OPT OpenIR  > 光谱成像技术研究室
Zero-Shot Learning Using Synthesised Unseen Visual Data with Diffusion Regularisation
Long, Yang1; Liu, Li2,3; Shen, Fumin4; Shao, Ling2,3; Li, Xuelong5
作者部门光学影像学习与分析中心
2018-10
发表期刊IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
ISSN0162-8828;1939-3539
卷号40期号:10页码:2498-2512
产权排序5
摘要

Sufficient training examples are the fundamental requirement for most of the learning tasks. However, collecting well-labelled training examples is costly. Inspired by Zero-shot Learning (ZSL) that can make use of visual attributes or natural language semantics as an intermediate level clue to associate low-level features with high-level classes, in a novel extension of this idea, we aim to synthesise training data for novel classes using only semantic attributes. Despite the simplicity of this idea, there are several challenges. First, how to prevent the synthesised data from over-fitting to training classes? Second, how to guarantee the synthesised data is discriminative for ZSL tasks? Third, we observe that only a few dimensions of the learnt features gain high variances whereas most of the remaining dimensions are not informative. Thus, the question is how to make the concentrated information diffuse to most of the dimensions of synthesised data. To address the above issues, we propose a novel embedding algorithm named Unseen Visual Data Synthesis (UVDS) that projects semantic features to the high-dimensional visual feature space. Two main techniques are introduced in our proposed algorithm. (1) We introduce a latent embedding space which aims to reconcile the structural difference between the visual and semantic spaces, meanwhile preserve the local structure. (2) We propose a novel Diffusion Regularisation (DR) that explicitly forces the variances to diffuse over most dimensions of the synthesised data. By an orthogonal rotation (more precisely, an orthogonal transformation), DR can remove the redundant correlated attributes and further alleviate the over-fitting problem. On four benchmark datasets, we demonstrate the benefit of using synthesised unseen data for zero-shot learning. Extensive experimental results suggest that our proposed approach significantly outperforms the state-of-the-art methods.

关键词Zero-shot Learning Data Synthesis Diffusion Regularisation Visual-semantic Embedding Object Recognition
DOI10.1109/TPAMI.2017.2762295
收录类别SCI ; EI
语种英语
WOS记录号WOS:000443875500016
出版者IEEE COMPUTER SOC
EI入藏号20174304296448
引用统计
被引频次:37[WOS]   [WOS记录]     [WOS相关记录]
文献类型期刊论文
条目标识符http://ir.opt.ac.cn/handle/181661/30620
专题光谱成像技术研究室
通讯作者Shao, Ling
作者单位1.Univ Newcastle, Sch Comp Sci, OpenLab, Newcastle Upon Tyne NE4 5TG, Tyne & Wear, England
2.Incept Inst Artificial Intelligence, Abu Dhabi, U Arab Emirates
3.Univ East Anglia, Sch Comp Sci, Norwich NR4 7TJ, Norfolk, England
4.Univ Elect Sci & Technol China, Sch Comp Sci & Engn, Chengdu 611731, Sichuan, Peoples R China
5.Chinese Acad Sci, Xian Inst Opt & Precis Mech, Xian 710119, Shaanxi, Peoples R China
推荐引用方式
GB/T 7714
Long, Yang,Liu, Li,Shen, Fumin,et al. Zero-Shot Learning Using Synthesised Unseen Visual Data with Diffusion Regularisation[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2018,40(10):2498-2512.
APA Long, Yang,Liu, Li,Shen, Fumin,Shao, Ling,&Li, Xuelong.(2018).Zero-Shot Learning Using Synthesised Unseen Visual Data with Diffusion Regularisation.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,40(10),2498-2512.
MLA Long, Yang,et al."Zero-Shot Learning Using Synthesised Unseen Visual Data with Diffusion Regularisation".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 40.10(2018):2498-2512.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
Zero-Shot Learning U(1288KB)期刊论文出版稿限制开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[Long, Yang]的文章
[Liu, Li]的文章
[Shen, Fumin]的文章
百度学术
百度学术中相似的文章
[Long, Yang]的文章
[Liu, Li]的文章
[Shen, Fumin]的文章
必应学术
必应学术中相似的文章
[Long, Yang]的文章
[Liu, Li]的文章
[Shen, Fumin]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。