{"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/learning-complexity-aware-cascades-for-deep","title":"Learning Complexity-Aware Cascades for Deep Pedestrian Detection","arxiv_id":"1507.05348","date":"2015-07-19","proceeding":"ICCV 2015 12","authors":["Zhaowei Cai","Mohammad Saberian","Nuno Vasconcelos"],"abstract":"The design of complexity-aware cascaded detectors, combining features of very\ndifferent complexities, is considered. A new cascade design procedure is\nintroduced, by formulating cascade learning as the Lagrangian optimization of a\nrisk that accounts for both accuracy and complexity. A boosting algorithm,\ndenoted as complexity aware cascade training (CompACT), is then derived to\nsolve this optimization. CompACT cascades are shown to seek an optimal\ntrade-off between accuracy and complexity by pushing features of higher\ncomplexity to the later cascade stages, where only a few difficult candidate\npatches remain to be classified. This enables the use of features of vastly\ndifferent complexities in a single detector. In result, the feature pool can be\nexpanded to features previously impractical for cascade design, such as the\nresponses of a deep convolutional neural network (CNN). This is demonstrated\nthrough the design of a pedestrian detector with a pool of features whose\ncomplexities span orders of magnitude. The resulting cascade generalizes the\ncombination of a CNN with an object proposal mechanism: rather than a\npre-processing stage, CompACT cascades seamlessly integrate CNNs in their\nstages. This enables state of the art performance on the Caltech and KITTI\ndatasets, at fairly fast speeds.","url_abs":"http://arxiv.org/abs/1507.05348v1","url_pdf":"http://arxiv.org/pdf/1507.05348v1.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":"pedestrian-detection","task_name":"Pedestrian Detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/pedestrian-detection-on-caltech","task":"Pedestrian Detection","dataset":"Caltech","model":"CompACT-Deep","rank_in_archive_order":26,"of":33,"metrics":{"Reasonable Miss Rate":"11.75"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1507.05348","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}