语义分割的相对难度提炼

IF 7.3 2区 计算机科学 Q1 COMPUTER SCIENCE, INFORMATION SYSTEMS
Dong Liang, Yue Sun, Yun Du, Songcan Chen, Sheng-Jun Huang
{"title":"语义分割的相对难度提炼","authors":"Dong Liang, Yue Sun, Yun Du, Songcan Chen, Sheng-Jun Huang","doi":"10.1007/s11432-023-4061-2","DOIUrl":null,"url":null,"abstract":"<p>Current knowledge distillation (KD) methods primarily focus on transferring various structured knowledge and designing corresponding optimization goals to encourage the student network to imitate the output of the teacher network. However, introducing too many additional optimization objectives may lead to unstable training, such as gradient conflicts. Moreover, these methods ignored the guidelines of relative learning difficulty between the teacher and student networks. Inspired by human cognitive science, in this paper, we redefine knowledge from a new perspective — the student and teacher networks’ relative difficulty of samples, and propose a pixel-level KD paradigm for semantic segmentation named relative difficulty distillation (RDD). We propose a two-stage RDD framework: teacher-full evaluated RDD (TFE-RDD) and teacher-student evaluated RDD (TSE-RDD). RDD allows the teacher network to provide effective guidance on learning focus without additional optimization goals, thus avoiding adjusting learning weights for multiple losses. Extensive experimental evaluations using a general distillation loss function on popular datasets such as Cityscapes, CamVid, Pascal VOC, and ADE20k demonstrate the effectiveness of RDD against state-of-the-art KD methods. Additionally, our research showcases that RDD can integrate with existing KD methods to improve their upper performance bound. Codes are available at https://github.com/sunyueue/RDD.git.</p>","PeriodicalId":21618,"journal":{"name":"Science China Information Sciences","volume":"51 1","pages":""},"PeriodicalIF":7.3000,"publicationDate":"2024-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Relative difficulty distillation for semantic segmentation\",\"authors\":\"Dong Liang, Yue Sun, Yun Du, Songcan Chen, Sheng-Jun Huang\",\"doi\":\"10.1007/s11432-023-4061-2\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>Current knowledge distillation (KD) methods primarily focus on transferring various structured knowledge and designing corresponding optimization goals to encourage the student network to imitate the output of the teacher network. However, introducing too many additional optimization objectives may lead to unstable training, such as gradient conflicts. Moreover, these methods ignored the guidelines of relative learning difficulty between the teacher and student networks. Inspired by human cognitive science, in this paper, we redefine knowledge from a new perspective — the student and teacher networks’ relative difficulty of samples, and propose a pixel-level KD paradigm for semantic segmentation named relative difficulty distillation (RDD). We propose a two-stage RDD framework: teacher-full evaluated RDD (TFE-RDD) and teacher-student evaluated RDD (TSE-RDD). RDD allows the teacher network to provide effective guidance on learning focus without additional optimization goals, thus avoiding adjusting learning weights for multiple losses. Extensive experimental evaluations using a general distillation loss function on popular datasets such as Cityscapes, CamVid, Pascal VOC, and ADE20k demonstrate the effectiveness of RDD against state-of-the-art KD methods. Additionally, our research showcases that RDD can integrate with existing KD methods to improve their upper performance bound. Codes are available at https://github.com/sunyueue/RDD.git.</p>\",\"PeriodicalId\":21618,\"journal\":{\"name\":\"Science China Information Sciences\",\"volume\":\"51 1\",\"pages\":\"\"},\"PeriodicalIF\":7.3000,\"publicationDate\":\"2024-08-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Science China Information Sciences\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://doi.org/10.1007/s11432-023-4061-2\",\"RegionNum\":2,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"COMPUTER SCIENCE, INFORMATION SYSTEMS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Science China Information Sciences","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1007/s11432-023-4061-2","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}
引用次数: 0

摘要

目前的知识提炼(KD)方法主要侧重于传输各种结构化知识并设计相应的优化目标,以鼓励学生网络模仿教师网络的输出。然而,引入过多的额外优化目标可能会导致训练不稳定,如梯度冲突。此外,这些方法忽略了教师网络和学生网络之间相对学习难度的准则。受人类认知科学的启发,本文从一个新的角度--学生和教师网络样本的相对难度--重新定义知识,并提出了一种像素级的语义分割 KD 范式,命名为相对难度提炼(RDD)。我们提出了一个两阶段 RDD 框架:教师全评估 RDD(TFE-RDD)和师生评估 RDD(TSE-RDD)。RDD 允许教师网络为学习重点提供有效指导,而无需额外的优化目标,从而避免了为多重损失调整学习权重。我们使用通用蒸馏损失函数对 Cityscapes、CamVid、Pascal VOC 和 ADE20k 等流行数据集进行了广泛的实验评估,证明了 RDD 与最先进的 KD 方法相比的有效性。此外,我们的研究还表明,RDD 可以与现有的 KD 方法相结合,以提高其性能上限。代码见 https://github.com/sunyueue/RDD.git。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Relative difficulty distillation for semantic segmentation

Current knowledge distillation (KD) methods primarily focus on transferring various structured knowledge and designing corresponding optimization goals to encourage the student network to imitate the output of the teacher network. However, introducing too many additional optimization objectives may lead to unstable training, such as gradient conflicts. Moreover, these methods ignored the guidelines of relative learning difficulty between the teacher and student networks. Inspired by human cognitive science, in this paper, we redefine knowledge from a new perspective — the student and teacher networks’ relative difficulty of samples, and propose a pixel-level KD paradigm for semantic segmentation named relative difficulty distillation (RDD). We propose a two-stage RDD framework: teacher-full evaluated RDD (TFE-RDD) and teacher-student evaluated RDD (TSE-RDD). RDD allows the teacher network to provide effective guidance on learning focus without additional optimization goals, thus avoiding adjusting learning weights for multiple losses. Extensive experimental evaluations using a general distillation loss function on popular datasets such as Cityscapes, CamVid, Pascal VOC, and ADE20k demonstrate the effectiveness of RDD against state-of-the-art KD methods. Additionally, our research showcases that RDD can integrate with existing KD methods to improve their upper performance bound. Codes are available at https://github.com/sunyueue/RDD.git.

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
Science China Information Sciences
Science China Information Sciences COMPUTER SCIENCE, INFORMATION SYSTEMS-
CiteScore
12.60
自引率
5.70%
发文量
224
审稿时长
8.3 months
期刊介绍: Science China Information Sciences is a dedicated journal that showcases high-quality, original research across various domains of information sciences. It encompasses Computer Science & Technologies, Control Science & Engineering, Information & Communication Engineering, Microelectronics & Solid-State Electronics, and Quantum Information, providing a platform for the dissemination of significant contributions in these fields.
×
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:481959085
Book学术官方微信