{"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/eco-efficient-convolution-operators-for","title":"ECO: Efficient Convolution Operators for Tracking","arxiv_id":"1611.09224","date":"2016-11-28","proceeding":"CVPR 2017 7","authors":["Martin Danelljan","Goutam Bhat","Fahad Shahbaz Khan","Michael Felsberg"],"abstract":"In recent years, Discriminative Correlation Filter (DCF) based methods have\nsignificantly advanced the state-of-the-art in tracking. However, in the\npursuit of ever increasing tracking performance, their characteristic speed and\nreal-time capability have gradually faded. Further, the increasingly complex\nmodels, with massive number of trainable parameters, have introduced the risk\nof severe over-fitting. In this work, we tackle the key causes behind the\nproblems of computational complexity and over-fitting, with the aim of\nsimultaneously improving both speed and performance.\n  We revisit the core DCF formulation and introduce: (i) a factorized\nconvolution operator, which drastically reduces the number of parameters in the\nmodel; (ii) a compact generative model of the training sample distribution,\nthat significantly reduces memory and time complexity, while providing better\ndiversity of samples; (iii) a conservative model update strategy with improved\nrobustness and reduced complexity. We perform comprehensive experiments on four\nbenchmarks: VOT2016, UAV123, OTB-2015, and TempleColor. When using expensive\ndeep features, our tracker provides a 20-fold speedup and achieves a 13.0%\nrelative gain in Expected Average Overlap compared to the top ranked method in\nthe VOT2016 challenge. Moreover, our fast variant, using hand-crafted features,\noperates at 60 Hz on a single CPU, while obtaining 65.0% AUC on OTB-2015.","url_abs":"http://arxiv.org/abs/1611.09224v2","url_pdf":"http://arxiv.org/pdf/1611.09224v2.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":"eco-efficient-convolution-operators-for","repo_url":"https://github.com/fengyang95/pyCFTrackers","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"eco-efficient-convolution-operators-for","repo_url":"https://github.com/martin-danelljan/ECO","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"GPL-3.0"}},{"paper_slug":"eco-efficient-convolution-operators-for","repo_url":"https://github.com/2023-MindSpore-1/ms-code-212/tree/main/ecolite","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"eco-efficient-convolution-operators-for","repo_url":"https://github.com/2023-MindSpore-4/Code3/tree/main/ecolite","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"eco-efficient-convolution-operators-for","repo_url":"https://github.com/MindSpore-MS-Code2/code0/tree/main/ecolite","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":null,"task_name":"CPU"},{"task_slug":"diversity","task_name":"Diversity"},{"task_slug":"visual-object-tracking","task_name":"Visual Object Tracking"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/visual-object-tracking-on-trackingnet","task":"Visual Object Tracking","dataset":"TrackingNet","model":"ECO","rank_in_archive_order":36,"of":40,"metrics":{"Accuracy":"56.13","Normalized Precision":"62.14","Precision":"48.86"},"uses_additional_data":false},{"leaderboard":"/sota/visual-object-tracking-on-vot201718","task":"Visual Object Tracking","dataset":"VOT2017/18","model":"ECO","rank_in_archive_order":13,"of":15,"metrics":{"Expected Average Overlap (EAO)":"0.280"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1611.09224","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}