SIA OpenIR  > 数字工厂研究室
基于全局交互的图像语义理解方法研究
Alternative TitleResearch on Image Semantic Understanding Method Based on Global Interaction
库涛1,2; 熊艳彬1,2,3; 杨楠1,2,3; 林乐新1,2; 朱珠4
Department数字工厂研究室
Source Publication控制与决策
ISSN1001-0920
2019
Pages1-9
Contribution Rank1
Funding Organization国家重点研发计划(2017YFB0306401) ; 国家自然科学基金(61803367)
Keyword卷积神经网络 循环神经网络 图像语义理解 全局交互机制 数据正则化 门控循环单元GRU
Abstract本文针对图像语义生成过程中图像信息易模糊的问题,研究了基于全局交互的图像语义理解方法。提出了基于双向门控循环单元(Gated Recurrent Unit,GRU)和图像信息全局交互相结合的图像语义生成模型,并在模型中提出了将图像和文本数据进行正则化处理以及采用文本向量映射表示文本信息的方法用于指导语义生成。实验结果表明,所提出的模型提高了图像语义描述的内容丰富度、准确性和逻辑性;将数据正则化处理、采用文本向量映射方式可以较大程度的解决数据稀疏和偏态问题;采用GUR单元可以进一步降低模型参数规模加快算法收敛速度,结合正则化及Dropout率可以有效抑制模型过拟合。
Other AbstractAiming at the problem that image information is easy to be blurred in the process of image semantic generation, this paper studies an image semantic understanding method based on global interaction. An image semantic generation model based on Gated Recurrent Unit (GRU) and global intersection of image information is proposed in this paper. Experimental results show that the proposed model improves the content richness, accuracy and logicality of semantic description. Data sparse and skewness can be solved to a large extent by regularization and word vector mapping. GUR element can further reduce the size of model parameters and accelerate the speed of algorithm convergence. The combination of regularization and Dropout rate can effectively suppress model overfitting.
Language中文
Document Type期刊论文
Identifierhttp://ir.sia.cn/handle/173321/25452
Collection数字工厂研究室
Corresponding Author库涛
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院机器人与智能制造创新研究院
3.中国科学院大学
4.辽宁大学
Recommended Citation
GB/T 7714
库涛,熊艳彬,杨楠,等. 基于全局交互的图像语义理解方法研究[J]. 控制与决策,2019:1-9.
APA 库涛,熊艳彬,杨楠,林乐新,&朱珠.(2019).基于全局交互的图像语义理解方法研究.控制与决策,1-9.
MLA 库涛,et al."基于全局交互的图像语义理解方法研究".控制与决策 (2019):1-9.
Files in This Item: Download All
File Name/Size DocType Version Access License
基于全局交互的图像语义理解方法研究.pd(1330KB)期刊论文出版稿开放获取CC BY-NC-SAView Download
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[库涛]'s Articles
[熊艳彬]'s Articles
[杨楠]'s Articles
Baidu academic
Similar articles in Baidu academic
[库涛]'s Articles
[熊艳彬]'s Articles
[杨楠]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[库涛]'s Articles
[熊艳彬]'s Articles
[杨楠]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: 基于全局交互的图像语义理解方法研究.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.