[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-292336-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-292336-en":131},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","video-efficient-foundation-models-academic-phd-thesis","Video-Efficient Foundation Models - Academic PhD Thesis","","Video-Efficient Foundation Models presents a doctoral research thesis on learning representations for video and action understanding with an emphasis on efficiency and transfer/generalization. The work develops methods such as feature-supervised action modality transfer, skeleton-contrastive 3D action representation learning, and tubelet-contrastive self-supervision for video-efficient generalization. It also studies how benchmark sensitivity affects video self-supervised learning, including sensitivity factors related to downstream domains, samples, actions, and tasks. Experiments cover multiple dataset settings, ablations, and comparisons to state-of-the-art.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/video-efficient-foundation-models-academic-phd-thesis/292336/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/video-efficient-foundation-models-academic-phd-thesis/292336.png","ImageObject",300,407,{"name":42,"@type":43},"Sophia Brooks","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-18","2026-09-17",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",12,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What is the main theme of the thesis?","Question",{"text":63,"@type":64},"The thesis focuses on video-efficient foundation models and learning effective representations for video and action understanding, with attention to efficiency, transfer, and generalization.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"Which methods are introduced in the thesis?",{"text":68,"@type":64},"It presents feature-supervised action modality transfer, skeleton-contrastive 3D action representation learning, tubelet-contrastive self-supervision for video-efficient generalization, and an analysis of benchmark sensitivity in video self-supervised learning.",{"name":70,"@type":61,"acceptedAnswer":71},"How does the thesis evaluate its ideas?",{"text":72,"@type":64},"Evaluation includes experimental setups with target and source datasets, multiple ablation studies, generalization tests across domains and modalities, and comparisons against state-of-the-art approaches.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},292336,1789650216,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,120,123,127],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":25,"category_name":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":25,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":112,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":113,"slug":114},8,30,"research-report",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":25,"category_name":129,"show_sort_weight":97,"slug":130},19,"General","general",{"code":4,"msg":82,"data":132},{"doc_id":79,"user_id":133,"nickname":42,"user_avatar":134,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":140,"language":141,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":12,"update_tm":80,"read_time":145,"preview_page_count":4},962084925636,"https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d","V ideo-Efficient Foundation Models Fida Mohammad Thoker  \nVideo-Efficient Foundation Models  \nVideo-Efficient Foundation Models  \nFida Mohammad Thoker  \nThis book was typeset by the author using LATEX2\\# .  \nThe cover was generated by Adobe-FireFly (Model Firefly Image 2 beta) with the following prompts:  \n• Front: “A 3D render of a glowing sphere inside a data center on a black background and with a prominent artificial deep neural network embedded in the sphere. Concepts: Futuristic, Color and Tone: Monochromatic, Themes: Digital art”  \n• Back: “Bottom view of GPU clusters in a single-color. Style: Image generated from the above front prompt, Concepts: Futuristic, Color and Tone: Monochromatic, Themes: Digital art”  \nCopyright © 2023 by Fida Mohammad Thoker.  \nAll rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission from the author.  \nVideo-Efficient Foundation Models  \nACADEMISCH PROEFSCHRIFT  \nter verkrijging van de graad van doctor aan de Universiteit van Amsterdam  \nop gezag van de Rector Magnificus [prof. dr. ir. P.P.C.C. Verbeek](prof. dr. ir. P.P.C.C. Verbeek)  \nten overstaan van een door het College voor Promoties ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel op vrijdag 8 december 2023, te 10:00 uur  \ndoor Fida Mohammad Thoker  \ngeboren te Tral, Jammu Kashmir  \nii  \nPromotiecommissie  \nPromotor: prof. dr. C. G. M. Snoek  \nCopromotor: dr. H.R. Doughty  \nOverige leden: dr. Y. M. Asano dr. E. Gavves  \ndr. P.S.M. Mettes [prof. dr. ir. H.E. Bal](prof. dr. ir. H.E. Bal)[ ](prof. dr. ir. H.E. Bal)[prof. dr](prof. dr). J. Gall  \nUniversiteit van Amsterdam  \nUniversiteit van Amsterdam  \nUniversiteit van Amsterdam  \nUniversiteit van Amsterdam  \nUniversiteit van Amsterdam Vrije Universiteit Amsterdam University of Bonn  \nFaculteit der Natuurwetenschappen, Wiskunde en Informatica  \nThe work described in this thesis has been carried out within the graduate school ASCI, at the Video & Image Sense Lab.  \niii  \nTo my cherished son, Ibaad, may this thesis stand as a testament to the power of perseverance and the pursuit of knowledge, inspiring you to reach for the stars in your  \nown journey.  \nv  \nContents  \n1 Introduction 1  \n1.1 Problem Statement .......................... 3  \n1.2 Research Questions .......................... 5  \n1.3 Publications, Co-authorship, and Roles ................ 9  \n2 Feature-Supervised Action Modality Transfer 11  \n2.1 Introduction .............................. 11  \n2.2 Related Work ............................. 12  \n2.2.1 Modalities for Action Recognition ............. 12  \n2.2.2 Knowledge Transfer ..................... 12  \n2.3 Proposed Method ........................... 14  \n2.3.1 Cross-Modal Teacher-Student ................ 14  \n2.3.2 Feature-Supervised Granularity ............... 15  \n2.3.2.1 Clip-to-Clip .................... 15  \n2.3.2.2 Video-to-Clip ................... 16  \n2.3.2.3 Video-to-Video .................. 17  \n2.4 Experimental Setup .......................... 17  \n2.4.1 Target Datasets ........................ 17  \n2.4.2 Source Datasets ........................ 17  \n2.4.3 Implementation Details .................... 19  \n2.5 Results ................................. 19  \n2.5.1 Ablation ............................ 19  \n2.5.1.1 Which source modality? .............. 20  \n2.5.1.2 Which loss function? ............... 21  \n2.5.1.3 Which granularity? ................ 21  \n2.5.1.4 Which layers to transfer? ............. 21  \n2.5.2 Domain Generalization .................... 22  \n2.5.3 Modality Generalization ................... 22  \n2.5.4 Task Generalization ...................... 23  \n2.5.5 Comparison with the state-of-the-art ............. 24  \n2.6 Conclusion .............................. 26  \n3 Skeleton-Contrastive 3D Action Representation Learning 27  \n3.1 Introduction ......................","cbCaibF0BuHMeXq7","https://ap.wps.com/l/cbCaibF0BuHMeXq7","pdf",42615541,147,"English","# 1 Introduction\n## 1.1 Problem Statement\n## 1.2 Research Questions\n## 1.3 Publications, Co-authorship, and Roles\n# 2 Feature-Supervised Action Modality Transfer\n## 2.1 Introduction\n## 2.2 Related Work\n## 2.3 Proposed Method\n## 2.4 Experimental Setup\n## 2.5 Results\n## 2.6 Conclusion\n# 3 Skeleton-Contrastive 3D Action Representation Learning\n## 3.1 Introduction\n## 3.2 Related Work\n## 3.3 Skeleton-Contrastive Learning\n## 3.4 Experiments\n## 3.5 Conclusion\n# 4 How Severe is Benchmark-Sensitivity in Video Self-Supervised Learning?\n## 4.1 Introduction\n## 4.2 Identifying Benchmark Sensitivity\n## 4.7 SEVERE-benchmark\n## 4.8 Observations, Limitations and Recommendations\n# 5 Tubelet-Contrastive Self-Supervision for Video-Efficient Generalization","[{\"question\":\"What is the main theme of the thesis?\",\"answer\":\"The thesis focuses on video-efficient foundation models and learning effective representations for video and action understanding, with attention to efficiency, transfer, and generalization.\"},{\"question\":\"Which methods are introduced in the thesis?\",\"answer\":\"It presents feature-supervised action modality transfer, skeleton-contrastive 3D action representation learning, tubelet-contrastive self-supervision for video-efficient generalization, and an analysis of benchmark sensitivity in video self-supervised learning.\"},{\"question\":\"How does the thesis evaluate its ideas?\",\"answer\":\"Evaluation includes experimental setups with target and source datasets, multiple ablation studies, generalization tests across domains and modalities, and comparisons against state-of-the-art approaches.\"}]","Video-Efficient Foundation Models - Academic PhD Thesis | PDF",370]