[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128489-en":3,"doc-seo-128489-105":30,"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":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":27,"seo_description":14,"update_tm":28,"read_time":29},128489,13056712833777,"Logic","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","Semantic Guided Multi-Future Human Motion Prediction - Master’s Thesis","This master’s thesis enhances a machine learning model for predicting multiple potential future human movements by embedding semantic information into the input pipeline. The work develops semantic class labeling strategies and evaluates the approach through comprehensive performance metrics. Comparative experiments contrast models using only kinematic data with models augmented by semantics in a realistic scenario where about 40% of data lacks semantic inputs, showing consistent error reductions. Model design is analyzed to ensure sufficient parameter capacity, supporting long-horizon prediction for human-robot cooperation in Industry 5.0.","UNIVERSITY OF PADUA  \nDEPARTMENT OF ENGINEERING AND INDUSTRIAL SYSTEMS MANAGEMENT  \nMASTER’S DEGREE IN MECHATRONIC ENGINEERING  \nMASTER’S THESIS  \nSEMANTIC GUIDED MULTI-FUTURE HUMAN MOTION PREDICTION  \nSupervisor: Stefano Michieletto  \nCo-supervisor: Michael Vanuzzo  \nCo-supervisor: Mattia Guidolin  \nStudent: Francesco Borsatti 2007675-IMC  \nACADEMIC YEAR: 2022-23  \nThe primary objective of this thesis is to enhance the accuracy of a machine learning model for predicting multiple potential future human movements. This has been achieved by incorporating semantic information into the input data, contributing to the broader goal of advancing safer and efficient human-robot cooperation within an Industry 5.0 context.  \nWe analyzed diverse performance metrics and explored multiple aspects of the problem, from the integration of semantic data into the preprocessing phase, to the development of semantic class labeling strategies and comprehensive evaluation methodologies.  \nWe demonstrated the significance of semantic context in motion prediction through a comparative analysis of models utilizing kinematic data alone and those augmented with semantic information. We conducted experiments on one of the most recent and significant datasets in the literature, simulating a realistic scenario where approximately 40% of the data lacked semantic information. Remarkably, even in this setting, the model with the most parameter capacity enhanced by semantic information (128 kin+sem) outperformed both the kinematic-only counterpart and the zero-velocity baseline model.  \nIn particular, \"128 kin+sem\" reduced by 3% the cumulative error over the \"128 kin\" and exhibited a 10% error reduction against the zero-velocity over one second of prediction timespan.  \nThe importance of a careful model design is highlighted, by showing why ensuring sufficient parameter capacity is necessary to effectively accommodate the augmented input data when semantic information is introduced.  \nRegarding the practical applications of our model, it is important to consider that for cooperative robotic planning, the initial moments of motion prediction hold relatively less significance. The primary focus lies in achieving accurate predictions for a range of potential outcomes in long-term motion prediction.  \nWhile our research has primarily centered around cooperative robotic applications, we also expect that our methodologies can be applied to diverse fields beyond the initial scope. The prediction of future body movements opens up possibilities for offline utilization in non-realtime tasks, such as generating realistic human motion. In particular, motion prediction models hold potential in generating partial movements, allowing us to leverage a limited amount of available data to generate new data points.  \nIn terms of performance evaluation, we measured the time it took to compute the inference of predictions using the metrics script. The  \ntested models exhibited slightly longer inference times compared to the duration of the predicted sequence, since the models were designed with offline testing in mind, but can be optimized for real-time with software and hardware adaptations.  \nThe findings of this study lay the foundation for future research endeavors, as numerous deep learning models that solely rely on kinematic information could potentially achieve groundbreaking results by effectively incorporating semantic information. This study represents an initial step in showcasing the influential role of semantics in enhancing the prediction of human motion.  \nThe problem with quotes found on the internet is that they are often not true.  \n—Abraham Lincoln  \nACKNOWLEDGMENTS  \nI would like to extend my heartfelt gratitude to the individuals who have played a significant role in my thesis journey. Their support, guidance, and contributions have been invaluable, and I am truly grateful.  \nFirst, I would like to express my deepest appreciation to Stefano Michieletto for pr","cbCairXB87MlOtKz","https://ap.wps.com/l/cbCairXB87MlOtKz","pdf",1297791,1,70,"English","en",105,"# Introduction\n## Problem Description\n### Collaborative robotics\n### Anticipation of human motion\n## Methods\n### Semantics in the context of human movement\n### Machine Learning, Artificial Intelligence\n## State of the Art\n### Review of existing methods for human motion prediction\n### Human motion datasets\n# Description of multiverse and multipose models\n## Fundamentals\n### Sequence-to-Sequence neural networks\n### Search algorithms\n### Graph attention networks\n## Original Multiverse model\n### Overview of the model\n### Architecture\n### Description of loss function\n## MultiPose: human motion prediction\n### Human motion prediction\n### Dimensionality reduction\n### Range Saturation and Prediction Combination\n## Modifications to the MultiPose model\n### Preprocessing\n### Training\n### Post-processing","[{\"question\":\"How does the thesis incorporate semantic information into human motion prediction?\",\"answer\":\"Semantic information is integrated into the preprocessing/input pipeline, together with semantic class labeling strategies, to guide prediction beyond kinematic-only cues.\"},{\"question\":\"What evidence demonstrates the benefit of semantics compared with kinematics only?\",\"answer\":\"Experiments compare kinematic-only models against semantic-augmented models, including a scenario where roughly 40% of data lacks semantic information, and show semantic models reduce cumulative and one-second prediction errors.\"},{\"question\":\"Why is sufficient parameter capacity important when adding semantic inputs?\",\"answer\":\"The thesis highlights that model design must provide enough capacity to accommodate the augmented input representation; otherwise the semantic benefit cannot be fully leveraged.\"}]","Semantic Guided Multi-Future Human Motion Prediction - Master’s Thesis | PDF",1786001356,176,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"semantic-guided-multi-future-human-motion-prediction-masters-thesis","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/semantic-guided-multi-future-human-motion-prediction-masters-thesis/128489/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-24","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},"How does the thesis incorporate semantic information into human motion prediction?","Question",{"text":76,"@type":77},"Semantic information is integrated into the preprocessing/input pipeline, together with semantic class labeling strategies, to guide prediction beyond kinematic-only cues.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What evidence demonstrates the benefit of semantics compared with kinematics only?",{"text":81,"@type":77},"Experiments compare kinematic-only models against semantic-augmented models, including a scenario where roughly 40% of data lacks semantic information, and show semantic models reduce cumulative and one-second prediction errors.",{"name":83,"@type":74,"acceptedAnswer":84},"Why is sufficient parameter capacity important when adding semantic inputs?",{"text":85,"@type":77},"The thesis highlights that model design must provide enough capacity to accommodate the augmented input representation; otherwise the semantic benefit cannot be fully leveraged.","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":24},{"code":4,"msg":5,"data":93},[94,98,102,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":21,"slug":104},"Exam","exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]