[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-156999-105":3,"detail-sidebar-cat-0-en-105":73,"doc-detail-156999-en":123},{"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":66,"head_meta":68,"extra_data":70,"updated_unix":72},105,"en","hypnos-highly-precise-foreground-focused-diffusion-finetuning-for-inanimate-objects","HYPNOS - Highly Precise Foreground-focused Diffusion Finetuning for Inanimate Objects","","Personalized diffusion-based text-to-image generation relies on near-perfect reconstruction under limited reference samples, yet common finetuning approaches struggle to preserve foreground object consistency while encouraging diverse, controllable backgrounds. Overfitting further reduces foreground control when prompt information is ambiguously applied to both foreground and background regions. Hypnos addresses these issues through content-centric prompting and an added foreground-focused discriminative module integrated into the diffusion model and trained with targeted supervision. Experiments show more robust results and better foreground-background disentanglement.",{"@graph":14,"@context":65},[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/hypnos-highly-precise-foreground-focused-diffusion-finetuning-for-inanimate-objects/156999/",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/hypnos-highly-precise-foreground-focused-diffusion-finetuning-for-inanimate-objects/156999.png","ImageObject",300,407,{"name":42,"@type":43},"Ezra","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-28","2026-08-29",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",12,{"@type":57,"mainEntity":58},"FAQPage",[59],{"name":60,"@type":61,"acceptedAnswer":62},"What results does the paper report for HYPNOS compared to former techniques?","Question",{"text":63,"@type":64},"Experimental results indicate more robust performance and visually pleasing images, with clearer disentanglement effects between subject (foreground) and scene (background).","Answer","https://schema.org",{"og:url":32,"og:type":67,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":69,"canonical":32},"index,follow",{"doc_id":71,"site_id":7},156999,1787970983,{"code":4,"msg":74,"data":75},"success",[76,80,84,88,93,98,103,107,112,115,119],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":77,"show_sort_weight":78,"slug":79},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":81,"show_sort_weight":82,"slug":83},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Exam",70,"exam",{"id":89,"doc_module":4,"doc_module_name":25,"category_name":90,"show_sort_weight":91,"slug":92},5,"Comic",60,"comic",{"id":94,"doc_module":4,"doc_module_name":25,"category_name":95,"show_sort_weight":96,"slug":97},6,"Technology",50,"technology",{"id":99,"doc_module":4,"doc_module_name":25,"category_name":100,"show_sort_weight":101,"slug":102},7,"Healthcare",40,"healthcare",{"id":104,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":105,"slug":106},8,30,"research-report",{"id":108,"doc_module":4,"doc_module_name":25,"category_name":109,"show_sort_weight":110,"slug":111},9,"Religion & Spirituality",20,"religion-spirituality",{"id":110,"doc_module":4,"doc_module_name":25,"category_name":113,"show_sort_weight":110,"slug":114},"World Cup","world-cup",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":116,"slug":118},10,"Lifestyle","lifestyle",{"id":120,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":89,"slug":122},19,"General","general",{"code":4,"msg":74,"data":124},{"doc_id":71,"user_id":125,"nickname":42,"user_avatar":126,"doc_module":4,"category_id":104,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":127,"file_id":128,"file_url":129,"file_type":130,"file_size":131,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":132,"language":133,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":134,"faqs":135,"seo_title":136,"seo_description":12,"update_tm":72,"read_time":101,"preview_page_count":4},1099514068035,"https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c","This ACCV 2024 workshop paper, provided here by the Computer Vision Foundation, is the author-created version. The content of this paper is identical to the content of the officially published ACCV 2024 LNCS version of the paper as available on SpringerLink: [https://link.springer.com/conference/accv](https://link.springer.com/conference/accv)  \nHYPNOS : Highly Precise Foreground-focused Diffusion Finetuning for Inanimate Objects  \nOliverio Theophilus Nathanael 1, Jonathan Samuel Lumentut1, Nicholas Hans Muliawan1, Edbert Valencio Angky1, Felix Indra Kurniadi2, Alfi Yusrotis Zakiyyah 1, and Jeklin Harefa1  \n1 Universitas Bina Nusantara, Jakarta, Indonesia  \n2 Università di Pisa, Pisa, Italy  \nAbstract. In recent years, personalized diffusion-based text-to-image  \ngenerative tasks have been a hot topic in computer vision studies. A ro  \nbust diffusion model is determined by its ability to perform near-perfect  \nreconstruction of certain product outcomes given few related input sam  \nples. Unfortunately, the current prominent diffusion-based finetuning  \ntechnique falls short in maintaining the foreground object consistency  \nwhile being constrained to produce diverse backgrounds in the image  \noutcome. In the worst scenario, the overfitting issue may occur, meaning  \nthat the foreground object is less controllable due to the condition above,  \nfor example, the input prompt information is transferred ambiguously to  \nboth foreground and background regions, instead of the supposed back  \nground region only. To tackle the issues above, we proposed Hypnos,  \na highly precise foreground-focused diffusion finetuning technique. On  \nthe image level, this strategy works best for inanimate object generation  \ntasks, and to do so, Hypnos implements two main approaches, namely:  \n(i) a content-centric prompting strategy and (ii) the utilization of our ad  \nditional foreground-focused discriminative module. The utilized module  \nis connected with the diffusion model and finetuned with our proposed  \nset of supervision mechanism. Combining the strategies above yielded  \nto the foreground-background disentanglement capability of the diffu  \nsion model. Our experimental results showed that the proposed strategy  \ngave a more robust performance and visually pleasing results compared  \nto the former technique. For better elaborations, we also provided ex  \ntensive studies to assess the fruitful outcomes above, which reveal how  \npersonalization behaves in regard to several training conditions.  \nKeywords: Generative model · Stable Diffusion · Dreambooth  \n1 Introduction  \nPersonalization for text-to-image (T2I) Diffusion Model has been a hot field of study and rapidly popularized among researchers in computer vision studies as well as practitioners and even hobbyist. Personalization is the key for generative  \n2 O. T. Nathanael et al.  \n(a) Entangled Foreground-Background (b) Hypnos generated images  \nFig. 1: The current (a) widely used method [21] is prone to subject-scene entanglement and overfitting as shown by the yellow lines. Our proposed method visually (b) shows a realistic disentanglement effect between subject and scene at the foreground and background, respectively.  \nmodels to be widely used as a tool for various use cases. This personalization in the Diffusion Model has the same spirit as grounding Large Language Models (LLMs), where it opens the possibility of being reliably incorporated in diverse real-world environments. With reliable personalization in the Diffusion Model, it opens several potentials for recreational purposes and even in the business world, such as creating product advertisements and campaign photos, which mostly highlight the foreground objects (foreground-focused objective) .  \nOne particular distinctive family of these techniques, namely Dreambooth [21], enables the ability to finetune Diffusion models on a specific image without losing prior knowledge. At the moment Dreambooth is one of the most popular personalization technique","cbCaid63JSh0q98X","https://ap.wps.com/l/cbCaid63JSh0q98X","pdf",2669480,16,"English","# Introduction\n## Personalization for text-to-image diffusion models\n## Dreambooth and observed limitations\n## Proposed HYPNOS approach","[{\"question\":\"What results does the paper report for HYPNOS compared to former techniques?\",\"answer\":\"Experimental results indicate more robust performance and visually pleasing images, with clearer disentanglement effects between subject (foreground) and scene (background).\"}]","HYPNOS - Highly Precise Foreground-focused Diffusion Finetuning for Inanimate Objects | PDF"]