Iterative Manifold Embedding Layer Learned by Incomplete Data for Large-Scale Image Retrieval

Iterative Manifold Embedding Layer Learned by Incomplete Data for Large-Scale Image Retrieval
复制标题

用于大规模图像检索的不完整数据学习的迭代流形嵌入层

DOI:
10.1109/tmm.2018.2883860
复制
发表时间:
2017-07
影响因子:
7.3
通讯作者:
Xiao Baihua
Xiao Baihua
中科院分区:
计算机科学1区
文献类型:
--
作者:
Xu Jian;Wang Chunheng;Qi Chengzuo;Shi Cunzhao;Xiao Baihua

文献摘要

参考文献

被引文献

相似文献

现有的流形学习方法不适用于图像检索任务,因为它们大多无法处理查询图像,而且计算成本更高,特别是对于大型数据库。因此,我们提出迭代流形嵌入(IME)层,该层的权值通过无监督策略离线学习,以探索不完整数据的内在流形。在包含27000张图像的大型数据库上,IME层在查询时嵌入原始表示的速度比其他流形学习方法快120倍以上。我们将位于高维空间流形上的数据库图像的原始描述符迭代嵌入到基于流形的表示中,以生成离线学习阶段的IME表示。根据原始描述符和数据库图像的IME表示,通过脊回归估计IME层的权重。在在线检索阶段,我们使用IME层以可忽略的时间成本(每张图像2 ms)映射查询图像的原始表示。我们在五个公共标准数据集上进行了图像检索实验。所提出的IME层显著优于相关的降维方法和流形学习方法。在没有后处理的情况下,我们的IME层实现了对大多数数据集进行后处理的最先进图像检索方法的性能提升,并且需要更少的计算成本。代码可在https://github.com/XJhaoren/IME_layer上获得。
Existing manifold learning methods are not appropriate for image retrieval tasks, because most of them are unable to process query images and they have much greater computational cost especially for large-scale database. Therefore, we propose the iterative manifold embedding (IME) layer, of which the weights are learned offline by an unsupervised strategy, to explore the intrinsic manifolds by incomplete data. On the large-scale database that contains 27 000 images, the IME layer is more than 120 times faster than other manifold learning methods to embed the original representations at query time. We embed the original descriptors of database images that lie on manifold in a high-dimensional space into manifold-based representations iteratively to generate the IME representations in an offline learning stage. According to the original descriptors and the IME representations of database images, we estimate the weights of the IME layer by ridge regression. In the online retrieval stage, we employ the IME layer to map the original representation of a query image with an ignorable time cost (2 ms per image). We experiment on five public standard datasets for image retrieval. The proposed IME layer significantly outperforms the related dimension reduction methods and manifold learning methods. Without postprocessing, our IME layer achieves a boost in the performance of state-of-the-art image retrieval methods with postprocessing on most datasets, and needs less computational cost. The code is available at https://github.com/XJhaoren/IME_layer.
DOI: 10.1201/b10345-5
发表时间: 2010-11
期刊: Encyclopedia of Autism Spectrum Disorders
影响因子: --
作者:
Kim-Anh Lê Cao;Z. Welham
通讯作者: Kim-Anh Lê Cao;Z. Welham
DOI: 10.1023/b:visi.0000029664.99615.94
发表时间: 2004-11-01
影响因子: 19.5
作者:
Lowe, DG
通讯作者: Lowe, DG
用于图像检索的上下文查询扩展
DOI: 10.1109/tmm.2014.2305909
发表时间: 2014-06
期刊: IEEE Trans. on Multimedia
影响因子: --
作者:
Zhang Yongdong;Tan Jianlong;Guo Li;Li Jintao
通讯作者: Li Jintao
DOI: 10.1109/cvpr.2007.383266
发表时间: 2007-06
期刊: 2007 IEEE Conference on Computer Vision and Pattern Recognition
影响因子: --
作者:
Florent Perronnin;C. Dance
通讯作者: Florent Perronnin;C. Dance
DOI: 10.1109/tpami.2011.235
发表时间: 2012-09-01
影响因子: 23.6
作者:
Jegou, Herve;Perronnin, Florent;Schmid, Cordelia
通讯作者: Schmid, Cordelia