Answers are grounded in this paper and cite their sources.
As a result of the massive expansion of multimodal data in recent years, researchers have developed a strong interest in the problem of cross-modal retrieval. Finding instances from multimodal data that are semantically related is the aim of cross-modal retrieval. However, the distribution of data across different modalities is inconsistent and there is a heterogeneity gap. At the same time, the semantic distinction between low-level and high-level semantic information makes cross-modal retrieval work challenging. This paper proposes a deep adversarial common subspace learning approach based on real-valued representation learning for image-to-text cross-modal retrieval., which combines adversarial learning with common subspaces to get imagetext feature representations of the same dimension in the common space. Secondly, it effectively utilizes label prediction and triplet loss to conduct constraint learning on the network, further improving the retrieval accuracy. Finally, the method achieves promising results on two public datasets on cross-modal retrieval tasks, proving its effective in cross-modal retrieval.
No takes yet. Share an insight, caveat, or question.
Ma et al. (2023) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: