Abstract
Several studies have reported effective image-based methods for detecting fungal diseases in corn. However, development of an AI-based offline system remains to be accomplished, which supports the SDGs no. 9 (Industry, Innovation, and Infrastructure) and 15 (Life on Land). The HuskyLens AI image sensor was incrementally trained on 1,620 maize leaf images which was classified into healthy and three (3) fungal disease categories, with severity level annotation. The HuskyLens AI achieved an overall accuracy of 0.89, precision of 0.90, recall of 0.99, and F1-score of 0.94, which indicates good classification performance. Furthermore, the integrated SIM800L GSM module that allowed offline SMS Alerts showed device responsiveness, with LCD response time achieving 8.13 seconds and SMS delivery time achieving 24.84 seconds, establishing system efficiency. The high-level post-implementation results of the acceptability, adaptability and SMS alert intervention performance of the device indicates that the offline AI-based system provides an effective, accessible, and sustainable solution for early maize disease detection in low-connectivity farming areas, supporting refined crop management and resource efficiency. Future studies for prototype development should expand the dataset to increase accuracy, include additional maize diseases, and refine training, SMS alerts, and power efficiency. The model should be field tested across locations and through multiple crop cycles for further validation. Additionally, the model should be distributed across the region to support smallholder farmers, extension workers, and local governments in low-connectivity areas.