期刊文献+

基于VGG的高准确率翻拍广告图片检索系统 被引量:1

High Accuracy Image Retrieval System of Recapture Advertisement Based on VGG
下载PDF
导出
摘要 随着信息技术的发展,图像检索技术应用越来越广泛,对图像检索技术的适用性要求也越来越高,不但要对主体明确的图像进行检索,同时对以翻拍广告图像为例的受环境等多种外界因素影响的图像也可以进行准确检索。实际中,如何摒弃这些外界干扰并迅速的从海量的数据中准确地匹配到所需的图像,成为图像检索的一大热点问题。本文采用VGG网络模型提取图像特征,利用迁移学习的思想将在ImageNet数据集上训练好的网络模型根据目标数据集的特点进行微调,使得在已有的大数据集上训练好的网络模型进一步适用于新的数据集。采用余弦相似度度量进行相似度比较。结果表明,卷积神经网络可以提取更丰富的特征,相对传统分类方法可以提升速度和准确率。 With the development of information technology, the application of image retrieval technology has become more and more extensive, and the requirements for the applicability of image retrieval technology are getting more higher. It can not only retrieve images with a clear subject, but also accurately retrieve images that are affected by a variety of external factors such as the recapture advertising images. In practice, how to get rid of these external interferences and match the required images from the massive data quickly and accurately has become a hot issue in image retrieval. This paper uses the VGG network model to extract image features, and uses the idea of transfer learning to fine-tune the network model trained on the ImageNet data set, so that the network model trained on the existing large data set is applied to the new Data set. Use cosine similarity measure to compare similarity. The results show that the convolutional neural network can extract richer features, which can improve the speed and accuracy compared with traditional classification methods.
作者 赵朋悦 王翾 刘守训 ZHAO Peng-yue;WANG Xuan;LIU Shou-xun(School of information and Communication Engineering,Communication University of China,Beijing 10024,China)
机构地区 中国传媒大学
出处 《无线通信技术》 2022年第2期32-36,40,共6页 Wireless Communication Technology
关键词 图像检索 深度学习 迁移学习 余弦相似度度量 VGG image retrieval deep learning transfer learning cosine similarity measure VGG
  • 相关文献

参考文献2

二级参考文献14

  • 1Edkins J, Graham M. Content-based image retrieval[R]. University of Northumbria at Newcastle: TechnicalReport, 1999.
  • 2Oliva A, Antonio T. Modeling the shape of the scene: aholistic representation of the spatial envelope [ J ]. Inter-national Journal of Computer Vision, 2001,42 ( 3 ):145-175.
  • 3Lowe D G. Distinctive image features from scale-invariantkeypoints [ J]. International Journal of Computer Vision,2004, 60(2) : 91-110.
  • 4Dalai N , Bill T. Histograms of oriented gradients for hu-man detection [ C] // CVPR. San Diego,USA: IEEE,2005: 886-893.
  • 5Datta R,Joshi D. Image retrieval: ideas, influences,and trends of the new age [J]. ACM Computing Surveys(CSUR),2008,40(2) : 5.
  • 6Hinton G E, Ruslan R S. Reducing the dimensionality ofdata with neural networks [J]. Science, 2006, 313(5786) : 504-507.
  • 7Cun L, Boser B. Handwritten digit recognition with aback-propagation network [C]// Advances in Neural In-formation Processing Systems. Denver, Colorado, USA:MIT, 1990: 396-404.
  • 8Alex K, Ilya S. Imagenet classification with deep convo-lutional neural networks [ C] H Advances in Neural Infor-mation Processing Systems. Lake Tahoe, Nevada, USA:MIT, 2012: 1097-1105.
  • 9Deng Jia,Dong Wei. Imagenet: a large-scale hierarchi-cal image database [ C ] CVPR. Miami, USA: IEEE,2009: 248-255.
  • 10Ali S R, Hossein A, Josephine S, et al. Cnn featuresoff-the-shelf : an astounding baseline for recognition [ C ]//CVPRW. Columbus, USA: IEEE, 2014: 512-519.

共引文献35

同被引文献8

引证文献1

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部