Field-acquired RGB image dataset of potato (Solanum tuberosum) leaves with late blight (Phytophthora infestans) and healthy leaves from Cajamarca, Peru
Description
This dataset contains 484 RGB images of potato (Solanum tuberosum) leaves acquired under real field conditions in Caserío Sapalí, Santa Cruz province, Cajamarca, Peru (latitude -7.3565, longitude -78.4739; 3150 m a.s.l.), a high-altitude Andean environment with high humidity favorable to late blight. Images are organized into two classes: Healthy (216 images) and Late blight caused by Phytophthora infestans (268 images). Images were captured with a low-cost smartphone (HONOR CMA-LX3, 4000 × 3000 px, JPG) at a standardized height of 1.5 m above ground level, under diffuse natural light and avoiding strong shadows. Field conditions include natural backgrounds and lighting variability. Collection period: [dates]. Potato variety: [variety]. Class labels were assigned by [visual inspection by ... / laboratory confirmation]. Duplicate and near-duplicate images were removed using perceptual hashing (pHash) and SHA-256 verification. The dataset includes: (1) original images by class; (2) images preprocessed to 600 × 600 px (proportional resizing and central crop); (3) the training/validation/test split used in the associated study; (4) a README describing the structure and protocol. It can be reused for plant disease detection, image classification, transfer learning and explainability (e.g., Grad-CAM) research, and for evaluating models under real-field, low-cost acquisition conditions. Associated article: Sanchez-Sanchez, J., Minguillo-Rubio, C., Arcila-Diaz, J. (2026). Development of a Machine Learning technique for the detection of late blight in potato leaves. Journal of Agriculture and Food Research, 29, 103109. https://doi.org/10.1016/j.jafr.2026.103109 If you use this dataset, please cite both the dataset and the associated article.