{"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/good-features-to-correlate-for-visual","title":"Good Features to Correlate for Visual Tracking","arxiv_id":"1704.06326","date":"2017-04-20","proceeding":null,"authors":["Erhan Gundogdu","A. Aydin Alatan"],"abstract":"During the recent years, correlation filters have shown dominant and\nspectacular results for visual object tracking. The types of the features that\nare employed in these family of trackers significantly affect the performance\nof visual tracking. The ultimate goal is to utilize robust features invariant\nto any kind of appearance change of the object, while predicting the object\nlocation as properly as in the case of no appearance change. As the deep\nlearning based methods have emerged, the study of learning features for\nspecific tasks has accelerated. For instance, discriminative visual tracking\nmethods based on deep architectures have been studied with promising\nperformance. Nevertheless, correlation filter based (CFB) trackers confine\nthemselves to use the pre-trained networks which are trained for object\nclassification problem. To this end, in this manuscript the problem of learning\ndeep fully convolutional features for the CFB visual tracking is formulated. In\norder to learn the proposed model, a novel and efficient backpropagation\nalgorithm is presented based on the loss function of the network. The proposed\nlearning framework enables the network model to be flexible for a custom\ndesign. Moreover, it alleviates the dependency on the network trained for\nclassification. Extensive performance analysis shows the efficacy of the\nproposed custom design in the CFB tracking framework. By fine-tuning the\nconvolutional parts of a state-of-the-art network and integrating this model to\na CFB tracker, which is the top performing one of VOT2016, 18% increase is\nachieved in terms of expected average overlap, and tracking failures are\ndecreased by 25%, while maintaining the superiority over the state-of-the-art\nmethods in OTB-2013 and OTB-2015 tracking datasets.","url_abs":"http://arxiv.org/abs/1704.06326v2","url_pdf":"http://arxiv.org/pdf/1704.06326v2.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":"good-features-to-correlate-for-visual","repo_url":"https://github.com/egundogdu/CFCF","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"object","task_name":"Object"},{"task_slug":"object-tracking","task_name":"Object Tracking"},{"task_slug":"visual-object-tracking","task_name":"Visual Object Tracking"},{"task_slug":"visual-tracking","task_name":"Visual Tracking"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/visual-object-tracking-on-vot2016","task":"Visual Object Tracking","dataset":"VOT2016","model":"CFCF","rank_in_archive_order":2,"of":6,"metrics":{"Expected Average Overlap (EAO)":"0.3903"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}