FieldMaizeLeaf-8: A real-field early vegetative maize leaf stress image dataset from Bangladesh

Maize (Zea mays L.) is a rapidly expanding cereal crop in Bangladesh, but its yield is threatened by biotic diseases, insect pests, and nutrient-deficiency disorders that are difficult to distinguish visually, particularly at the early vegetative stage. Most publicly available leaf-image datasets used to train computer-vision models for crop diagnosis are captured under controlled laboratory conditions with plain backgrounds, which limits the field-readiness of models trained on them. FieldMaizeLeaf-8 addresses this gap by providing 3975 RGB photographs of maize leaves aged approximately 1-2 months, captured directly in commercial maize fields in Daudkandi, Cumilla, Bangladesh, using a consumer smartphone under natural daylight, background, and viewing-angle conditions, with no laboratory staging. Images are organised into eight classes: one healthy-leaf class and seven stress classes, comprising three biotic conditions (Fall Armyworm damage, Leaf Spot, and Northern Corn Leaf Blight) and four abiotic nutrient-deficiency conditions (Nitrogen, Phosphorus, Potassium, and Magnesium deficiency). Class labels assigned in the field were subsequently reviewed and validated by a plant-pathology expert. The dataset retains natural variability in illumination, leaf orientation, and symptom severity, and is intended to serve as a real-field benchmark for developing and evaluating deep-learning models for maize stress recognition, transfer learning, and explainable artificial intelligence (XAI) applications in precision agriculture.