Dataset: Forage grasses in crop fields from ultra-high spatial resolution UAV-based imagery

Simple item page
cg.authorship.typesCGIAR single centreen
cg.contributor.initiativeAccelerated Breeding
cg.contributor.initiativeSustainable Animal Productivity
cg.coverage.regionSouth America
cg.creator.identifierJuan Andrés Cardoso Arango: 0000-0002-0252-4655en
cg.creator.identifierRosa Noemi Jauregui: 0000-0002-1058-3735en
cg.creator.identifierRodrigo Camelo: 0000-0003-0563-6123en
cg.creator.identifierAndres Felipe Ruiz-Hurtado: 0000-0003-1293-8736en
cg.creator.identifierDarwin Alexis Arrechea-Castillo: 0000-0002-2395-2181en
cg.identifier.doihttps://doi.org/10.7910/dvn/dbgufwen
cg.reviewStatusInternal Reviewen
cg.subject.alliancebiovciatCROP PRODUCTIONen
cg.subject.alliancebiovciatHEALTHen
cg.subject.alliancebiovciatNUTRITIONen
cg.subject.impactAreaClimate adaptation and mitigation
cg.subject.impactAreaEnvironmental health and biodiversity
cg.subject.impactAreaNutrition, health and food security
dc.contributor.authorCardoso Arango, Juan Andresen
dc.contributor.authorJauregui, Rosa Noemien
dc.contributor.authorCamelo-Munevar, Rodrigo Andresen
dc.contributor.authorRuiz-Hurtado, Andres Felipeen
dc.contributor.authorArrechea-Castillo, Darwin Alexisen
dc.date.accessioned2025-05-02T00:58:46Zen
dc.date.available2025-05-02T00:58:46Zen
dc.identifier.urihttps://hdl.handle.net/10568/174405
dc.titleDataset: Forage grasses in crop fields from ultra-high spatial resolution UAV-based imageryen
dcterms.abstractThis dataset contains orthomosaics and individual Regions of Interest (ROIs) of forage grasses in crop fields from experimental trials of CIAT’s tropical forages breeding program; and annotations in Common Objects in Context (COCO) format derived from that data. The ROIs were manually annotated on UAV imagery and exported in common objects in context (COCO) format compatible with different machine learning models and architectures. 9,554 ROIs in the geospatial data and 12,365 annotations of forage grasses in COCO format. Methodology: The dataset was generated through a multi-step process beginning with data acquisition of forages crop fields via UAV flights (DJI Phantom 4 Multispectral drone) with RTK determining the geolocation. These images were processed in Agisoft Metashape to generate georeferenced orthomosaics as raster files. Manual annotation of forage grasses ROIs was performed in QGIS and the geospatial data for 8 different orthomosaics was later converted to COCO format using custom python scripting. To ensure compatibility witch COCO standards and optimize training efficiency, the large orthomosaics where clipped to the annotations’ extents with additional 1% spatial buffer and split into tiles with a maximum dimension close to 1024 pixels for the larger side and 25% overlap.en
dcterms.accessRightsOpen Access
dcterms.bibliographicCitationCardoso Arango, J.A.; Jauregui, R.N.; Camelo-Munevar, R.A.; Ruiz-Hurtado, A.F.; Arrechea-Castillo, D.A. (2025) Dataset: Forage grasses in crop fields from ultra-high spatial resolution UAV-based imagery. https://doi.org/10.7910/DVN/DBGUFWen
dcterms.issued2025en
dcterms.languageen
dcterms.licenseCC-BY-NC-4.0
dcterms.subjectmachine learningen
dcterms.subjectunmanned aerial vehiclesen
dcterms.subjectimageryen
dcterms.subjectfeed cropsen
dcterms.typeDataset

Files

Collections

Alliance Bioversity CIAT Datasets
Alliance Research Lever 6: Crops for Nutrition and Health
CGIAR Initiative on Accelerated Breeding
CGIAR Initiative on Sustainable Animal Productivity