Eggplant Leaf Disease Detection Using Deep Learning Approach

Abstract

Eggplant (Solanum melongena), an important vegetable crop, is heavily infested with high incidence of diseases like Leaf Spot, Mosaic Virus, White Mold, Wilt, Small Leaf disorder, and Insect Pest damage on the foliage. These diseases greatly restrict yield and crop quality, particularly in areas where farm support is not available at the right time. Conventional disease diagnosis relies on expert visual examination, which is normally time-consuming, subjective, and not accessible to farmers in remote or under-equipped locations. This study suggests a deep learning-based mobile system for automatic, real-time identification of seven most significant eggplant leaf statuses: Healthy, LeafSpot, MosaicVirus, InsectPest, SmallLeaf, WhiteMold, and Wilt. A high-quality image dataset of 2,000 images was captured from actual agricultural fields and labeled with the consultation of expert agronomists and after augmentation we work on 9,800 augmented images of dataset. To help generalize the model more effectively, several data augmentation strategies were used. Among the models tried, the best was ResNet50 with 99% accuracy and F1-score of 0.98. The best model was chosen and converted to TensorFlow Lite, compressing the model size to 15 MB with quick, on-device inference appropriate for mobile apps. A simple Android mobile app was built, which allowed the farmer to take leaf images and get real-time disease classification and agronomic advice without internet connectivity.