{"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/residual-features-and-unified-prediction","title":"Residual Features and Unified Prediction Network for Single Stage Detection","arxiv_id":"1707.05031","date":"2017-07-17","proceeding":null,"authors":["Kyoungmin Lee","Jae-Seok Choi","Jisoo Jeong","Nojun Kwak"],"abstract":"Recently, a lot of single stage detectors using multi-scale features have\nbeen actively proposed. They are much faster than two stage detectors that use\nregion proposal networks (RPN) without much degradation in the detection\nperformances. However, the feature maps in the lower layers close to the input\nwhich are responsible for detecting small objects in a single stage detector\nhave a problem of insufficient representation power because they are too\nshallow. There is also a structural contradiction that the feature maps have to\ndeliver low-level information to next layers as well as contain high-level\nabstraction for prediction. In this paper, we propose a method to enrich the\nrepresentation power of feature maps using Resblock and deconvolution layers.\nIn addition, a unified prediction module is applied to generalize output\nresults and boost earlier layers' representation power for prediction. The\nproposed method enables more precise prediction, which achieved higher score\nthan SSD on PASCAL VOC and MS COCO. In addition, it maintains the advantage of\nfast computation of a single stage detector, which requires much less\ncomputation than other detectors with similar performance. Code is available at\nhttps://github.com/kmlee-snu/run","url_abs":"http://arxiv.org/abs/1707.05031v4","url_pdf":"http://arxiv.org/pdf/1707.05031v4.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":"residual-features-and-unified-prediction","repo_url":"https://github.com/kmlee-snu/run","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"caffe2","reach":null}],"tasks":[{"task_slug":"prediction","task_name":"Prediction"},{"task_slug":"region-proposal","task_name":"Region Proposal"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"non-maximum-suppression","method_name":"Non Maximum Suppression"},{"method_slug":"ssd","method_name":"SSD"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}