{"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/context-aware-deep-feature-compression-for","title":"Context-aware Deep Feature Compression for High-speed Visual Tracking","arxiv_id":"1803.10537","date":"2018-03-28","proceeding":"CVPR 2018 6","authors":["Jongwon Choi","Hyung Jin Chang","Tobias Fischer","Sangdoo Yun","Kyuewang Lee","Jiyeoup Jeong","Yiannis Demiris","Jin Young Choi"],"abstract":"We propose a new context-aware correlation filter based tracking framework to\nachieve both high computational speed and state-of-the-art performance among\nreal-time trackers. The major contribution to the high computational speed lies\nin the proposed deep feature compression that is achieved by a context-aware\nscheme utilizing multiple expert auto-encoders; a context in our framework\nrefers to the coarse category of the tracking target according to appearance\npatterns. In the pre-training phase, one expert auto-encoder is trained per\ncategory. In the tracking phase, the best expert auto-encoder is selected for a\ngiven target, and only this auto-encoder is used. To achieve high tracking\nperformance with the compressed feature map, we introduce extrinsic denoising\nprocesses and a new orthogonality loss term for pre-training and fine-tuning of\nthe expert auto-encoders. We validate the proposed context-aware framework\nthrough a number of experiments, where our method achieves a comparable\nperformance to state-of-the-art trackers which cannot run in real-time, while\nrunning at a significantly fast speed of over 100 fps.","url_abs":"http://arxiv.org/abs/1803.10537v1","url_pdf":"http://arxiv.org/pdf/1803.10537v1.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":"context-aware-deep-feature-compression-for","repo_url":"https://github.com/jongwon20000/TRACA","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"feature-compression","task_name":"Feature Compression"},{"task_slug":"visual-object-tracking","task_name":"Visual Object Tracking"},{"task_slug":"visual-tracking","task_name":"Visual Tracking"},{"task_slug":"high","task_name":"Vocal Bursts Intensity Prediction"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/visual-object-tracking-on-vot201718","task":"Visual Object Tracking","dataset":"VOT2017/18","model":"TRACA","rank_in_archive_order":15,"of":15,"metrics":{"Expected Average Overlap (EAO)":"0.137"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1803.10537","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}