[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85778-en":3,"doc-seo-85778-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85778,2336464648322,"Aria","https://ap-avatar.wpscdn.com/avatar/2200025388227c56fec?_k=1778556882303663488",8,"Research & Report","UniPose9D Universal Category-Agnostic Object Pose Estimation","Object pose estimation in 3D vision often overfits to benchmarks and generalizes poorly to new categories and unseen scenes. UniPose9D introduces a category-agnostic foundation model that predicts full 9D pose (rotation, translation, and metric size) from an instance mask/ROI using RGB-D or RGB with predicted depth. It samples point pairs from observed geometry, uses DINOv2 and PointNet features to predict NOCS coordinates, and estimates pose with an adaptive point-pair RANSAC N-hop Kabsch–Umeyama. Flow matching addresses symmetry ambiguities and large-scale training curation improves in-the-wild generalization across six datasets, matching or surpassing specialist methods.","UniPose9D: Universal Category-Agnostic Object Pose  \nEstimation  \nYang You  \nStanford University [yangyou@stanford.edu](yangyou@stanford.edu)  \nYi Du  \nStanford University [duyi@stanford.edu](duyi@stanford.edu)  \nCole Harrison  \nAmazon  \n[colha@amazon.com](colha@amazon.com)  \narXiv :2607 .09985v1 [ cs .CV] 10 Jul 2026  \nLeonidas Guibas  \nStanford University  \n[guibas@cs.stanford.edu](guibas@cs.stanford.edu)  \nAbstract  \nObject pose estimation is a fundamental problem in 3D vision. Although recent state-of-the-art approaches achieve strong performance, they often overfit to existing benchmarks and exhibit limited generalization to novel categories and unseen scenes. We propose UniPose9D, a category-agnostic foundation model for 9D object pose estimation: given an instance mask/ROI and either an RGB-D observation or an RGB image with predicted depth, the model estimates rotation, translation, and metric size without category labels, CAD models, mean-shape priors, or reference views. Specifically, UniPose9D samples point pairs from the observed object geometry and uses DINOv2 and PointNet features to predict NOCS coordinates for each pair. To improve accuracy, we introduce a point-pair-based RANSAC N-hop Kabsch–Umeyama algorithm with an adaptive threshold. We further employ flow matching to address symmetric ambiguities and construct a large-scale training set by curating and aligning pose annotations from existing public datasets. Experiments across six datasets show that a single unified model can match or surpass specialist methods while generalizing to unseen objects and in-the-wild scenarios. Our code and model are available at [https://github.com/qq456cvb/UniPose9D](https://github.com/qq456cvb/UniPose9D).  \nFigure 1: UniPose9D. Given an instance mask/ROI and either RGB-D input (left four examples) or RGB input with predicted depth (right three examples), we predict a full 9D pose and metric 3D bounding box without category labels, CAD models, mean-shape priors, or reference views. Examples include cluttered tabletops, severe occlusions, a robotic manipulation scene, and in-the-wild photographs of novel objects. Yellow cuboids denote predicted metric boxes, and the red, green, and blue axes indicate orientation. The same model generalizes across scenes and categories.  \nPreprint.  \n1 Introduction  \nObject pose estimation is a fundamental problem in computer vision with wide impact on robotic manipulation [7, 36, 38, 44] and augmented reality [22, 23, 28] . Traditional instance-level methods [8, 32, 39] are tailored to specific objects and typically rely on textured CAD models to synthesize training data, which prevents generalization to unseen instances at test time. Category-level approaches [2, 15, 17, 33, 47] remove the need for instance-wise training and meshes, yet they usually assume category-level supervision, category-specific training, or mean-shape priors and therefore remain tied to predefined taxonomies.  \nTo mitigate these limitations, recent works aim to estimate the pose of arbitrary objects at inference time by providing a reference 3D textured model or a set of reference images. While this setting enables generalization to unseen objects, it still requires scanning the entire object or carefully selecting reference viewpoints, which is impractical under occlusion and in robotic scenarios where the camera cannot freely move around the target.  \nThis motivates us to ask: do we truly need instance-level meshes/reference views or category labels (e.g., a category name or template)? We call a model category-agnostic if, at inference time, it estimates 9D pose for a detected object without receiving its category name, a category template, a mean shape, an instance mesh, or reference views. As discussed in Orient Anything [35], most objects have a naturally defined canonical orientation that is unique (or a finite set, for symmetric objects) and aligned with human perception. Consequently, a single observation shou","cbCaithhnsvd83lS","https://ap.wps.com/l/cbCaithhnsvd83lS","pdf",36904880,1,20,"English","en",105,"# Abstract\n# Introduction\n## Motivation: beyond benchmark overfitting\n## Category-agnostic pose estimation definition\n## Method overview and core components","[{\"question\":\"What problem does UniPose9D address in object pose estimation?\",\"answer\":\"UniPose9D targets limited generalization of existing 3D pose methods, which often overfit to benchmarks and fail on novel categories and unseen scenes.\"},{\"question\":\"What inputs and outputs does UniPose9D use to estimate pose?\",\"answer\":\"Given an instance mask/ROI, UniPose9D takes RGB-D directly or RGB with predicted depth, and outputs rotation, translation, and metric object size (full 9D pose and a metric 3D bounding box).\"},{\"question\":\"How does UniPose9D avoid reliance on CAD models, category labels, or reference views?\",\"answer\":\"It predicts NOCS coordinates by sampling point pairs from observed geometry and using DINOv2 and PointNet features, then recovers pose via a point-pair-based adaptive RANSAC N-hop Kabsch–Umeyama algorithm, without category supervision or instance meshes.\"}]",1784206223,50,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"unipose9d-universal-category-agnostic-object-pose-estimation","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/unipose9d-universal-category-agnostic-object-pose-estimation/85778/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does UniPose9D address in object pose estimation?","Question",{"text":75,"@type":76},"UniPose9D targets limited generalization of existing 3D pose methods, which often overfit to benchmarks and fail on novel categories and unseen scenes.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What inputs and outputs does UniPose9D use to estimate pose?",{"text":80,"@type":76},"Given an instance mask/ROI, UniPose9D takes RGB-D directly or RGB with predicted depth, and outputs rotation, translation, and metric object size (full 9D pose and a metric 3D bounding box).",{"name":82,"@type":73,"acceptedAnswer":83},"How does UniPose9D avoid reliance on CAD models, category labels, or reference views?",{"text":84,"@type":76},"It predicts NOCS coordinates by sampling point pairs from observed geometry and using DINOv2 and PointNet features, then recovers pose via a point-pair-based adaptive RANSAC N-hop Kabsch–Umeyama algorithm, without category supervision or instance meshes.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,114,119,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":28,"slug":113},6,"Technology","technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":21,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":21,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":106,"slug":136},19,"General","general"]