{"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/strong-weak-distribution-alignment-for","title":"Strong-Weak Distribution Alignment for Adaptive Object Detection","arxiv_id":"1812.04798","date":"2018-12-12","proceeding":"CVPR 2019 6","authors":["Kuniaki Saito","Yoshitaka Ushiku","Tatsuya Harada","Kate Saenko"],"abstract":"We propose an approach for unsupervised adaptation of object detectors from\nlabel-rich to label-poor domains which can significantly reduce annotation\ncosts associated with detection. Recently, approaches that align distributions\nof source and target images using an adversarial loss have been proven\neffective for adapting object classifiers. However, for object detection, fully\nmatching the entire distributions of source and target images to each other at\nthe global image level may fail, as domains could have distinct scene layouts\nand different combinations of objects. On the other hand, strong matching of\nlocal features such as texture and color makes sense, as it does not change\ncategory level semantics. This motivates us to propose a novel method for\ndetector adaptation based on strong local alignment and weak global alignment.\nOur key contribution is the weak alignment model, which focuses the adversarial\nalignment loss on images that are globally similar and puts less emphasis on\naligning images that are globally dissimilar. Additionally, we design the\nstrong domain alignment model to only look at local receptive fields of the\nfeature map. We empirically verify the effectiveness of our method on four\ndatasets comprising both large and small domain shifts. Our code is available\nat \\url{https://github.com/VisionLearningGroup/DA_Detection}","url_abs":"http://arxiv.org/abs/1812.04798v3","url_pdf":"http://arxiv.org/pdf/1812.04798v3.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":"strong-weak-distribution-alignment-for","repo_url":"https://github.com/VisionLearningGroup/DA_Detection","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"strong-weak-distribution-alignment-for","repo_url":"https://github.com/harsh-99/SCL","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"unsupervised-domain-adaptation","task_name":"Unsupervised Domain Adaptation"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/unsupervised-domain-adaptation-on-cityscapes-1","task":"Unsupervised Domain Adaptation","dataset":"Cityscapes to Foggy Cityscapes","model":"SWDA","rank_in_archive_order":19,"of":22,"metrics":{"mAP@0.5":"34.8"},"uses_additional_data":false},{"leaderboard":"/sota/unsupervised-domain-adaptation-on-sim10k-to-2","task":"Unsupervised Domain Adaptation","dataset":"SIM10K to BDD100K","model":"SWDA","rank_in_archive_order":3,"of":3,"metrics":{"mAP@0.5":"42.9"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1812.04798","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}