{"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/depthwise-convolution-is-all-you-need-for","title":"Depthwise Convolution is All You Need for Learning Multiple Visual Domains","arxiv_id":"1902.00927","date":"2019-02-03","proceeding":null,"authors":["Yunhui Guo","Yandong Li","Rogerio Feris","Liqiang Wang","Tajana Rosing"],"abstract":"There is a growing interest in designing models that can deal with images\nfrom different visual domains. If there exists a universal structure in\ndifferent visual domains that can be captured via a common parameterization,\nthen we can use a single model for all domains rather than one model per\ndomain. A model aware of the relationships between different domains can also\nbe trained to work on new domains with less resources. However, to identify the\nreusable structure in a model is not easy. In this paper, we propose a\nmulti-domain learning architecture based on depthwise separable convolution.\nThe proposed approach is based on the assumption that images from different\ndomains share cross-channel correlations but have domain-specific spatial\ncorrelations. The proposed model is compact and has minimal overhead when being\napplied to new domains. Additionally, we introduce a gating mechanism to\npromote soft sharing between different domains. We evaluate our approach on\nVisual Decathlon Challenge, a benchmark for testing the ability of multi-domain\nmodels. The experiments show that our approach can achieve the highest score\nwhile only requiring 50% of the parameters compared with the state-of-the-art\napproaches.","url_abs":"http://arxiv.org/abs/1902.00927v2","url_pdf":"http://arxiv.org/pdf/1902.00927v2.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":"depthwise-convolution-is-all-you-need-for","repo_url":"https://github.com/yunhuiguo/Depthwise_Convolution_for_Multiple_Domain_Learning","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"all","task_name":"All"},{"task_slug":"continual-learning","task_name":"Continual Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/continual-learning-on-visual-domain-decathlon","task":"Continual Learning","dataset":"visual domain decathlon (10 tasks)","model":"Depthwise Soft Sharing","rank_in_archive_order":2,"of":14,"metrics":{"decathlon discipline (Score)":"3507"},"uses_additional_data":false},{"leaderboard":"/sota/continual-learning-on-visual-domain-decathlon","task":"Continual Learning","dataset":"visual domain decathlon (10 tasks)","model":"Depthwise Sharing","rank_in_archive_order":4,"of":14,"metrics":{"decathlon discipline (Score)":"3234"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1902.00927","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}