Apple Leaf Disease Detection using Machine Learning and Deep Learning Models

Abstract

Apple leaf disease detection is of critical importance for enhancing productivity in agriculture while reducing losses caused by diseases. Conventional apple leaf disease detection techniques are often inefficient and inaccurate because of environmental factors and the similarity of symptoms in the early stages of diseases. In this paper, a novel intelligent multi-model-based apple leaf disease detection system is proposed for enhanced accuracy and robustness of the proposed method. In the proposed method, a wide range of preprocessing techniques are employed for background removal, segmentation of the apple leaves, and extraction of disease region information. Various models such as MobileNetV2, ResNet50, DenseNet121, EfficientNetB0, Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbors (KNN) are implemented and compared for enhanced accuracy of the proposed method. An adaptive ensemble method is also proposed for combining the results of different models for enhanced generalization performance of the proposed method. From the experimental results, it is observed that the proposed method based on SVM with deep feature extraction using MobileNetV2 achieves a high accuracy of 99.1%. Furthermore, a novel disease severity estimation module is also proposed for quantification of the percentage of disease infection in the apple leaves using HSV color space segmentation techniques.