[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-153382-105":59,"doc-detail-153382-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","domes-to-drones-self-supervised-active-triangulation-for-3d-human-pose-reconstruction","Domes to Drones - Self-Supervised Active Triangulation for 3D Human Pose Reconstruction","","Existing systems estimate multiple 2D human poses in images reliably, but monocular 3D pose estimation remains ill-posed because depth ambiguities and occlusions limit what can be inferred from a single viewpoint. With multiple cameras or an active observer that changes viewpoint, 3D reconstruction from 2D measurements becomes tractable using multi-view geometry. The paper introduces ACTOR, an active triangulation agent combining a 2D pose network and a reinforcement learning policy that adaptively selects informative camera viewpoints via self-supervised training, improving 3D accuracy on complex multi-person scenes and suggesting a path toward drone-based observation.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/domes-to-drones-self-supervised-active-triangulation-for-3d-human-pose-reconstruction/153382/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/domes-to-drones-self-supervised-active-triangulation-for-3d-human-pose-reconstruction/153382.png","ImageObject",300,407,{"name":92,"@type":93},"Rowan","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-17","2026-08-27",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",11,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"Why is single-image 3D human pose estimation ill-posed?","Question",{"text":113,"@type":114},"It is ill-posed due to depth ambiguities and occlusions, which prevent uniquely determining 3D structure from a single viewpoint.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"What is ACTOR and what problem does it address?",{"text":118,"@type":114},"ACTOR is an active triangulation agent for 3D human pose reconstruction. It addresses the view selection problem by choosing informative camera viewpoints to improve reconstruction accuracy in complex scenes.",{"name":120,"@type":111,"acceptedAnswer":121},"How is ACTOR trained without pose annotations?",{"text":122,"@type":114},"Given a pre-trained 2D pose estimator, ACTOR trains its viewpoint selection policy in a self-supervised manner, avoiding the need for 2D or 3D pose annotations.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},153382,1787875006,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":105,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":130,"read_time":144,"preview_page_count":4},1099514067415,"https://ap-avatar.wpscdn.com/avatar/100002539d78ffe74a7?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779092875211072502","Domes to Drones: Self-Supervised Active Triangulation for 3D Human Pose Reconstruction  \nAleksis Pirinen, 1 􀀃 Erik Gärtner 1􀀃 and Cristian Sminchisescu 1 ;2  \n1Department of Mathematics, Faculty of Engineering, Lund University  \n2 Google Research  \n{aleksis .pirinen, erik .gartner, cristian .sminchisescu}@math .lth .se  \nAbstract  \nExisting state-of-the-art estimation systems can detect 2d poses of multiple people in images quite reliably. In contrast, 3d pose estimation from a single image is ill-posed due to occlusion and depth ambiguities. Assuming access to multiple cameras, or given an active system able to position itself to observe the scene from multiple viewpoints, reconstructing 3d pose from 2d measurements becomes well-posed within the framework of standard multi-view geometry. Less clear is what is an informative set of viewpoints for accurate 3d reconstruction, particularly in complex scenes, where people are occluded by others or by scene objects. In order to address the view selection problem in a principled way, we here introduce ACTOR, an active triangulation agent for 3d human pose reconstruction. Our fully trainable agent consists of a 2d pose estimation network (any of which would work) and a deep reinforcement learning-based policy for camera viewpoint selection. The policy predicts observation viewpoints, the number of which varies adaptively depending on scene content, and the associated images are fed to an underlying pose estimator. Importantly, training the view selection policy requires no annotations – given a pre-trained 2d pose estimator, ACTOR is trained in a self-supervised manner. In extensive evaluations on complex multi-people scenes filmed in a Panoptic dome, under multiple viewpoints, we compare our active triangulation agent to strong multi-view baselines, and show that ACTOR produces significantly more accurate 3d pose reconstructions. We also provide a proof-ofconcept experiment indicating the potential of connecting our view selection policy to a physical drone observer.  \n1 Introduction  \nEstimating 2d and 3d human pose from given images or video is an active research area, with deep learning playing a prominent role in most of today's state-of-the-art pose and shape estimation models [2, 4, 21–23, 27, 34] . Monocular 3d pose estimation is however ill-posed [26] due to depth ambiguities, and these cannot always be resolved by priors or by increasing a feed-forward model's predictive power. Given access to multiple cameras, or given an active observer which can capture images from multiple viewpoints, reconstructing 3d pose from 2d estimates however becomes tractable within the framework of standard multi-view geometry. An active setup for triangulating 2d estimates also addresses many common practical issues, such as partial observability due to occlusion, either self-induced or due to other people or objects.  \nGiven sufficiently many viewpoints, 3d pose reconstructions from 2d estimates can be made robust and accurate, and such results have even been used as (pseudo)ground-truth [17, 33] . While inferring  \n􀀃 Denotes equal contribution, order determined by coin flip.  \n33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Vancouver, Canada.  \nnext camera location (􀀢 \\#, 􀀢 %)  \n􀀧∗)  \nFigure 1: Overview of ACTOR, our active triangulation agent for 3d human pose reconstruction. A random view is initially given and the image is fed to a 2d body joint predictor yielding estimates for all visible people (X ti) and the core of the agent's state (Bt ) . The policy network predicts camera locations until all joints have been triangulated, then switches to the next active-view at time t + 1 . The predicted camera location is encoded via spherical angles relative to the agent's location on the viewing sphere, and the closest camera is selected. When the agent is done it outputs Y t?, the final 3d reconstructions of all people in the scene, which are obtained by a combination of sp","cbCaicEzCkDQ4NLO","https://ap.wps.com/l/cbCaicEzCkDQ4NLO","pdf",6130766,"English","# Abstract\n# Introduction\n## Problem setting: 2D vs 3D pose\n## Active viewpoint selection and self-supervision\n# Figure 1: ACTOR overview\n## Agent workflow\n# Related Work\n## Lifting 2D to 3D","[{\"question\":\"Why is single-image 3D human pose estimation ill-posed?\",\"answer\":\"It is ill-posed due to depth ambiguities and occlusions, which prevent uniquely determining 3D structure from a single viewpoint.\"},{\"question\":\"What is ACTOR and what problem does it address?\",\"answer\":\"ACTOR is an active triangulation agent for 3D human pose reconstruction. It addresses the view selection problem by choosing informative camera viewpoints to improve reconstruction accuracy in complex scenes.\"},{\"question\":\"How is ACTOR trained without pose annotations?\",\"answer\":\"Given a pre-trained 2D pose estimator, ACTOR trains its viewpoint selection policy in a self-supervised manner, avoiding the need for 2D or 3D pose annotations.\"}]","Domes to Drones - Self-Supervised Active Triangulation for 3D Human Pose Reconstruction | PDF",28]