{"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/conditional-similarity-networks","title":"Conditional Similarity Networks","arxiv_id":"1603.07810","date":"2016-03-25","proceeding":"CVPR 2017 7","authors":["Andreas Veit","Serge Belongie","Theofanis Karaletsos"],"abstract":"What makes images similar? To measure the similarity between images, they are\ntypically embedded in a feature-vector space, in which their distance preserve\nthe relative dissimilarity. However, when learning such similarity embeddings\nthe simplifying assumption is commonly made that images are only compared to\none unique measure of similarity. A main reason for this is that contradicting\nnotions of similarities cannot be captured in a single space. To address this\nshortcoming, we propose Conditional Similarity Networks (CSNs) that learn\nembeddings differentiated into semantically distinct subspaces that capture the\ndifferent notions of similarities. CSNs jointly learn a disentangled embedding\nwhere features for different similarities are encoded in separate dimensions as\nwell as masks that select and reweight relevant dimensions to induce a subspace\nthat encodes a specific similarity notion. We show that our approach learns\ninterpretable image representations with visually relevant semantic subspaces.\nFurther, when evaluating on triplet questions from multiple similarity notions\nour model even outperforms the accuracy obtained by training individual\nspecialized networks for each notion separately.","url_abs":"http://arxiv.org/abs/1603.07810v3","url_pdf":"http://arxiv.org/pdf/1603.07810v3.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":"conditional-similarity-networks","repo_url":"https://github.com/andreasveit/conditional-similarity-networks","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"conditional-similarity-networks","repo_url":"https://github.com/Maryeon/asen","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"conditional-similarity-networks","repo_url":"https://github.com/dr-lingxiao/ag-man","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"conditional-similarity-networks","repo_url":"https://github.com/mvasil/fashion-compatibility","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"conditional-similarity-networks","repo_url":"https://github.com/rxtan2/Learning-Similarity-Conditions","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":null,"task_name":"Triplet"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}