Research Article
Critical Separation Hashing for Cross-Modal Retrieval
@INPROCEEDINGS{10.1007/978-3-031-36011-4_15, author={Zening Wang and Yungong Sun and Liang Liu and Ao Li}, title={Critical Separation Hashing for Cross-Modal Retrieval}, proceedings={6GN for Future Wireless Networks. 5th EAI International Conference, 6GN 2022, Harbin, China, December 17-18, 2022, Proceedings, Part I}, proceedings_a={6GN}, year={2023}, month={7}, keywords={Unsupervised Cross-modal Smilarity Matrix}, doi={10.1007/978-3-031-36011-4_15} }- Zening Wang
Yungong Sun
Liang Liu
Ao Li
Year: 2023
Critical Separation Hashing for Cross-Modal Retrieval
6GN
Springer
DOI: 10.1007/978-3-031-36011-4_15
Abstract
With the development of Internet technology, unimodal retrieval techniques are no longer suitable for the current environment, and mutual retrieval between multiple modalities is needed to obtain more complete information. Deep hashing has clearly become a simpler and faster method in cross-modal hashing. In recent years, unsupervised cross-modal hashing has received increasing attention. However, existing methods fail to exploit the common information across modalities, thus resulting in information wastage. In this paper, we propose a new critical separation cross-modal hashing (CSCH) for unsupervised cross-modal retrieval, which explores the similarity information across modalities by highlighting the similarity between instances to help the network learn the hash function, and we carefully design the loss function by introducing the likelihood loss commonly used in supervised learning into the loss function. Extensive experiments on two cross-modal retrieval datasets show that CSCH has better performance.
