Maize is a vital staple food in Uganda, eaten in various forms. However, it is often attacked by different pests and diseases that require immediate attention when detected to minimize agricultural losses. Recently, deep learning has proven to be very efficient, especially in computer vision. However, the basic model frameworks cannot efficiently capture complex image patterns. This study leverages YOLOv8 (You Only Look Once, Version 8), an advanced deep-learning model, to detect and classify maize leaf diseases using spectral maize leaf data from Uganda. The dataset comprises 37,217 maize leaf images with healthy samples and four disease classes, including Fall Army Worm (FAW), Maize Leaf Blight (MLB), Maize Lethal Necrosis (MLN), and Maize Streak Virus (MSV). Data augmentation techniques were adopted to enhance the model’s robustness, including random cropping, flipping, adjustments, RGB shifts, and color jittering. Thanks to these augmentation techniques, the model’s accuracy improved from 94% to 99.7%; this underscores the effectiveness of data augmentation in boosting the model’s generalization ability and robustness. The model’s high accuracy highlights its potential to significantly assist in early disease detection, ultimately increasing agricultural output.
