{"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/learning-to-detect-and-track-visible-and","title":"Learning to Detect and Track Visible and Occluded Body Joints in a Virtual World","arxiv_id":"1803.08319","date":"2018-03-22","proceeding":"ECCV 2018 9","authors":["Matteo Fabbri","Fabio Lanzi","Simone Calderara","Andrea Palazzi","Roberto Vezzani","Rita Cucchiara"],"abstract":"Multi-People Tracking in an open-world setting requires a special effort in\nprecise detection. Moreover, temporal continuity in the detection phase gains\nmore importance when scene cluttering introduces the challenging problems of\noccluded targets. For the purpose, we propose a deep network architecture that\njointly extracts people body parts and associates them across short temporal\nspans. Our model explicitly deals with occluded body parts, by hallucinating\nplausible solutions of not visible joints. We propose a new end-to-end\narchitecture composed by four branches (visible heatmaps, occluded heatmaps,\npart affinity fields and temporal affinity fields) fed by a time linker feature\nextractor. To overcome the lack of surveillance data with tracking, body part\nand occlusion annotations we created the vastest Computer Graphics dataset for\npeople tracking in urban scenarios by exploiting a photorealistic videogame. It\nis up to now the vastest dataset (about 500.000 frames, almost 10 million body\nposes) of human body parts for people tracking in urban scenarios. Our\narchitecture trained on virtual data exhibits good generalization capabilities\nalso on public real tracking benchmarks, when image resolution and sharpness\nare high enough, producing reliable tracklets useful for further batch data\nassociation or re-id modules.","url_abs":"http://arxiv.org/abs/1803.08319v3","url_pdf":"http://arxiv.org/pdf/1803.08319v3.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":"learning-to-detect-and-track-visible-and","repo_url":"https://github.com/fabbrimatteo/JTA-Dataset","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"learning-to-detect-and-track-visible-and","repo_url":"https://github.com/fabbrimatteo/jta-mods","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[],"methods":[],"datasets_introduced":[{"slug":"jta","name":"JTA","full_name":"Joint Track Auto"}],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.08319","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}