Machine Learning-Driven Object Detection Techniques for Automated Identification and Spatial Localization of Plant Diseases

Disease detection in plants is a key aspect of precision agriculture that facilitates early detection and effective control of plant diseases. This research proposes a better approach to plant disease classification through a hybrid framework using deep learning and machine learning algorithms. Feature extraction was carried out using EfficientNetB0, Vision Transformers (ViTs), and handcrafted features to enhance the model's ability to effectively differentiate between healthy and diseased leaves. Advanced data preprocessing techniques, such as KNN imputation for dealing with missing values and data augmentation, improved model generalization and resilience. Among the classification models experimented with, the hybrid CNN + XGBoost model achieved the highest accuracy of 96.3%, outperforming EfficientNetB0 (95.7%) and Vision Transformers (94.8%). The results indicate the potency of combining deep learning feature extraction with machine learning classifiers to achieve outstanding classification performance. This is a scalable, efficient, and high-precision solution for automated plant disease detection, which serves the growth of advanced agricultural practice.