[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-124635-en":3,"doc-seo-124635-105":30,"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":4,"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},124635,549758146520,"Patrick","https://ap-avatar.wpscdn.com/avatar/80002397d8c0411e94?_k=1775819394049821470",8,"Research & Report","Vision Based Machine Learning Algorithms for Out-of-Distribution Generalisation","Machine learning drives many computer vision tasks such as segmentation, classification, detection, and reconstruction, yet models usually work well only in the training domain. When a classifier trained on one domain is evaluated on an unseen out-of-distribution (OOD) domain, performance can drop significantly, creating challenges for domain generalisation, domain adaptation, and domain shifting. The study compares vision-based domain-specific and domain-generalised approaches using PACS and OfficeHome and introduces an implementation pipeline, showing that simple CNN-based deep learning models generalise poorly under domain shifting.","Vision Based Machine Learning Algorithms for Out-of-Distribution Generalisation  \nHamza Riaz 1 and Alan F. Smeaton2  \n1 School of Computing  \n2 Insight Centre for Data Analytics, Dublin City University, Glasnevin, Dublin 9, Ireland.  \n[hamza.riaz2@mail.dcu.ie](hamza.riaz2@mail.dcu.ie)  \nAbstract. There are many computer vision applications including object segmentation, classification, object detection, and reconstruction for which machine learning (ML) shows state-of-the-art performance. Nowadays, we can build ML tools for such applications with real-world accuracy. However, each tool works well within the domain in which it has been trained and developed. Often, when we train a model on a dataset in one specific domain and test on another unseen domain known as an out of distribution (OOD) dataset, models or ML tools show a decrease in performance. For instance, when we train a simple classifier on real-world images and apply that model on the same classes but with a different domain like cartoons, paintings or sketches then the performance of ML tools disappoints. This presents serious challenges of domain generalisation (DG), domain adaptation (DA), and domain shifting. To enhance the power of ML tools, we can rebuild and retrain models from scratch or we can perform transfer learning. In this paper, we present a comparison study between vision-based technologies for domain-specific and domain-generalised methods. In this research we highlight that simple convolutional neural network (CNN) based deep learning methods perform poorly when they have to tackle domain shifting. Experiments are conducted on two popular vision-based benchmarks, PACS and OfficeHome. We introduce an implementation pipeline for domain generalisation methods and conventional deep learning models. The outcome confirms that CNN-based deep learning models show poor generalisation compare to other extensive methods.  \nKeywords: Vision Machine Learning, Domain Generalisation, Domain Adaptation, Domain Shifting, Domain Specific Learning  \n1 Introduction  \nThe field of machine learning (ML) has created tremendous success stories by solving many complex problems like object classification, detection, segmentation and reconstruction in videos, natural language processing (NLP), medical image analysis, robotics, and many more. These developments in ML algorithmsand databases, and the fusion of various fields of ML help researchers achieve  \n2 Hamza Riaz and Alan F. Smeaton  \nhigh-level goals. The majority of current applications are built on what we call traditional ML where we usually have millions of example datapoints with labels to train a model under supervised learning (SL) .  \nSince 2011, with the help of deep learning (DL) which is a sub-domain of ML that deals with various datasets to automatically extract features, scientists are now using DL to handle various supervised and unsupervised learning problems as described in [1] . Pure DL-based models are static systems, and have many problems including overfitting, they commonly need huge datasets, have data biases, and they do not have significant potential for generalisation and domain adaptation [2] .  \nPrevious works illustrate that the majority of the time ML tools fail to generalise when processing out of distribution (OOD) data. The main reasons for wanting to design and analyse such generalised tools are applications like vision based autonomous systems, and medical imaging [2] . For example, when only a few conditions change during an inference process in image processing such as light variations, shapes, locations, or the pose of objects, then models perform poorly because they did not have interaction with similar variations during their training and thus did not learn how to perform under such unpredictable circumstances [3–5] . The work in [6] conveys information about the collapse of ML tools for generalisation of OOD data which actually happens when ML models learn fake correlations inste","cbCaiibNGeAP2pei","https://ap.wps.com/l/cbCaiibNGeAP2pei","pdf",2141377,1,16,"English","en",105,"# Introduction\n## Domain shift and OOD generalisation\n## Domain generalisation vs domain adaptation","[{\"question\":\"Why do models lose performance on out-of-distribution (OOD) data in vision tasks?\",\"answer\":\"Because the model is trained to fit one domain and then tested on a different unseen domain, it fails to handle variations it did not encounter during training, leading to domain shift and reduced generalisation.\"},{\"question\":\"What is the key difference between domain generalisation (DG) and domain adaptation (DA)?\",\"answer\":\"DG learns from multiple source domains without access to target-domain data during training, while DA can use target-domain data to maximise performance on that target domain.\"},{\"question\":\"Which benchmarks and model types are used to evaluate generalisation in the study?\",\"answer\":\"Experiments are conducted on PACS and OfficeHome, and the comparison includes domain generalisation methods and conventional deep learning models, with results indicating weaker performance for simple CNN-based approaches under domain shifting.\"}]","Vision Based Machine Learning Algorithms for Out-of-Distribution Generalisation | PDF",1785893441,40,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"vision-based-machine-learning-algorithms-for-out-of-distribution-generalisation","",{"@graph":36,"@context":85},[37,54,68],{"@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/vision-based-machine-learning-algorithms-for-out-of-distribution-generalisation/124635/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why do models lose performance on out-of-distribution (OOD) data in vision tasks?","Question",{"text":75,"@type":76},"Because the model is trained to fit one domain and then tested on a different unseen domain, it fails to handle variations it did not encounter during training, leading to domain shift and reduced generalisation.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the key difference between domain generalisation (DG) and domain adaptation (DA)?",{"text":80,"@type":76},"DG learns from multiple source domains without access to target-domain data during training, while DA can use target-domain data to maximise performance on that target domain.",{"name":82,"@type":73,"acceptedAnswer":83},"Which benchmarks and model types are used to evaluate generalisation in the study?",{"text":84,"@type":76},"Experiments are conducted on PACS and OfficeHome, and the comparison includes domain generalisation methods and conventional deep learning models, with results indicating weaker performance for simple CNN-based approaches under domain shifting.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"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":29,"slug":118},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]