[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83074-en":3,"doc-seo-83074-105":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},83074,1374391975076,"Riley","https://ap-avatar.wpscdn.com/avatar/14000253ca4ec9f6853?x-image-process=image/resize,m_fixed,w_180,h_180&k=1783305029341752051",8,"Research & Report","Visual graphs for image classification: does the structure affect performance","Deep learning models deliver strong results in visual tasks but often fail to fully encode intrinsic visual structure, overlooking spatial, topological, and semantic cues present in images. Graph neural networks provide a framework to represent such structure, yet visual graph construction for these models remains only partially understood. This study addresses the gap by systematically comparing current graph construction techniques under a fixed three-layer GCN architecture. Empirical results show how the network structure impacts performance and emphasize that earlier computational stages, especially graph construction, strongly shape downstream effectiveness.","arXiv :2607 .06295v 1 [ cs .CV] 7 Jul 2026  \nVisual graphs for image classification: does the structure affect performance?  \nAlessandra Ibba 1  \nUniversity of Sassari-Department of Engineering, Sassari, Italy {[a.ibba26}@studenti.uniss.it](a.ibba26}@studenti.uniss.it)  \nAbstract. Deep learning models have emerged in machine learning and related fields, demonstrating astonishing performance in various visual tasks. Despite their great success, however, these models are unable to fully encode intrinsic visual structures, and often ignore the spatial, topological, and semantic information contained within an image. Graph neural networks offer a good framework to face this aspect, but their effective use for visual tasks has been only partly explored and mainly starting from a limited perspective. This work aims to address this gap by conducting a systematic comparison of current graph construction techniques within the context of a fixed three-layer GCN architecture.  \nThrough an empirical study, it demonstrates in particular how the network structure affects performance and provides an important methodological contribution regarding the computational stages preceding graph utilization, which will be strongly influenced by the structure itself.  \nKeywords: Graph Neural Networks · Computer Vision · Image Classification · Graph Construction · Graph Topology · Graph Sparsification  \n1 Introduction  \nIn recent years, both computer vision and natural language processing have greatly benefited from the success of convolutional neural networks (CNNs) [25] . A key factor contributing to this success is the ability of CNNs to exploit the inherent structural properties of specific data types. In the case of images, their grid-like organization allows convolutional layers to extract high-level features and build hierarchical representations, thereby providing substantial expressive power. Despite these advantages, CNNs often struggle to fully capture the relationships among objects within a scene and to effectively incorporate contextual information, which can be crucial in many applications. Moreover, CNN-based solutions are typically characterized by a large number of parameters and high computational complexity, regardless of the actual image content. Vision Transformers (ViTs) [5] partially address these issues by implementing self-attention mechanisms to directly model the relationships between pairs of image patches. However, they fail to fully preserve spatial layout; furthermore, they are unable to handle fine details and precise localization due to their very nature, which relies on patches of significant size [7] . In addition, like CNNs, ViTs are  \n2 A. Ibba  \nparameter-heavy and computationally demanding. Graph neural networks offer a compelling framework to address these limitations. By operating on non-regular graph structures, they are able to model structural relationships within the data and reveal intrinsic connections among different parts of an image. However, theirregular nature of graphs also poses challenges, as convolution and filtering operations on graphs are not as straightforward or well-defined as they are in the image domain [23] . Graph Convolutional Networks (GCN), introduced by [11], are an efficient variant of convolutional neural networks for graphs. The input data of a GNN is a graph, that is a graph-structured representation of visual data in which nodes correspond to visual entities and edges encode relationships between them, such as spatial adjacency, geometric proximity, semantic similarity, or contextual interaction. Note that the set of nodes typically carries visual features extracted from images but these features can vary greatly from simple pixels to regions or other structured representations (sometimes denoted as descriptors) of image areas. Information carried by nodes and edges commonly takes the name of “embedding”; from an intuitive perspective, a node embedding answers to the basic ques","cbCaidxTTN3Q9Pei","https://ap.wps.com/l/cbCaidxTTN3Q9Pei","pdf",1163580,1,14,"English","en",105,"# Introduction\n# State of the art\n## Nodes definition","[{\"question\":\"What problem does the paper address in visual image classification models?\",\"answer\":\"It targets the limitation that common deep learning approaches do not fully encode intrinsic visual structure, such as spatial, topological, and semantic information in images.\"},{\"question\":\"How does the study evaluate the impact of graph design?\",\"answer\":\"It performs an empirical, systematic comparison of graph construction techniques while keeping a fixed three-layer GCN architecture constant across experiments.\"},{\"question\":\"Which aspect is emphasized as having strong influence on final performance?\",\"answer\":\"The paper highlights the computational stages before graph utilization—particularly how the network structure and graph construction choices affect downstream performance.\"}]",1784185031,35,{"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},"visual-graphs-for-image-classification-does-the-structure-affect-performance","",{"@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/visual-graphs-for-image-classification-does-the-structure-affect-performance/83074/",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 the paper address in visual image classification models?","Question",{"text":75,"@type":76},"It targets the limitation that common deep learning approaches do not fully encode intrinsic visual structure, such as spatial, topological, and semantic information in images.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the study evaluate the impact of graph design?",{"text":80,"@type":76},"It performs an empirical, systematic comparison of graph construction techniques while keeping a fixed three-layer GCN architecture constant across experiments.",{"name":82,"@type":73,"acceptedAnswer":83},"Which aspect is emphasized as having strong influence on final performance?",{"text":84,"@type":76},"The paper highlights the computational stages before graph utilization—particularly how the network structure and graph construction choices affect downstream performance.","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},[93,97,101,105,110,115,120,123,128,131,135],{"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":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]