{"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/why-do-linear-svms-trained-on-hog-features","title":"Why do linear SVMs trained on HOG features perform so well?","arxiv_id":"1406.2419","date":"2014-06-10","proceeding":null,"authors":["Hilton Bristow","Simon Lucey"],"abstract":"Linear Support Vector Machines trained on HOG features are now a de facto\nstandard across many visual perception tasks. Their popularisation can largely\nbe attributed to the step-change in performance they brought to pedestrian\ndetection, and their subsequent successes in deformable parts models. This\npaper explores the interactions that make the HOG-SVM symbiosis perform so\nwell. By connecting the feature extraction and learning processes rather than\ntreating them as disparate plugins, we show that HOG features can be viewed as\ndoing two things: (i) inducing capacity in, and (ii) adding prior to a linear\nSVM trained on pixels. From this perspective, preserving second-order\nstatistics and locality of interactions are key to good performance. We\ndemonstrate surprising accuracy on expression recognition and pedestrian\ndetection tasks, by assuming only the importance of preserving such local\nsecond-order interactions.","url_abs":"http://arxiv.org/abs/1406.2419v1","url_pdf":"http://arxiv.org/pdf/1406.2419v1.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":"why-do-linear-svms-trained-on-hog-features","repo_url":"https://github.com/RashadGarayev/PersonDetection","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"why-do-linear-svms-trained-on-hog-features","repo_url":"https://github.com/dansthemanwhosakid/image-classification-cancer","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"pedestrian-detection","task_name":"Pedestrian Detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}