{"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/high-speed-tracking-with-kernelized","title":"High-Speed Tracking with Kernelized Correlation Filters","arxiv_id":"1404.7584","date":"2014-04-30","proceeding":null,"authors":["João F. Henriques","Rui Caseiro","Pedro Martins","Jorge Batista"],"abstract":"The core component of most modern trackers is a discriminative classifier,\ntasked with distinguishing between the target and the surrounding environment.\nTo cope with natural image changes, this classifier is typically trained with\ntranslated and scaled sample patches. Such sets of samples are riddled with\nredundancies -- any overlapping pixels are constrained to be the same. Based on\nthis simple observation, we propose an analytic model for datasets of thousands\nof translated patches. By showing that the resulting data matrix is circulant,\nwe can diagonalize it with the Discrete Fourier Transform, reducing both\nstorage and computation by several orders of magnitude. Interestingly, for\nlinear regression our formulation is equivalent to a correlation filter, used\nby some of the fastest competitive trackers. For kernel regression, however, we\nderive a new Kernelized Correlation Filter (KCF), that unlike other kernel\nalgorithms has the exact same complexity as its linear counterpart. Building on\nit, we also propose a fast multi-channel extension of linear correlation\nfilters, via a linear kernel, which we call Dual Correlation Filter (DCF). Both\nKCF and DCF outperform top-ranking trackers such as Struck or TLD on a 50\nvideos benchmark, despite running at hundreds of frames-per-second, and being\nimplemented in a few lines of code (Algorithm 1). To encourage further\ndevelopments, our tracking framework was made open-source.","url_abs":"http://arxiv.org/abs/1404.7584v3","url_pdf":"http://arxiv.org/pdf/1404.7584v3.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":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/Paulpauladq/HighSpeedFaceTracker","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/SonginCV/GMPHD_MAF","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/SonginCV/GMPHD_SAF","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/SonginCV/MAF_HDA","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/Superlee506/Correlation_Filtering_Tracking_Lib","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"GPL-3.0"}},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/ZC119/richardson_tracking","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/fanshixiong/tracker","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/yumaloop/CV_report2019","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"high-speed-tracking-with-kernelized","repo_url":"https://github.com/oylz/FDSSTTest","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"video-object-tracking","task_name":"Video Object Tracking"},{"task_slug":"high","task_name":"Vocal Bursts Intensity Prediction"},{"task_slug":"regression-1","task_name":"regression"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/video-object-tracking-on-nv-vot211","task":"Video Object Tracking","dataset":"NT-VOT211","model":"KCF(HOG)","rank_in_archive_order":38,"of":43,"metrics":{"AUC":"24.10","Precision":"32.06"},"uses_additional_data":false},{"leaderboard":"/sota/video-object-tracking-on-nv-vot211","task":"Video Object Tracking","dataset":"NT-VOT211","model":"CSK","rank_in_archive_order":41,"of":43,"metrics":{"AUC":"21.31","Precision":"26.51"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1404.7584","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}