Paddy Crop RGB Drone Data

This dataset contains 224\u00d7224 pixel RGB sub-images of paddy crop fields, captured using the RGB sensor of the DJI Mavic 3M drone. Each sub-image exclusively represents a paddy crop region and is labelled as either healthy or unhealthy based on visual indicators such as leaf color, density, canopy structure, and texture. The images are organized into class-specific folders to support supervised learning. A Python script is also provided to train and evaluate a Vision Transformer-based enhancement-guided dual-branch classification model (MoCE-ViT) that incorporates contrastive learning, image enhancement, and momentum encoding. This dataset supports research in UAV-based agricultural monitoring, crop health classification, and domain-generalizable learning.