[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128683-en":3,"doc-seo-128683-105":31,"detail-sidebar-cat-0-en-105":92},{"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":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":28,"seo_description":14,"update_tm":29,"read_time":30},128683,962084928432,"Emma Wilson","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Algorithms for 3D hand pose extraction - a novel implementation","In this work a marker-less 3D hand pose extraction solution is built using only a single RGB video source. State-of-the-art computer vision models are leveraged for hand keypoint prediction, combined with a custom depth estimation algorithm to reconstruct full 3D space. The approach is packaged as a Python library supporting real-time pose estimation and is evaluated on two datasets, including a task-focused set and a common academic dataset. Results show accuracy limits with average errors above 10 cm, exceeding the few-millimeters target, while motivating future work.","UNIVERSIT`A DEGLI STUDI DI PADOVA  \nMaster’s Degree in Computer Engineering Department of Information Engineering  \nAlgorithms for 3D hand pose  \nextraction: a novel implementation  \nProfessor:  \nMenegatti Emanuele Supervisor:  \nSteiner Florian  \nStudent: Trapanotto Martino  \nM. 2044729  \nA.A. 2023/2024  \n2  \nAbstract  \nIn this work we construct a marker-less 3D hand pose extraction solution relying solely on a single RGB video source. We leverage available solutions for state-of-the-art computer vision models, coupled with a custom depth estimation algorithm to reconstruct full 3D space data. This system is packaged in a Python library capable of real time pose estimation, and is evaluated over two different datasets, one reflecting our intended task and a common academic dataset. The overall results shows limits in the accuracy of the model, measured on average above 10 cm, exceeding the desired precision of a few millimeters, but we acknowledge the relevancy of the results in the larger context of the field, and discuss potential avenues for future work.  \nContents  \nList of Figures 5  \n1 Introduction 7  \n1.1 Marker less and marker based techniques ....................... 8  \n1.2 2D, 3D and 2.5D poses ................................. 10  \n1.3 Multiview, Depth cameras and RGB cameras .................... 11  \n1.4 Hand pose specific problems .............................. 17  \n1.5 Conclusions ....................................... 18  \n2 State of the Art 21  \n2.1 InterNet ......................................... 23  \n2.2 PeCLR .......................................... 26  \n2.3 Mediapipe ........................................ 28  \n2.4 A2J-Transformer .................................... 31  \n2.5 HTT ........................................... 33  \n2.6 HDR ........................................... 35  \n2.7 Conclusions ....................................... 37  \n3 Custom solution 39  \n3.1 Depth estimation via apparent size .......................... 41  \n3.1.1 Calibration phase ................................ 44  \n3.1.2 3D pose reconstruction ............................. 45  \n3.2 Overall resulting library ................................ 46  \n3.2.1 Data export and visualization ......................... 47  \n3.3 Evaluations and Results ................................ 49  \n3.4 Conclusions ....................................... 58  \nBibliography 61  \nList of Figures  \n1.1 Examples of markers and marker based pose estimation solutions [Wang et al., 2023]  \n[Merlau et al., 2023][Wade et al., 2022] ........................ 8  \n1.2 Examples of markerless based pose estimation results [Christian Zimmermann and Brox, 2019]  \n[Bartol et al., 2020] ................................... 9  \n1.3 Examples of depth images [Tompson et al., ] [Garcia-Hernando et al., 2018] . . . . 12  \n1.4 Examples of multi view setups [Moon et al., 2020][Christian Zimmermann and Brox, 2019] 15  \n2.1 Architecture of the InterNet model [Moon et al., 2020] ............... 23  \n2.2 Examples of InterNet results [Moon et al., 2020] ................... 24  \n2.3 Example of contrastive learning: the distance represent similarity scores [Schroff et al., 2015] 26  \n2.4 Overview of contrastive learning in PeCLR ...................... 27  \n2.5 Overview of the architecture Mediapipe Hands, constructed using the Mediapipe Framework [Zhang et al., 2020] ............................ 29  \n2.6 Overview of the internal architecture of the A2J-Transformer model [Jiang et al., 2023] 32  \n2.7 Overview of the internal architecture of the HTT model [Wen et al., 2023] . . . . 34  \n2.8 Overview of the internal architecture of the HDR model [Meng et al., 2022] . . . . 35  \n3.1 Map of the keypoints generated by Mediapipe [med, 2024] ............. 41  \n3.2 Visualization of the perspective geometry effect ................... 42  \n3.3 Visualization of the 2.5D keypoints reference system ................. 43  \n3.4 Folder structure of the Python package ........................ 46  ","cbCaidBumSTDm6Qn","https://ap.wps.com/l/cbCaidBumSTDm6Qn","pdf",17806594,2,1,67,"English","en",105,"# Abstract\n# Contents\n# List of Figures\n# 1 Introduction\n## 1.1 Marker less and marker based techniques\n## 1.2 2D, 3D and 2.5D poses\n## 1.3 Multiview, Depth cameras and RGB cameras\n## 1.4 Hand pose specific problems\n## 1.5 Conclusions\n# 2 State of the Art\n## 2.1 InterNet\n## 2.2 PeCLR\n## 2.3 Mediapipe\n## 2.4 A2J-Transformer\n## 2.5 HTT\n## 2.6 HDR\n## 2.7 Conclusions\n# 3 Custom solution\n## 3.1 Depth estimation via apparent size\n## 3.2 Overall resulting library\n## 3.3 Evaluations and Results\n## 3.4 Conclusions\n# Bibliography","[{\"question\":\"What data source does the proposed 3D hand pose extraction method use?\",\"answer\":\"The method relies solely on a single RGB video source, without markers.\"},{\"question\":\"How is full 3D hand space reconstructed in the system?\",\"answer\":\"State-of-the-art vision models provide keypoint information, while a custom depth estimation algorithm reconstructs 3D space.\"},{\"question\":\"How is the approach evaluated and what accuracy is achieved?\",\"answer\":\"The system is evaluated on two datasets (task-oriented and a common academic dataset). Average errors are reported above 10 cm, which is beyond the intended few-millimeters precision.\"}]","Algorithms for 3D hand pose extraction - a novel implementation | PDF",1786002632,169,{"code":4,"msg":32,"data":33},"ok",{"site_id":25,"language":24,"slug":34,"title":13,"keywords":35,"description":14,"schema_data":36,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":29},"algorithms-for-3d-hand-pose-extraction-a-novel-implementation","",{"@graph":37,"@context":86},[38,54,69],{"@type":39,"itemListElement":40},"BreadcrumbList",[41,45,48,51],{"item":42,"name":43,"@type":44,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":46,"name":47,"@type":44,"position":20},"https://docshare.wps.com/document/","Document",{"item":49,"name":12,"@type":44,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":44,"position":53},"https://docshare.wps.com/document/algorithms-for-3d-hand-pose-extraction-a-novel-implementation/128683/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":42,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-25","2026-08-06",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What data source does the proposed 3D hand pose extraction method use?","Question",{"text":76,"@type":77},"The method relies solely on a single RGB video source, without markers.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How is full 3D hand space reconstructed in the system?",{"text":81,"@type":77},"State-of-the-art vision models provide keypoint information, while a custom depth estimation algorithm reconstructs 3D space.",{"name":83,"@type":74,"acceptedAnswer":84},"How is the approach evaluated and what accuracy is achieved?",{"text":85,"@type":77},"The system is evaluated on two datasets (task-oriented and a common academic dataset). Average errors are reported above 10 cm, which is beyond the intended few-millimeters precision.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":21,"doc_module":4,"doc_module_name":47,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":47,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":47,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":47,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":47,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":47,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":47,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":47,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":47,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":47,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":47,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]