{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/modality-specific-cross-modal-similarity","title":"Modality-specific Cross-modal Similarity Measurement with Recurrent Attention Network","arxiv_id":"1708.04776","date":"2017-08-16","proceeding":null,"authors":["Yuxin Peng","Jinwei Qi","Yuxin Yuan"],"abstract":"Nowadays, cross-modal retrieval plays an indispensable role to flexibly find\ninformation across different modalities of data. Effectively measuring the\nsimilarity between different modalities of data is the key of cross-modal\nretrieval. Different modalities such as image and text have imbalanced and\ncomplementary relationships, which contain unequal amount of information when\ndescribing the same semantics. For example, images often contain more details\nthat cannot be demonstrated by textual descriptions and vice versa. Existing\nworks based on Deep Neural Network (DNN) mostly construct one common space for\ndifferent modalities to find the latent alignments between them, which lose\ntheir exclusive modality-specific characteristics. Different from the existing\nworks, we propose modality-specific cross-modal similarity measurement (MCSM)\napproach by constructing independent semantic space for each modality, which\nadopts end-to-end framework to directly generate modality-specific cross-modal\nsimilarity without explicit common representation. For each semantic space,\nmodality-specific characteristics within one modality are fully exploited by\nrecurrent attention network, while the data of another modality is projected\ninto this space with attention based joint embedding to utilize the learned\nattention weights for guiding the fine-grained cross-modal correlation\nlearning, which can capture the imbalanced and complementary relationships\nbetween different modalities. Finally, the complementarity between the semantic\nspaces for different modalities is explored by adaptive fusion of the\nmodality-specific cross-modal similarities to perform cross-modal retrieval.\nExperiments on the widely-used Wikipedia and Pascal Sentence datasets as well\nas our constructed large-scale XMediaNet dataset verify the effectiveness of\nour proposed approach, outperforming 9 state-of-the-art methods.","url_abs":"http://arxiv.org/abs/1708.04776v1","url_pdf":"http://arxiv.org/pdf/1708.04776v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"modality-specific-cross-modal-similarity","repo_url":"https://github.com/PKU-ICST-MIPL/MCSM_TIP2018","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"torch","reach":null}],"tasks":[{"task_slug":"cross-modal-retrieval","task_name":"Cross-Modal Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"sentence","task_name":"Sentence"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1708.04776","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}