[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85084-en":3,"doc-seo-85084-105":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},85084,1099514067438,"River Wang","https://ap-avatar.wpscdn.com/avatar/100002539ee87300030?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780474512215547542",8,"Research & Report","Predicting Viticulture Potential through an Ensemble of U-Net and a Geospatial Foundation Model","Agricultural potential assessment underpins sustainable land management and agricultural planning, especially when field surveys are costly and need frequent updates under changing conditions like climate variability. ImageCLEF AI4Agri 2026 Subtask 1 focuses on predicting viticulture suitability in Southern France using Sentinel-2 multispectral imagery. A Georgia Tech DS@GT ARC ensemble combines U-Net semantic segmentation with Prithvi-2.0, achieving ±1 accuracy of 68.32 and ranking 2nd among 7 teams.","Predicting Viticulture Potential through an Ensemble of U-Net and a Geospatial Foundation Model  \nImageCLEF AI4Agri Subtask 1 at CLEF 2026  \nJorge Ignacio Perez 1, * , Hwaai Kang Kee1, * and Lucas Rassbach1, * 1 Georgia Institute of Technology, North Ave NW, Atlanta, GA 30332  \nAbstract  \nDetermining agricultural potential is fundamental to sustainable land management and agricultural planning. Remote sensing data is increasingly valuable as an avenue for agricultural potential due to the cost of traditional methods (surveys, in-situ measurements, soil testing, etc) . ImageCLEF AI4Agri 2026: Subtask 1 is concerned with the prediction of viticulture potential in Southern France. The DS@GT ARC’s submission for Subtask 1 introduces an ensemble of U-Net and a Geospatial Foundation Model (Prithvi-2.0) . Our best model achieved a ±1 accuracy of 68.32 on the leaderboard, ranking 2nd among 7 teams. The implementation for this work is publicly available [at github.com/dsgt-arc/imageclef-ai4agri-2026](at github.com/dsgt-arc/imageclef-ai4agri-2026) .  \nKeywords  \nImageCLEF 2026, AI4Agri 2026, remote sensing, Agricultural Potential, Earth Observation, Precision Agriculture, Multi-temporal Imagery, Hyperspectral Data, Semantic Segmentation, Foundation Models, Vision Transformers, ViT  \n1. Introduction  \nThe agricultural suitability of land is an important topic for agricultural entities such as farmers and policymakers, as well as for economic development efforts. If the agricultural potential of land is understood, a determination can be made as to what kind of agriculture should be facilitated, leading to better ecological conditions, higher crop yield, and improved economic outcomes. Agricultural suitability is typically done via physical surveying, a costly, time-consuming, and manual process [1] . A compounding disadvantage of the time-consuming and manual process of physical surveys is that the surveys must be reassessed frequently to capture the changing conditions of a region, especially when one considers climate change. The gains to be realized from the creation of a digital and automated process are profound. Subtask 1 of the ImageCLEF AI4Agri 2026 challenge [2] utilizes the AgriPotential dataset [1] to classify Sentinel-2 satellite multispectral imagery, where each pixel is labeled with an agricultural suitability score from 1 to 5 for viticulture. Exploratory data analysis (EDA) indicates that temporal modeling appears to be important due to varying Normalized Difference Vegetation Index (NDVI) values across time frames. We hypothesize that temporal modeling is important and will lead to greater performance gains in this task.  \nThe Data Science at Georgia Tech Applied Research Competitions (DS@GT ARC) team has developed a solution via an ensemble machine learning model consisting of a U-Net [3] semantic segmentation model and the Prithvi [4] geospatial foundation model. Prithvi acts as the student in a teacher-student model, with the U-Net as the teacher for unlabeled pixels. This approach led to acceptable results and demonstrates additional avenues that could be pursued to improve performance further.  \nCLEF 2026 Working Notes, 21 – 24 September 2026, Jena, Germany  \n* Corresponding author.  \n$ [jperez333@gatech.edu](jperez333@gatech.edu) (J. I. Perez); [hkee7@gatech.edu](hkee7@gatech.edu) (H. K. Kee); [lrassbach3@gatech.edu](lrassbach3@gatech.edu) (L. Rassbach)  \n􀀚 0009-0001-3367-7646 (J. I. Perez); 0009-0003-7613-5905 (H. K. Kee); 0009-0008-8096-7700 (L. Rassbach)  \n © 2026 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0) .  \n2. Related Work  \n2.1. Prithvi-EO-2.0  \nThe paper by Szwarcman et al details an Earthly Observation (EO) pre-trained model, Prithvi-EO-2.0 [4] . EO methods have been revolutionized by the introduction of AI systems trained on large unlabeled satellite imagery datasets. The authors created a new foundational model for EO ma","cbCaibJDnkMtdJJd","https://ap.wps.com/l/cbCaibJDnkMtdJJd","pdf",712149,1,10,"English","en",105,"# Abstract\n# Introduction\n# Related Work\n## Prithvi-EO-2.0\n## U-Net\n# Methodology","[{\"question\":\"What task does ImageCLEF AI4Agri 2026 Subtask 1 address?\",\"answer\":\"It predicts viticulture potential in Southern France using Sentinel-2 multispectral imagery, assigning each pixel a suitability score from 1 to 5.\"},{\"question\":\"What models are combined in the proposed ensemble?\",\"answer\":\"The solution uses a weighted ensemble of a U-Net semantic segmentation model and the geospatial foundation model Prithvi-2.0.\"},{\"question\":\"How well did the best model perform on the leaderboard?\",\"answer\":\"The best model reached ±1 accuracy of 68.32 and ranked 2nd among 7 teams.\"}]",1784200966,25,{"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},"predicting-viticulture-potential-through-an-ensemble-of-u-net-and-a-geospatial-foundation-model","",{"@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/predicting-viticulture-potential-through-an-ensemble-of-u-net-and-a-geospatial-foundation-model/85084/",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 task does ImageCLEF AI4Agri 2026 Subtask 1 address?","Question",{"text":75,"@type":76},"It predicts viticulture potential in Southern France using Sentinel-2 multispectral imagery, assigning each pixel a suitability score from 1 to 5.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What models are combined in the proposed ensemble?",{"text":80,"@type":76},"The solution uses a weighted ensemble of a U-Net semantic segmentation model and the geospatial foundation model Prithvi-2.0.",{"name":82,"@type":73,"acceptedAnswer":83},"How well did the best model perform on the leaderboard?",{"text":84,"@type":76},"The best model reached ±1 accuracy of 68.32 and ranked 2nd among 7 teams.","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},[93,97,101,105,110,115,120,123,128,131,134],{"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":21,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":21,"slug":133},"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]