[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-116908-en":3,"doc-seo-116908-105":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},116908,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1785132997149421697",8,"Research & Report","Data Models for Dataset Drift Controls in Machine Learning With Optical Images","Camera images are ubiquitous in machine learning research and support critical public services such as medicine and environmental surveying. Robustness limits have restricted reliable deployment because performance can drop when training and deployment data differ. Existing validation methods address drift but do not model the primary object of interest explicitly: the data. This work builds differentiable data models by combining machine learning with physical optics, enabling drift synthesis, gradient-based tolerancing, drift forensics, and drift optimization to improve data generation and downstream machine vision performance. Two public datasets support the study.","Data Models for Dataset Drift Controls in Machine Learning With Optical Images  \nLuis Oala ∗  \nFraunhofer HHI and Dotphoton AG  \nMarco Aversa∗  \nDotphoton AG and University of Glasgow  \nGabriel Nobis  \nFraunhofer HHI  \nKurt Willis  \nFraunhofer HHI  \nYoan Neuenschwander  \nHEPIA/HES-SO  \nMichèle Buck  \nKlinikum rechts der Isar  \nChristian Matek  \nHelmholtz Zentrum Munich  \nJérôme Extermann  \nHEPIA/HES-SO  \nEnrico Pomarico  \nHEPIA/HES-SO  \nWojciech Samek  \nFraunhofer HHI  \nRoderick Murray-Smith  \nUniversity of Glasgow  \nChristoph Clausen  \nDotphoton AG  \nBruno Sanguinetti  \n[luis. oala@dotphoton. com](luis. oala@dotphoton. com)  \n[marco.aversa@dotphoton. com](marco.aversa@dotphoton. com)  \n[gabriel.nobis@hhi.fraunhofer. de](gabriel.nobis@hhi.fraunhofer. de)  \n[kurt.willis@hhi.fraunhofer. de](kurt.willis@hhi.fraunhofer. de)  \n[yoan. neuenschwander@hesge. ch](yoan. neuenschwander@hesge. ch)  \n[michele.kyncl@tum. de](michele.kyncl@tum. de)  \n[christian. matek@helmholtz-muenchen. de](christian. matek@helmholtz-muenchen. de)  \n[jerome. extermann@hesge. ch](jerome. extermann@hesge. ch)  \n[enrico.pomarico@hesge. ch](enrico.pomarico@hesge. ch)  \n[wojciech.samek@hhi.fraunhofer. de](wojciech.samek@hhi.fraunhofer. de)  \n[roderick. murray-smith@glasgow. ac.uk](roderick. murray-smith@glasgow. ac.uk)  \n[christoph. clausen@dotphoton. com](christoph. clausen@dotphoton. com)  \n[bruno. sanguinetti@dotphoton. com](bruno. sanguinetti@dotphoton. com)  \nDotphoton AG  \nReviewed on OpenReview: [https: // openreview. net/ forum? id= I4IkGmgFJz](https: // openreview. net/ forum? id= I4IkGmgFJz)  \n∗ Equal contribution  \nAbstract  \nCamera images are ubiquitous in machine learning research. They also play a central role in the delivery of important public services spanning medicine or environmental surveying.  \nHowever, the application of machine learning models in these domains has been limited because of robustness concerns. A primary failure mode are performance drops due to differences between the training and deployment data. While there are methods to prospectively validate the robustness of machine learning models to such dataset drifts, existing approaches do not account for explicit models of machine learning’s primary object of interest:  \nthe data. This limits our ability to study and understand the relationship between data generation and downstream machine learning model performance in a physically accurate manner. In this study, we demonstrate how to overcome this limitation by pairing traditional machine learning with physical optics to obtain explicit and differentiable data models. We demonstrate how such data models can be constructed for image data and used to control downstream machine learning model performance related to dataset drift. The findings are distilled into three applications. First, drift synthesis enables the controlled generation of physically faithful drift test cases to power model selection and targeted generalization.  \nSecond, the gradient connection between machine learning task model and data model allows advanced, precise tolerancing of task model sensitivity to changes in the data generation.  \nThese drift forensics can be used to precisely specify the acceptable data environments in which a task model may be run. Third, drift optimization opens up the possibility to create drifts that can help the task model learn better faster, effectively optimizing the data generating process itself to support the downstream machine vision task. This is an interesting upgrade to existing imaging pipelines which traditionally have been optimized tobe consumed by human users but not machine learning models. The data models require access to raw sensor images as commonly processed at scale in industry domains such as microscopy, biomedicine, autonomous vehicles or remote sensing. Alongside the data model code we release two datasets to the public that we collected as part of this work. In total, the two datasets, Raw-Microscopy an","cbCaipr9NCj3mQdg","https://ap.wps.com/l/cbCaipr9NCj3mQdg","pdf",51036075,1,54,"English","en",105,"# Abstract\n# Introduction\n## Motivation and problem statement\n## Camera image datasets in ML\n## Robustness concerns and deployment gaps","[{\"question\":\"Why is dataset drift a major problem for machine learning on image data?\",\"answer\":\"Models can experience performance drops when the deployment data differs from the training data. This robustness failure mode directly limits reliable use in sensitive application domains.\"},{\"question\":\"How does the study use physical optics to address drift validation?\",\"answer\":\"It pairs traditional machine learning with physical optics to construct explicit, differentiable data models. These models make physically faithful drift test cases and enable deeper control over downstream model behavior.\"},{\"question\":\"What are the three key applications derived from the data models?\",\"answer\":\"The study highlights drift synthesis for controlled test cases and targeted generalization, gradient connections for precise tolerancing of task sensitivity, and drift optimization that can improve how data is generated to better support machine vision tasks.\"}]","Data Models for Dataset Drift Controls in Machine Learning With Optical Images | PDF",1785672428,136,{"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},"data-models-for-dataset-drift-controls-in-machine-learning-with-optical-images","",{"@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/data-models-for-dataset-drift-controls-in-machine-learning-with-optical-images/116908/",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","2026-08-02",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 is dataset drift a major problem for machine learning on image data?","Question",{"text":75,"@type":76},"Models can experience performance drops when the deployment data differs from the training data. This robustness failure mode directly limits reliable use in sensitive application domains.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the study use physical optics to address drift validation?",{"text":80,"@type":76},"It pairs traditional machine learning with physical optics to construct explicit, differentiable data models. These models make physically faithful drift test cases and enable deeper control over downstream model behavior.",{"name":82,"@type":73,"acceptedAnswer":83},"What are the three key applications derived from the data models?",{"text":84,"@type":76},"The study highlights drift synthesis for controlled test cases and targeted generalization, gradient connections for precise tolerancing of task sensitivity, and drift optimization that can improve how data is generated to better support machine vision tasks.","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},[93,97,101,105,110,115,120,123,128,131,135],{"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":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"]