Deep Graph-neighbor Coherence Preserving Network for Unsupervised Cross-modal Hashing
Open accessThe DOI matches its Crossref record; 43 of the 43 references it cites were found in a registry or an index.
Abstract
Unsupervised cross-modal hashing (UCMH) has become a hot topic recently. Current UCMH focuses on exploring data similarities. However, current UCMH methods calculate the similarity between two data, mainly relying on the two data's cross-modal features. These methods suffer from inaccurate similarity problems that result in a suboptimal retrieval Hamming space, because the cross-modal features between the data are not sufficient to describe the complex data relationships, such as situations where two data have different feature representations but share the inherent concepts. In this paper, we devise a deep graph-neighbor coherence preserving network (DGCPN). Specifically, DGCPN stems from graph models and explores graph-neighbor coherence by consolidating the information between data and their neighbors. DGCPN regulates comprehensive similarity preserving losses by exploiting three types of data similarities (i.e., the graph-neighbor coherence, the coexistent similarity, and the intra- and inter-modality consistency) and designs a half-real and half-binary optimization strategy to reduce the quantization errors during hashing. Essentially, DGCPN addresses the inaccurate similarity problem by exploring and exploiting the data's intrinsic relationships in a graph. We conduct extensive experiments on three public UCMH datasets. The experimental results demonstrate the superiority of DGCPN, e.g., by improving the mean average precision from 0.722 to 0.751 on MIRFlickr-25K using 64-bit hashing codes to retrieval texts from images. We will release the source code package and the trained model on https://github.com/Atmegal/DGCPN.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 2×
“2020), we use a pre-trained VGG-19 (Simonyan and Zisserman 2014) to extract image features and use the bagof-words (BoW) model to extract text features.”
MethodImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 1×
“We note that the feature-extracting subnetworks can marry other backbones, such as AlexNet (Krizhevsky, Sutskever, and Hinton 2012) and the Latent Dirichlet allocation (LDA) model (Blei, Ng, and Jordan 2003).”
BackgroundImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 2×
“Following Hu et al., for all datasets, the images are represented by 4096-dimensional deep features extracted from VGG-19 pretrained on ImageNet (Deng et al. 2009).”
Method(2003) · Journal of Machine Learning Research · doi:10.5555/944919.944937 · Cited by 32,337
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2002) · Advances in Neural Information Processing Systems 14 · doi:10.7551/mitpress/1120.003.0082 · Cited by 12,223
NUS-WIDE: a real-world web image database from National University of Singapore
Chua, T.-S., Tang, J., Hong, R., Li, H., Luo, Z., & Zheng, Y. (2009) · Proceedings of the ACM International Conference on Image and Video Retrieval · doi:10.1145/1646396.1646452 · Cited by 3,025 · cited 2×
“2010), MIRFlickr-25K (Huiskes and Lew 2008), and NUS-WIDE (Chua et al. 2009).”
MethodImageNet: A large-scale hierarchical image database
Deng, J. H., Dong, W. F., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvprw.2009.5206848 · Cited by 2,714
The MIR flickr retrieval evaluation
Huiskes, M. J., & Lew, M. S. (2008) · Proceedings of the 1st ACM international conference on Multimedia information retrieval · doi:10.1145/1460096.1460104 · Cited by 1,836 · cited 5×
“2010), MIRFlickr-25K (Huiskes and Lew 2008), and NUS-WIDE (Chua et al.”
MethodInfluentialA new approach to cross-modal multimedia retrieval
Rasiwasia, N., Costa Pereira, J., Coviello, E., Doyle, G., Lanckriet, G. R. G., Levy, R., & Vasconcelos, N. (2010) · Proceedings of the 18th ACM international conference on Multimedia · doi:10.1145/1873951.1873987 · Cited by 1,334 · cited 7×
“The final parameters are: for Wikipedia: α =0.3, γ =0.3, λ 1 =1, λ 2 =1, β =900, and k =600; for MIRFlickr-25K α =0.01, γ =0.3, λ 1 =1, λ 2 =1, β =4000, k =2000; and for NUS-WIDE: α =0.1, γ =0.3, λ 1 =1, λ 2 =1, β =4500, k =2000.”
MethodInfluentialJiang, Q.-Y., & Li, W.-J. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.348 · Cited by 880 · cited 1×
“CMH methods can be roughly classified into two categories: supervised methods (Jiang and Li 2017) and un-supervised methods (Su, Zhong, and Zhang 2019).”
MethodAdversarial Cross-Modal Retrieval
Wang, B., Yang, Y., Xu, X., Hanjalic, A., & Shen, H. T. (2017) · Proceedings of the 25th ACM international conference on Multimedia · doi:10.1145/3123266.3123326 · Cited by 779 · cited 2×
“for a large volume of multimedia data (Feng, Wang, and Li 2014; Wang et al. 2017; Xie et al. 2020).”
BackgroundZhang, J., & Tao, D. (2021) · IEEE Internet of Things Journal · doi:10.1109/jiot.2020.3039359 · Cited by 774 · cited 1×
“Later, UCMH methods exploited deep neural networks (DNNs) (Wang et al. 2020; Zhang and Tao 2020).”
MethodSelf-citationCollective Matrix Factorization Hashing for Multimodal Data
Ding, G., Guo, Y., & Zhou, J. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.267 · Cited by 680 · cited 6×
“…text retrieval from an image query (I2T) and image retrieval from (Liu et al. 2017) 0.242 0.283 0.305 0.590 0.597 0.597 0.400 0.414 0.424 CMFH (Ding, Guo, and Zhou 2014) 0.163 0.176 0.163 0.588 0.592 0.594 0.483 0.488 0.486 LSSH (Zhou, Ding, and Guo 2014) 0.364 0.387 0.397 0.630 0.634 0.631…”
BackgroundMethodInfluentialMirrorGAN: Learning Text-To-Image Generation by Redescription
Qiao, T., Zhang, J., Xu, D., & Tao, D. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00160 · Cited by 644 · cited 1×
“Due to the explosion of multimedia data on the Internet (Yu, Rui, and Chen 2013; Yang et al. 2020c), cross-modal retrieval has become an essential technique in information retrieval (Yang et al. 2020c) and multi-modal data mining (Qiao et al. 2019b,a; Yang et al. 2020a,c; Li et al. 2020).”
BackgroundSelf-citationInter-media hashing for large-scale retrieval from heterogeneous data sources
Song, J., Yang, Y., Yang, Y., Huang, Z., & Shen, H. T. (2013) · Proceedings of the 2013 ACM SIGMOD International Conference on Management of Data · doi:10.1145/2463676.2465274 · Cited by 620 · cited 4×
“For example, IMH (Song et al. 2013), CMFH (Ding, Guo, and Zhou 2014), and MSFH (Fang, Zhang, and Ren 2019) learned hashing codes by preserving intra- and inter-graph consistency, using collective matrix factorization, and using multigraph regularized smooth matrix factorization, respectively.”
BackgroundMethodInfluentialLearning Hash Functions for Cross-View Similarity Search
Kumar, S., & Udupa, R. (2012) · Cited by 614
Cross-modal Retrieval with Correspondence Autoencoder
Feng, F., Wang, X., & Li, R. (2014) · Proceedings of the 22nd ACM international conference on Multimedia · doi:10.1145/2647868.2654902 · Cited by 560 · cited 1×
“…such a connection can be inferred by their neighbor coherence that both “rose” and “wedding” have high similarities of the “bride, rose”. mon spaces, and they suffer from high computation burden for a large volume of multimedia data (Feng, Wang, and Li 2014; Wang et al. 2017; Xie et al. 2020).”
BackgroundLatent semantic sparse hashing for cross-modal similarity search
Zhou, J., Ding, G., & Guo, Y. (2014) · Proceedings of the 37th international ACM SIGIR conference on Research & development in information retrieval · doi:10.1145/2600428.2609610 · Cited by 467 · cited 4×
“We compare DGCPN with nine UCMH methods: CVH (Kumar and Udupa 2011), FSH (Liu et al. 2017), CMFH (Ding, Guo, and Zhou 2014), LSSH (Zhou, Ding, and Guo 2014), UGACH (Zhang, Peng, and Yuan 2018), DJSRH (Su, Zhong, and Zhang 2019), UKD-SS (Hu et al. 2020), DSAH (Yang et al. 2020b), and JDSH (Liu et al. 2020).”
MethodInfluentialDeep Joint-Semantics Reconstructing Hashing for Large-Scale Unsupervised Cross-Modal Retrieval
Su, S., Zhong, Z., & Zhang, C. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00312 · Cited by 342 · cited 11×
“Both joint-semantics similarity (Su, Zhong, and Zhang 2019) and distribution-based joint-semantics similarity (Liu et al. 2020) consider data distributions.”
BackgroundMethodInfluentialLearning Cross-Media Joint Representation With Sparse and Semisupervised Regularization
Zhai, X., Peng, Y., & Xiao, J. (2014) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2013.2276704 · Cited by 294
Cross-Modality Binary Code Learning via Fusion Similarity Hashing
Liu, H., Ji, R., Wu, Y., Huang, F., & Zhang, B. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.672 · Cited by 251 · cited 4×
“We compare DGCPN with nine UCMH methods: CVH (Kumar and Udupa 2011), FSH (Liu et al. 2017), CMFH (Ding, Guo, and Zhou 2014), LSSH (Zhou, Ding, and Guo 2014), UGACH (Zhang, Peng, and Yuan 2018), DJSRH (Su, Zhong, and Zhang 2019), UKD-SS (Hu et al. 2020), DSAH (Yang et al. 2020b), and JDSH (Liu et al. 2020).”
MethodInfluentialUnsupervised Generative Adversarial Cross-Modal Hashing
Zhang, J., Peng, Y., & Yuan, M. (2018) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v32i1.11263 · Cited by 225
Liu, S., Qian, S., Guan, Y., Zhan, J., & Ying, L. (2020) · Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval · doi:10.1145/3397271.3401086 · Cited by 212 · cited 12×
“Both joint-semantics similarity (Su, Zhong, and Zhang 2019) and distribution-based joint-semantics similarity (Liu et al. 2020) consider data distributions.”
BackgroundMethodInfluentialA Comprehensive Survey on Cross-modal Retrieval
Wang, K., Yin, Q., Wang, W., Wu, S., & Wang, L. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1607.06215 · Cited by 208 · cited 4×
“Cross-modal retrieval (CMR) aims to retrieve relevant data when given queries that have different modalities of the retrieval data (Wang et al. 2016).”
BackgroundMethodInfluentialMulti-Task Consistency-Preserving Adversarial Hashing for Cross-Modal Retrieval
Xie, D., Deng, C., Li, C., Liu, X., & Tao, D. (2020) · IEEE Transactions on Image Processing · doi:10.1109/tip.2020.2963957 · Cited by 199 · cited 8×
“Nevertheless, such a connection can be inferred by their neighbor coherence that both “rose” and “wedding” have high similarities of the “bride, rose”. mon spaces, and they suffer from high computation burden for a large volume of multimedia data (Feng, Wang, and Li 2014; Wang et al. 2017; Xie et al. 2020).”
BackgroundInfluentialSelf-citationA Multiview-Based Parameter Free Framework for Group Detection
Li, X., Chen, M., Nie, F., & Wang, Q. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.11208 · Cited by 178 · cited 1×
“In UCMH, data similarities serve as critical considerations (Liang et al. 2016; Li et al. 2017).”
BackgroundCreating Something From Nothing: Unsupervised Knowledge Distillation for Cross-Modal Hashing
Hu, H., Xie, L., Hong, R., & Tian, Q. (2020) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr42600.2020.00319 · Cited by 172
Exploiting Click Constraints and Multi-view Features for Image Re-ranking
Yu, J., Rui, Y., & Chen, B. (2014) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2013.2284755 · Cited by 170 · cited 1×
“Due to the explosion of multimedia data on the Internet (Yu, Rui, and Chen 2013; Yang et al. 2020c), cross-modal retrieval has become an essential technique in information retrieval (Yang et al. 2020c) and multi-modal data mining (Qiao et al. 2019b,a; Yang et al. 2020a,c; Li et al. 2020).”
BackgroundSelf-citationUnsupervised Deep Hashing via Binary Latent Factor Models for Large-scale Cross-modal Retrieval
Wu, G., Lin, Z., Han, J., Liu, L., Ding, G., Zhang, B., & Shen, J. (2018) · Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2018/396 · Cited by 167 · cited 1×
“Nevertheless, the data similarities have been generally used as weights in the optimization functions and did not make full use of data similarities (Wu et al. 2018; Wang et al. 2020; Hu et al. 2020).”
MethodDeep Binary Reconstruction for Cross-Modal Hashing
Hu, D., Nie, F., & Li, X. (2018) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2018.2866771 · Cited by 138
Coupled CycleGAN: Unsupervised Hashing Network for Cross-Modal Retrieval
Li, C., Deng, C. D., Wang, L., Xie, D., & Liu, X. (2019) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v33i01.3301176 · Cited by 133
Learn, Imagine and Create: Text-to-Image Generation from Prior Knowledge
Qiao, T., Zhang, J., Xu, D., & Tao, D. (2019) · neural information processing systems · Cited by 125
Self-citationDeep Semantic-Alignment Hashing for Unsupervised Cross-Modal Retrieval
Yang, D., Wu, D., Zhang, W., Zhang, H., Li, B., & Wang, W. (2020) · International Conference on Multimedia Retrieval · doi:10.1145/3372278.3390673 · Cited by 119 · cited 15×
“In particular, DGCPN outperforms the second-best competitor, DSAH, in MIRFlickr-25K by 4.4%, 4.2%, and 4.0% on 16-bit, 32-bit, and 64-bit for I2T and by 3.1%, 3.9%, and 2.9% on 16-bit, 32-bit, and 64-bit for T2I, respectively.”
BackgroundMethodInfluentialMulti-pathway generative adversarial hashing for unsupervised cross-modal retrieval
Zhang, J., & Peng, Y. (2020) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2019.2922128 · Cited by 103
DeepFacePencil: Creating Face Images from Freehand Sketches
Li, Y., Chen, X., Yang, B., Chen, Z., Cheng, Z., & Zha, Z.-J. (2020) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3394171.3413684 · Cited by 62 · cited 2×
“2020c) and multi-modal data mining (Qiao et al. 2019b,a; Yang et al. 2020a,c; Li et al. 2020).”
BackgroundFang, Y., Zhang, H., & Ren, Y. (2019) · Knowledge-Based Systems · doi:10.1016/j.knosys.2019.02.004 · Cited by 49 · cited 2×
“For example, IMH (Song et al. 2013), CMFH (Ding, Guo, and Zhou 2014), and MSFH (Fang, Zhang, and Ren 2019) learned hashing codes by preserving intra-and inter-graph consistency, using collective matrix factorization, and using multigraph regularized smooth matrix factorization, respectively.”
BackgroundMethodLiong, V. E., Lu, J., & Tan, Y. (2018) · Pattern Recognition · doi:10.1016/j.patcog.2018.02.002 · Cited by 49
Unsupervised multi-graph cross-modal hashing for large-scale multimedia retrieval
Xie, L., Zhu, L., & Chen, G. (2016) · Multimedia Tools and Applications · doi:10.1007/s11042-016-3432-0 · Cited by 47 · cited 1×
“Later, continuous similarities between data were designed by using feature distances (Xie, Zhu, and Chen 2016; Fang, Zhang, and Ren 2019).”
MethodUnsupervised Deep Cross-modal Hashing with Virtual Label Regression
Wang, T., Zhu, L., Cheng, Z., Li, J., & Gao, Z. (2019) · Neurocomputing · doi:10.1016/j.neucom.2019.12.058 · Cited by 45 · cited 3×
“Later, UCMH methods exploited deep neural networks (DNNs) (Wang et al. 2020; Zhang and Tao 2020).”
MethodSelf-Paced Cross-Modal Subspace Matching
Liang, J., Li, Z., Cao, D., He, R., & Wang, J. (2016) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/2911451.2911527 · Cited by 36 · cited 1×
“In UCMH, data similarities serve as critical considerations (Liang et al. 2016; Li et al. 2017).”
BackgroundGenerative Attribute Manipulation Scheme for Flexible Fashion Search
Yang, X., Song, X., Han, X., Wen, H., Nie, J., & Nie, L. (2020) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3397271.3401150 · Cited by 34 · cited 16×
“For example, Hu, Nie, and Li designed a deep autoencoder with an adjusted Tanh function.”
BackgroundMethodInfluentialComprehensive Distance-Preserving Autoencoders for Cross-Modal Retrieval
Zhan, Y., Yu, J., Zhou, Y., Zhang, R., Tao, D., & Tian, Q. (2018) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3240508.3240607 · Cited by 32
Self-citationJoint Sketch-Attribute Learning for Fine-Grained Face Synthesis
Yang, B., Chen, X., Hong, R., Chen, Z., Li, Y., & Zha, Z.-J. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-37731-1_64 · Cited by 10
Cited by
Unsupervised Contrastive Cross-modal Hashing
Hu, P., Zhu, H., Lin, J., Peng, D., Zhao, Y.-P., & Peng, X. (2022) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2022.3177356 · Cited by 199
Metadata
- Authors
- Yu, J., Zhou, H., Zhan, Y., & Tao, D.
- Venue
- Proceedings of the AAAI Conference on Artificial Intelligence
- Published
- 2021-05-18
- Volume
- 35
- Issue
- 5
- Pages
- 4626-4634
- ISSN
- 2159-5399 (ISSN-L)
- 2374-3468
- OpenAlex
- W3175740157
- Cited by
- 197
- References
- 43
- FWCI
- 11.35
- Type
- conference paper
- Language
- en
- Open access
- diamond
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.