A Deep Learning Approach for Detecting Pests and Diseases in Maize Crops

Abstract

Maize is a major staple crop whose pests and diseases threaten food security, and conventional detection is slow, labor-intensive, and error-prone. This work presents an automated deep-learning system that classifies maize leaf diseases and insect damage from images, training DenseNet-121 and ResNet-50 on a publicly available, expert-labeled dataset collected under realistic field conditions and spanning 23 disease and insect classes. DenseNet-121 achieved the strongest per-class accuracy (especially on insect-related classes), while ResNet-50 offered consistent overall reliability. The results point to deploying such models on edge or mobile devices to give farmers continuous, accessible crop-health monitoring and decision support for precision agriculture.