{"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/zigzag-learning-for-weakly-supervised-object","title":"Zigzag Learning for Weakly Supervised Object Detection","arxiv_id":"1804.09466","date":"2018-04-25","proceeding":"CVPR 2018 6","authors":["Xiaopeng Zhang","Jiashi Feng","Hongkai Xiong","Qi Tian"],"abstract":"This paper addresses weakly supervised object detection with only image-level\nsupervision at training stage. Previous approaches train detection models with\nentire images all at once, making the models prone to being trapped in\nsub-optimums due to the introduced false positive examples. Unlike them, we\npropose a zigzag learning strategy to simultaneously discover reliable object\ninstances and prevent the model from overfitting initial seeds. Towards this\ngoal, we first develop a criterion named mean Energy Accumulation Scores (mEAS)\nto automatically measure and rank localization difficulty of an image\ncontaining the target object, and accordingly learn the detector progressively\nby feeding examples with increasing difficulty. In this way, the model can be\nwell prepared by training on easy examples for learning from more difficult\nones and thus gain a stronger detection ability more efficiently. Furthermore,\nwe introduce a novel masking regularization strategy over the high level\nconvolutional feature maps to avoid overfitting initial samples. These two\nmodules formulate a zigzag learning process, where progressive learning\nendeavors to discover reliable object instances, and masking regularization\nincreases the difficulty of finding object instances properly. We achieve 47.6%\nmAP on PASCAL VOC 2007, surpassing the state-of-the-arts by a large margin.","url_abs":"http://arxiv.org/abs/1804.09466v1","url_pdf":"http://arxiv.org/pdf/1804.09466v1.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":[],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"weakly-supervised-object-detection","task_name":"Weakly Supervised Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/weakly-supervised-object-detection-on-pascal-1","task":"Weakly Supervised Object Detection","dataset":"PASCAL VOC 2007","model":"ZLDN-L","rank_in_archive_order":24,"of":41,"metrics":{"MAP":"47.6"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-object-detection-on-pascal","task":"Weakly Supervised Object Detection","dataset":"PASCAL VOC 2012 test","model":"ZLDN-L","rank_in_archive_order":22,"of":32,"metrics":{"MAP":"42.9"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.09466","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}