{"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/grounded-human-object-interaction-hotspots","title":"Grounded Human-Object Interaction Hotspots from Video","arxiv_id":"1812.04558","date":"2018-12-11","proceeding":"ICCV 2019 10","authors":["Tushar Nagarajan","Christoph Feichtenhofer","Kristen Grauman"],"abstract":"Learning how to interact with objects is an important step towards embodied\nvisual intelligence, but existing techniques suffer from heavy supervision or\nsensing requirements. We propose an approach to learn human-object interaction\n\"hotspots\" directly from video. Rather than treat affordances as a manually\nsupervised semantic segmentation task, our approach learns about interactions\nby watching videos of real human behavior and anticipating afforded actions.\nGiven a novel image or video, our model infers a spatial hotspot map indicating\nhow an object would be manipulated in a potential interaction-- even if the\nobject is currently at rest. Through results with both first and third person\nvideo, we show the value of grounding affordances in real human-object\ninteractions. Not only are our weakly supervised hotspots competitive with\nstrongly supervised affordance methods, but they can also anticipate object\ninteraction for novel object categories.","url_abs":"http://arxiv.org/abs/1812.04558v2","url_pdf":"http://arxiv.org/pdf/1812.04558v2.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":"grounded-human-object-interaction-hotspots","repo_url":"https://github.com/Tushar-N/interaction-hotspots","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"human-object-interaction-detection","task_name":"Human-Object Interaction Detection"},{"task_slug":"object","task_name":"Object"},{"task_slug":"object-recognition","task_name":"Object Recognition"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"video-to-image-affordance-grounding","task_name":"Video-to-image Affordance Grounding"}],"methods":[],"datasets_introduced":[{"slug":"epic-hotspot","name":"EPIC-Hotspot","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/video-to-image-affordance-grounding-on-epic","task":"Video-to-image Affordance Grounding","dataset":"EPIC-Hotspot","model":"Hotspot","rank_in_archive_order":3,"of":3,"metrics":{"AUC-J":"0.79","KLD":"1.26","SIM":"0.40"},"uses_additional_data":false},{"leaderboard":"/sota/video-to-image-affordance-grounding-on-opra-1","task":"Video-to-image Affordance Grounding","dataset":"OPRA (28x28)","model":"Hotspot","rank_in_archive_order":4,"of":4,"metrics":{"AUC-J":"0.81","KLD":"1.47","SIM":"0.36"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1812.04558","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}