用于高光谱图像场景分类的轻量级深度全局-局部知识蒸馏网络

Q3 Materials Science
Yingxu LIU, Chunyu PU, Diankun XU, Yichuan YANG, Hong HUANG
{"title":"用于高光谱图像场景分类的轻量级深度全局-局部知识蒸馏网络","authors":"Yingxu LIU, Chunyu PU, Diankun XU, Yichuan YANG, Hong HUANG","doi":"10.37188/ope.20233117.2598","DOIUrl":null,"url":null,"abstract":"é’ˆå¯¹ç›®æ ‡åœºæ™¯å¤æ‚çš„ç©ºé—´å¸ƒå±€å’Œé«˜å ‰è°±å½±åƒå›ºæœ‰çš„ç©º-è°±ä¿¡æ¯å†—ä½™ç­‰æŒ‘æˆ˜ï¼Œæå‡ºäº†ç«¯åˆ°ç«¯çš„è½»é‡åŒ–æ·±åº¦å ¨å±€-局部知识蒸馏(Lightweight Deep Global-Local Knowledge Distillation,LDGLKD)网络。为探索空-è°±ç‰¹å¾çš„å ¨å±€åºåˆ—å±žæ€§ï¼Œæ•™å¸ˆæ¨¡åž‹è§†è§‰Transformer(Vision Transformer,ViTï¼‰è¢«ç”¨æ¥æŒ‡å¯¼è½»é‡åŒ–å­¦ç”Ÿæ¨¡åž‹è¿›è¡Œé«˜å ‰è°±å½±åƒåœºæ™¯åˆ†ç±»ã€‚LDGLKD选择预训练的VGG16作为学生模型来提取局部细节信息,将ViT和VGG16é€šè¿‡çŸ¥è¯†è’¸é¦ååŒè®­ç»ƒåŽï¼Œæ•™å¸ˆæ¨¡åž‹å°†æ‰€å­¦ä¹ åˆ°çš„è¿œç¨‹ä¸Šä¸‹æ–‡å ³ç³»å‘å°è§„æ¨¡å­¦ç”Ÿæ¨¡åž‹è¿›è¡Œä¼ é€’ã€‚LDGLKDå¯é€šè¿‡çŸ¥è¯†è’¸é¦ç»“åˆä¸Šè¿°ä¸¤ç§æ¨¡åž‹çš„ä¼˜ç‚¹ï¼Œåœ¨æ¬§æ¯”ç‰¹é«˜å ‰è°±å½±åƒåœºæ™¯åˆ†ç±»æ•°æ®é›†OHID-SCåŠå ¬å¼€çš„é«˜å ‰è°±é¥æ„Ÿå›¾åƒæ•°æ®é›†HSRS-SC上的最佳分类精度分别达到91.62%和97.96%。实验结果表明:LDGLKDç½‘ç»œå ·æœ‰è‰¯å¥½çš„åˆ†ç±»æ€§èƒ½ã€‚æ ¹æ®æ¬§æ¯”ç‰¹ç æµ·ä¸€å·å«æ˜Ÿæä¾›çš„é¥æ„Ÿæ•°æ®æž„å»ºçš„OHID-SCå¯ä»¥åæ˜ è¯¦ç»†çš„åœ°è¡¨è¦†ç›–æƒ å†µï¼Œå¹¶ä¸ºé«˜å ‰è°±åœºæ™¯åˆ†ç±»ä»»åŠ¡æä¾›æ•°æ®æ”¯æ’‘ã€‚","PeriodicalId":39778,"journal":{"name":"Guangxue Jingmi Gongcheng/Optics and Precision Engineering","volume":"18 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Lightweight deep global-local knowledge distillation network for hyperspectral image scene classification\",\"authors\":\"Yingxu LIU, Chunyu PU, Diankun XU, Yichuan YANG, Hong HUANG\",\"doi\":\"10.37188/ope.20233117.2598\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"é’ˆå¯¹ç›®æ ‡åœºæ™¯å¤æ‚çš„ç©ºé—´å¸ƒå±€å’Œé«˜å ‰è°±å½±åƒå›ºæœ‰çš„ç©º-è°±ä¿¡æ¯å†—ä½™ç­‰æŒ‘æˆ˜ï¼Œæå‡ºäº†ç«¯åˆ°ç«¯çš„è½»é‡åŒ–æ·±åº¦å ¨å±€-局部知识蒸馏(Lightweight Deep Global-Local Knowledge Distillation,LDGLKD)网络。为探索空-è°±ç‰¹å¾çš„å ¨å±€åºåˆ—å±žæ€§ï¼Œæ•™å¸ˆæ¨¡åž‹è§†è§‰Transformer(Vision Transformer,ViTï¼‰è¢«ç”¨æ¥æŒ‡å¯¼è½»é‡åŒ–å­¦ç”Ÿæ¨¡åž‹è¿›è¡Œé«˜å ‰è°±å½±åƒåœºæ™¯åˆ†ç±»ã€‚LDGLKD选择预训练的VGG16作为学生模型来提取局部细节信息,将ViT和VGG16é€šè¿‡çŸ¥è¯†è’¸é¦ååŒè®­ç»ƒåŽï¼Œæ•™å¸ˆæ¨¡åž‹å°†æ‰€å­¦ä¹ åˆ°çš„è¿œç¨‹ä¸Šä¸‹æ–‡å ³ç³»å‘å°è§„æ¨¡å­¦ç”Ÿæ¨¡åž‹è¿›è¡Œä¼ é€’ã€‚LDGLKDå¯é€šè¿‡çŸ¥è¯†è’¸é¦ç»“åˆä¸Šè¿°ä¸¤ç§æ¨¡åž‹çš„ä¼˜ç‚¹ï¼Œåœ¨æ¬§æ¯”ç‰¹é«˜å ‰è°±å½±åƒåœºæ™¯åˆ†ç±»æ•°æ®é›†OHID-SCåŠå ¬å¼€çš„é«˜å ‰è°±é¥æ„Ÿå›¾åƒæ•°æ®é›†HSRS-SC上的最佳分类精度分别达到91.62%和97.96%。实验结果表明:LDGLKDç½‘ç»œå ·æœ‰è‰¯å¥½çš„åˆ†ç±»æ€§èƒ½ã€‚æ ¹æ®æ¬§æ¯”ç‰¹ç æµ·ä¸€å·å«æ˜Ÿæä¾›çš„é¥æ„Ÿæ•°æ®æž„å»ºçš„OHID-SCå¯ä»¥åæ˜ è¯¦ç»†çš„åœ°è¡¨è¦†ç›–æƒ å†µï¼Œå¹¶ä¸ºé«˜å ‰è°±åœºæ™¯åˆ†ç±»ä»»åŠ¡æä¾›æ•°æ®æ”¯æ’‘ã€‚\",\"PeriodicalId\":39778,\"journal\":{\"name\":\"Guangxue Jingmi Gongcheng/Optics and Precision Engineering\",\"volume\":\"18 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-01-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Guangxue Jingmi Gongcheng/Optics and Precision Engineering\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.37188/ope.20233117.2598\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"Materials Science\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Guangxue Jingmi Gongcheng/Optics and Precision Engineering","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.37188/ope.20233117.2598","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"Materials Science","Score":null,"Total":0}
引用次数: 0

摘要

Deep Global-Local Knowledge Distillation:LDGLKDï¼ç½ç "为æ¢ç´¢ç©º-Transformer(变压器):Vision Transformer(视觉变压器):ViT(智能变压器LDGLKD "VGG16 "系列在LDGLKD中,你会发现很多新功能,比如 "我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么","我可以做什么"。SCHSRS-SC... 91.62%å97.96%的学生都在使用OHID-SCCACS。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Lightweight deep global-local knowledge distillation network for hyperspectral image scene classification
é’ˆå¯¹ç›®æ ‡åœºæ™¯å¤æ‚çš„ç©ºé—´å¸ƒå±€å’Œé«˜å ‰è°±å½±åƒå›ºæœ‰çš„ç©º-è°±ä¿¡æ¯å†—ä½™ç­‰æŒ‘æˆ˜ï¼Œæå‡ºäº†ç«¯åˆ°ç«¯çš„è½»é‡åŒ–æ·±åº¦å ¨å±€-局部知识蒸馏(Lightweight Deep Global-Local Knowledge Distillation,LDGLKD)网络。为探索空-è°±ç‰¹å¾çš„å ¨å±€åºåˆ—å±žæ€§ï¼Œæ•™å¸ˆæ¨¡åž‹è§†è§‰Transformer(Vision Transformer,ViTï¼‰è¢«ç”¨æ¥æŒ‡å¯¼è½»é‡åŒ–å­¦ç”Ÿæ¨¡åž‹è¿›è¡Œé«˜å ‰è°±å½±åƒåœºæ™¯åˆ†ç±»ã€‚LDGLKD选择预训练的VGG16作为学生模型来提取局部细节信息,将ViT和VGG16é€šè¿‡çŸ¥è¯†è’¸é¦ååŒè®­ç»ƒåŽï¼Œæ•™å¸ˆæ¨¡åž‹å°†æ‰€å­¦ä¹ åˆ°çš„è¿œç¨‹ä¸Šä¸‹æ–‡å ³ç³»å‘å°è§„æ¨¡å­¦ç”Ÿæ¨¡åž‹è¿›è¡Œä¼ é€’ã€‚LDGLKDå¯é€šè¿‡çŸ¥è¯†è’¸é¦ç»“åˆä¸Šè¿°ä¸¤ç§æ¨¡åž‹çš„ä¼˜ç‚¹ï¼Œåœ¨æ¬§æ¯”ç‰¹é«˜å ‰è°±å½±åƒåœºæ™¯åˆ†ç±»æ•°æ®é›†OHID-SCåŠå ¬å¼€çš„é«˜å ‰è°±é¥æ„Ÿå›¾åƒæ•°æ®é›†HSRS-SC上的最佳分类精度分别达到91.62%和97.96%。实验结果表明:LDGLKDç½‘ç»œå ·æœ‰è‰¯å¥½çš„åˆ†ç±»æ€§èƒ½ã€‚æ ¹æ®æ¬§æ¯”ç‰¹ç æµ·ä¸€å·å«æ˜Ÿæä¾›çš„é¥æ„Ÿæ•°æ®æž„å»ºçš„OHID-SCå¯ä»¥åæ˜ è¯¦ç»†çš„åœ°è¡¨è¦†ç›–æƒ å†µï¼Œå¹¶ä¸ºé«˜å ‰è°±åœºæ™¯åˆ†ç±»ä»»åŠ¡æä¾›æ•°æ®æ”¯æ’‘ã€‚
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
Guangxue Jingmi Gongcheng/Optics and Precision Engineering
Guangxue Jingmi Gongcheng/Optics and Precision Engineering Materials Science-Electronic, Optical and Magnetic Materials
CiteScore
2.40
自引率
0.00%
发文量
95
×
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