Dynamic label correlations and dual-semantic enhancement learning for cross-modal retrieval
Shaohua Teng, Ziye Fang, Zefeng Zheng, Naiqi Wu, Wei Zhang, Luyao Teng · Neurocomputing · 2025
With the rapid growth of multi-modal data, Cross-Modal Hashing (CMH) is widely applied due to its outstanding performance in both search and storage. Nevertheless, there are two issues to be further addressed: (1) most existing methods neglect dynamic learning of the importance of different labels; and (2) many methods fail to purify the consistency of data extracted from different feature spaces. For this purpose, we propose a method called Dynamic Label Correlations and Dual-Semantic Enhancement Learning for Cross-Modal Retrieval (DLCDE) in this study. This method is formed of two parts: Label Semantic Enhancement with Dynamic Label Reconstruction (LSEDLR) and Sample Semantic Enhancement with Consistency Purification and Structure Maintenance (SECPSM). The former first utilizes label-wise self-expression to dynamically explore the latent correlations between different labels and then employs a graph-based manifold regularizer to explore the structural relationships in the transformed label space to enhance label semantics, the latter leverages Hadamard-Product-based Matrix Factorization to enhance the common relationships between samples, thereby enhancing the sample semantics of the latent shared space. Moreover, dual-semantic enhancement learning is achieved by integrating enhanced label semantics and sample semantics in Distance-Distance Difference Minimization (DDDM). Numerous experiments on four benchmark datasets reveal that DLCDE surpasses a number of state-of-the-art CMH methods . The source code for DLCDE is publicly available at https://github.com/Fizzyf/DLCDE .