A Spatio-temporal Image Dataset of Maize Leaf Herbivory by Four Major Pests

This dataset contains 14,642 unaugmented images of maize leaf herbivory caused by four lepidopteran pests: Spodoptera frugiperda, Spodoptera litura, Mythimna separata, and Agrotis ypsilon. The image data records the physical progression of leaf damage from the 1st to the 6th larval instar. Annotations are provided in the accompanying CSV file across three levels: pest species, estimated instar, and chronological age (days post-hatching). Images were acquired using a flatbed scanner under standardized laboratory rearing conditions (25°C). The regions of interest (ROIs) were manually cropped to isolate the bite boundaries and do not conform to a fixed aspect ratio or resolution. The dataset supports research and algorithm evaluation in the following areas: Fine-Grained Visual Classification (FGVC) Multi-Task Learning (simultaneous species classification and age regression) Out-of-Distribution (OOD) generalization across developmental stages See the attached README.md for directory structure details and preprocessing instructions.