[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82856-en":3,"doc-seo-82856-105":29,"detail-sidebar-cat-0-en-105":90},{"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":13,"seo_description":14,"update_tm":27,"read_time":28},82856,2336464648322,"Aria","https://ap-avatar.wpscdn.com/avatar/2200025388227c56fec?_k=1778556882303663488",8,"Research & Report","FM-ChangeNet: Learning Change through Pathwise Feature Transport","FM-ChangeNet proposes a pathwise-supervised framework for change detection that converts bi-temporal reasoning into continuous transport in feature space. Encoded pre- and post-temporal representations are connected through intermediate latent states, where a time-conditioned velocity field models transformation dynamics along the trajectory. This continuum supervision yields denser, less ambiguous learning than endpoint-only methods and explicitly captures temporal evolution. The velocity magnitude becomes an interpretable spatial change cue, separating true structural variation from nuisance effects such as illumination shifts and misalignment. A hierarchical multi-scale network with cross-temporal alignment and a unified objective couples flow, trajectory, regularization, and segmentation losses, delivering robust results on remote sensing benchmarks.","arXiv :2607 .04750v 1 [ cs .AI] 6 Jul 2026  \nFM-ChangeNet: Learning Change through Pathwise  \nFeature Transport  \nRoie Kazoom  \nGoogle Research [kazoomroie@google.com](kazoomroie@google.com)  \nGeorge Leifman  \nGoogle Research [gleifman@google.com](gleifman@google.com)  \nGenady Beryozkin  \nGoogle Research [genady@google.com](genady@google.com)  \n􀜶1  \nෝ􀢜􀣂 (􀢠 􀢚, 􀢚)  \nෝ􀢜􀣂 (􀢠 􀢚, 􀢚)  \n􀜶0  \nFigure 1: Conceptual illustration of flow matching for change detection. Given a bi-temporal image pair (T0 , T1 ), the model defines an intermediate latent state zt = (1 − t)f 0 + tf1 along a continuous feature-space interpolation between the encoded pre and post-temporal representations. A time-conditioned velocity field vˆθ (zt , t) characterizes the local transport dynamics from T0 toward T1 , while its magnitude ∥vˆθ (zt , t)∥2 provides a spatially localized cue for change.  \nAbstract  \nWe present FM-ChangeNet, a pathwise-supervised framework for change detection that reformulates bi-temporal reasoning as continuous transport in feature space rather than static endpoint comparison. Given encoded pre and post-temporal representations, we construct intermediate latent states and learn a time-conditioned velocity field vˆθ (zt , t) along the transformation trajectory. This pathwise formulation constrains the predictor over a continuum of intermediate states, providing a denser and less ambiguous supervision signal than conventional endpoint-only segmentation and enabling the model to capture temporal evolution explicitly. The learned velocity field is not only a transport mechanism but also an interpretable representation of change: its magnitude serves as a spatially localized change cue that helps distinguish true structural variation from nuisance effects such as illumination shifts and spatial misalignment. We develop a hierarchical multi-scale architecture with cross-temporal alignment, time-conditioned coarse-to-fine flow decoding, and a unified objective that couples flow supervision, trajectory consistency, spatial regularization, and segmentation loss. Experiments on remote sensing benchmarks show that the proposed framework produces more structured and robust change representations while achieving state-of-the-art performance.  \nPreprint.  \nˆ  \nPre-temporal Observation (T0) Post-temporal Observation (T1) Velocity Magnitude (kk 2) Ground Truth (Y) Inferred Change Map (Y)  \n0.8  \n0.6  \n0.4  \n0.2  \n0.8  \n0.6  \n0.4  \n0.2  \n0.8  \n0.6  \n0.4  \n0.2  \nFigure 2: Change detection as feature-space transport. Given a bi-temporal pair (T0 , T1 ), the encoder extracts feature representations f0 = E (T0 ) and f1 = E (T1 ) . We model their relationship through a continuous latent path zt = (1 − t)f 0 + tf1 and learn a time-conditioned velocity field vˆθ (zt , t) using flow matching. The velocity magnitude ∥vˆθ (zt , t)∥2 provides a localized cue for change: unchanged regions require little transport, whereas structurally changed regions induce coherent high-magnitude responses. This transport-based representation aligns with the ground-truth mask Y and supports accurate prediction of the final change map ˆY .  \n1 Introduction  \nChange detection (CD) aims to identify meaningful surface variations between bi-temporal images captured over the same geographical area at different times. It is a fundamental problem in remote sensing, with applications in urban expansion monitoring [3, 28], disaster assessment [36], and environmental conservation. Reliably isolating changes remains challenging due to nuisance factors such as illumination variation, seasonal effects, and viewpoint shifts, particularly in high-resolution imagery [8, 33] . Moreover, detecting subtle or small-scale changes often leads to high false-alarm rates when relying on local spectral differences [15] . Modern CD methods are dominated by Siamese architectures, where bi-temporal inputs are processed independently and fused via concatenation, differencing, or attention mechanisms. Early works","cbCaifhtKSACsGal","https://ap.wps.com/l/cbCaifhtKSACsGal","pdf",13560561,1,30,"English","en",105,"# Abstract\n# Introduction","[{\"question\":\"What is FM-ChangeNet’s core idea for change detection?\",\"answer\":\"FM-ChangeNet reformulates bi-temporal change detection as continuous transport in feature space using a pathwise-supervised transformation instead of relying only on endpoint comparisons.\"},{\"question\":\"How does FM-ChangeNet model temporal evolution between two images?\",\"answer\":\"It constructs intermediate latent states along a continuous path and learns a time-conditioned velocity field that characterizes transport dynamics from the pre-temporal to post-temporal representation.\"},{\"question\":\"What does the velocity field magnitude represent in the method?\",\"answer\":\"The magnitude of the learned velocity field provides a spatially localized change cue: regions that remain unchanged require little transport, while structurally changed regions induce coherent high-magnitude responses.\"}]",1784183472,76,{"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":85,"head_meta":87,"extra_data":89,"updated_unix":27},"fm-changenet-learning-change-through-pathwise-feature-transport","",{"@graph":35,"@context":84},[36,53,67],{"@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/fm-changenet-learning-change-through-pathwise-feature-transport/82856/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What is FM-ChangeNet’s core idea for change detection?","Question",{"text":74,"@type":75},"FM-ChangeNet reformulates bi-temporal change detection as continuous transport in feature space using a pathwise-supervised transformation instead of relying only on endpoint comparisons.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does FM-ChangeNet model temporal evolution between two images?",{"text":79,"@type":75},"It constructs intermediate latent states along a continuous path and learns a time-conditioned velocity field that characterizes transport dynamics from the pre-temporal to post-temporal representation.",{"name":81,"@type":72,"acceptedAnswer":82},"What does the velocity field magnitude represent in the method?",{"text":83,"@type":75},"The magnitude of the learned velocity field provides a spatially localized change cue: regions that remain unchanged require little transport, while structurally changed regions induce coherent high-magnitude responses.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":21,"slug":120},"research-report",{"id":122,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]