{"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/adaptive-deep-learning-through-visual-domain","title":"Adaptive Deep Learning through Visual Domain Localization","arxiv_id":"1802.08833","date":"2018-02-24","proceeding":null,"authors":["Gabriele Angeletti","Barbara Caputo","Tatiana Tommasi"],"abstract":"A commercial robot, trained by its manufacturer to recognize a predefined\nnumber and type of objects, might be used in many settings, that will in\ngeneral differ in their illumination conditions, background, type and degree of\nclutter, and so on. Recent computer vision works tackle this generalization\nissue through domain adaptation methods, assuming as source the visual domain\nwhere the system is trained and as target the domain of deployment. All\napproaches assume to have access to images from all classes of the target\nduring training, an unrealistic condition in robotics applications. We address\nthis issue proposing an algorithm that takes into account the specific needs of\nrobot vision. Our intuition is that the nature of the domain shift experienced\nmostly in robotics is local. We exploit this through the learning of maps that\nspatially ground the domain and quantify the degree of shift, embedded into an\nend-to-end deep domain adaptation architecture. By explicitly localizing the\nroots of the domain shift we significantly reduce the number of parameters of\nthe architecture to tune, we gain the flexibility necessary to deal with subset\nof categories in the target domain at training time, and we provide a clear\nfeedback on the rationale behind any classification decision, which can be\nexploited in human-robot interactions. Experiments on two different settings of\nthe iCub World database confirm the suitability of our method for robot vision.","url_abs":"http://arxiv.org/abs/1802.08833v1","url_pdf":"http://arxiv.org/pdf/1802.08833v1.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":"adaptive-deep-learning-through-visual-domain","repo_url":"https://github.com/blackecho/LoAd-Network","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"},{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}