利用跨分辨率关系对比蒸馏技术识别低分辨率物体

Kangkai Zhang, Shiming Ge, Ruixin Shi, Dan Zeng
{"title":"利用跨分辨率关系对比蒸馏技术识别低分辨率物体","authors":"Kangkai Zhang, Shiming Ge, Ruixin Shi, Dan Zeng","doi":"arxiv-2409.02555","DOIUrl":null,"url":null,"abstract":"Recognizing objects in low-resolution images is a challenging task due to the\nlack of informative details. Recent studies have shown that knowledge\ndistillation approaches can effectively transfer knowledge from a\nhigh-resolution teacher model to a low-resolution student model by aligning\ncross-resolution representations. However, these approaches still face\nlimitations in adapting to the situation where the recognized objects exhibit\nsignificant representation discrepancies between training and testing images.\nIn this study, we propose a cross-resolution relational contrastive\ndistillation approach to facilitate low-resolution object recognition. Our\napproach enables the student model to mimic the behavior of a well-trained\nteacher model which delivers high accuracy in identifying high-resolution\nobjects. To extract sufficient knowledge, the student learning is supervised\nwith contrastive relational distillation loss, which preserves the similarities\nin various relational structures in contrastive representation space. In this\nmanner, the capability of recovering missing details of familiar low-resolution\nobjects can be effectively enhanced, leading to a better knowledge transfer.\nExtensive experiments on low-resolution object classification and\nlow-resolution face recognition clearly demonstrate the effectiveness and\nadaptability of our approach.","PeriodicalId":501480,"journal":{"name":"arXiv - CS - Multimedia","volume":"67 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Low-Resolution Object Recognition with Cross-Resolution Relational Contrastive Distillation\",\"authors\":\"Kangkai Zhang, Shiming Ge, Ruixin Shi, Dan Zeng\",\"doi\":\"arxiv-2409.02555\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Recognizing objects in low-resolution images is a challenging task due to the\\nlack of informative details. Recent studies have shown that knowledge\\ndistillation approaches can effectively transfer knowledge from a\\nhigh-resolution teacher model to a low-resolution student model by aligning\\ncross-resolution representations. However, these approaches still face\\nlimitations in adapting to the situation where the recognized objects exhibit\\nsignificant representation discrepancies between training and testing images.\\nIn this study, we propose a cross-resolution relational contrastive\\ndistillation approach to facilitate low-resolution object recognition. Our\\napproach enables the student model to mimic the behavior of a well-trained\\nteacher model which delivers high accuracy in identifying high-resolution\\nobjects. To extract sufficient knowledge, the student learning is supervised\\nwith contrastive relational distillation loss, which preserves the similarities\\nin various relational structures in contrastive representation space. In this\\nmanner, the capability of recovering missing details of familiar low-resolution\\nobjects can be effectively enhanced, leading to a better knowledge transfer.\\nExtensive experiments on low-resolution object classification and\\nlow-resolution face recognition clearly demonstrate the effectiveness and\\nadaptability of our approach.\",\"PeriodicalId\":501480,\"journal\":{\"name\":\"arXiv - CS - Multimedia\",\"volume\":\"67 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-04\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Multimedia\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.02555\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Multimedia","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.02555","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

由于缺乏信息细节,在低分辨率图像中识别物体是一项极具挑战性的任务。最近的研究表明,知识灌输方法可以通过对齐跨分辨率表征,有效地将知识从高分辨率的教师模型转移到低分辨率的学生模型。在本研究中,我们提出了一种跨分辨率关系对比蒸馏方法来促进低分辨率物体识别。我们的方法能让学生模型模拟训练有素的教师模型的行为,从而在识别高分辨率物体时达到较高的准确率。为了提取足够的知识,学生的学习受到了对比关系蒸馏损失的监督,这种损失保留了对比表示空间中各种关系结构的相似性。在低分辨率物体分类和低分辨率人脸识别方面的大量实验清楚地证明了我们方法的有效性和适应性。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Low-Resolution Object Recognition with Cross-Resolution Relational Contrastive Distillation
Recognizing objects in low-resolution images is a challenging task due to the lack of informative details. Recent studies have shown that knowledge distillation approaches can effectively transfer knowledge from a high-resolution teacher model to a low-resolution student model by aligning cross-resolution representations. However, these approaches still face limitations in adapting to the situation where the recognized objects exhibit significant representation discrepancies between training and testing images. In this study, we propose a cross-resolution relational contrastive distillation approach to facilitate low-resolution object recognition. Our approach enables the student model to mimic the behavior of a well-trained teacher model which delivers high accuracy in identifying high-resolution objects. To extract sufficient knowledge, the student learning is supervised with contrastive relational distillation loss, which preserves the similarities in various relational structures in contrastive representation space. In this manner, the capability of recovering missing details of familiar low-resolution objects can be effectively enhanced, leading to a better knowledge transfer. Extensive experiments on low-resolution object classification and low-resolution face recognition clearly demonstrate the effectiveness and adaptability of our approach.
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
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:481959085
Book学术官方微信