Apple Leaf Disease Detection using Deep Learning Based Pre-Trained Models

All four of these leaf diseases—Alternaria leaf spot, brown spot, mosaic, and grey spot—have had a major impact on apple yield. But there isn't yet a fast and accurate apple disease detector in the literature, which would slow down the industry's expansion. Using improved convolutional neural networks (CNNs), this study proposes a deep learning approach for immediate-form illness detection in apple leaves. The apple leaf disease dataset, which contains both standardized lab shots and complex field images, is initially constructed in this work using data augmentation and picture annotation techniques. Thus, we present a novel deep-CNN model based on the MobileNet architecture and ResMLP for disease detection in apple leaves. In the end, a dataset consisting of 962 images of infected apple leaves is used to train the suggested ResMLP model and MobileNet structure to detect these five common apple leaf illnesses. Experimental results show that the suggested framework has a detection performance of 98.25% mAP. By comparing the novel model to state-of-the-art methods, we find that it provides a high-performance solution for early apple leaf disease diagnosis, both in terms of accuracy and real-time detection speed.