AUTOMATED MULTI-CLASS PEST CLASSIFICATION IN AGRICULTURE USING EFFICIENTNET-B0 WITH TRANSFER LEARNING

Abstract

Accurate pest identification is critical for effective crop protection. Manual identification methods are time-consuming and require specialized expertise. We developed an automated classification system using EfficientNet-B0 transfer learning to recognize twelve agricultural pest species: ants, bees, beetles, caterpillars, earthworms, earwigs, grasshoppers, moths, slugs, snails, wasps, and weevils. Trained on 7,826 specimens, the model achieved 97.44% validation and 97.24% test accuracy. The system was deployed as a web application providing rapid pest identification for agricultural practitioners. This work bridges deep learning research with practical agricultural solutions.