Grapevine Downy Mildew Severity Dataset: 920 Field-Captured Images with Severity Grading Labels

A curated image dataset of grapevine (Vitis vinifera 'Thompson Seedless') leaves exhibiting varying degrees of downy mildew infection caused by Plasmopara viticola, captured under natural field conditions in Karnataka, India. Contents 920 preprocessed images (1024×1024 JPEG) organized by severity class (S0–S4) Severity labels for all 920 images (5 ordinal classes, 0–100% infection) Metadata including split manifests and augmentation provenance CSVs Severity Classes S0: Healthy (0%, n=99) S1: Mild (1–25%, n=409) S2: Moderate (26–50%, n=172) S3: Severe (51–75%, n=132) S4: Very severe (>75%, n=108) Collection Images captured using Nikon D5200 DSLR from three vineyard sites in Karnataka, India during 2021–2023 monsoon seasons. Severity annotated by two PhD plant pathologists (Cohen's κ = 0.83). Related Article K.E. Cholachgudda, R.C. Biradar, B.M. Kiran, M.K. Prasannakumar, “Automated severity grading of grapevine downy mildew in the field: a hybrid segmentation and feature-based machine learning approach,” Journal of Agriculture and Food Research (2026). Related Code github.com/kartikenc/grapevine-downy-mildew-severity