自引率
17.60%
发文量
23
审稿时长
>12 weeks
期刊介绍: The Journal of Ambient Intelligence and Smart Environments (JAISE) serves as a forum to discuss the latest developments on Ambient Intelligence (AmI) and Smart Environments (SmE). Given the multi-disciplinary nature of the areas involved, the journal aims to promote participation from several different communities covering topics ranging from enabling technologies such as multi-modal sensing and vision processing, to algorithmic aspects in interpretive and reasoning domains, to application-oriented efforts in human-centered services, as well as contributions from the fields of robotics, networking, HCI, mobile, collaborative and pervasive computing. This diversity stems from the fact that smart environments can be defined with a variety of different characteristics based on the applications they serve, their interaction models with humans, the practical system design aspects, as well as the multi-faceted conceptual and algorithmic considerations that would enable them to operate seamlessly and unobtrusively. The Journal of Ambient Intelligence and Smart Environments will focus on both the technical and application aspects of these.
期刊最新文献
Evaluation factors of adopting smart home IoT: The hybrid fuzzy MCDM approach for robot vacuum Hybrid fuzzy response threshold-based distributed task allocation in heterogeneous multi-robot environment From programming-to-modeling-to-prompts smart ubiquitous applications A UAV deployment strategy based on a probabilistic data coverage model for mobile CrowdSensing applications Memoization based priority-aware task management for QoS provisioning in IoT gateways
×
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
现在去查看 取消
×
提示
确定
0
微信
客服QQ
Book学术公众号 扫码关注我们
反馈
×
意见反馈
请填写您的意见或建议
请填写您的手机或邮箱
已复制链接
已复制链接
快去分享给好友吧!
我知道了
×
扫码分享
扫码分享
Book学术官方微信
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术
文献互助 智能选刊 最新文献 互助须知 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 Book学术 All rights reserved.
ghs 京公网安备 11010802042870号 京ICP备2023020795号-1