Plant Disease Detection: An Effective Approach Incorporating Deep Learning & Machine Learning Techniques

Plant diseases are a great concern nowadays for farmers because of their devastating capabilities in destroying plants and crops. The adverse impacts include food scarcity, a reduction in farmers’ livelihood, and a downward economy. Detecting plant diseases early can be important in mitigating the adverse impacts. However, trivial detection approaches are prone to error and can lead to a mis-treatment. Deep learning (DL) and machine learning (ML) techniques can be incorporated into an effective detection approach. This paper presents an effective detection approach using convolutional neural network (CNN)- based pre-trained DL models and ML models. The models are trained on plant leaf images. By using the approach, different types of plant diseases can be detected. CNN-based deep learning (DL) models such as MobileNetV2, ResNet101, EfficientNetB0, and DenseNet169 are considered feature extractors. Additionally, ML models such as Support Vector Machine(SVM), LightGBM, and XGBoost are considered classifiers. Additionally, a customdesigned DL-based classifier is also considered. The leaf images are collected from PlantVillage and PlantDoc datasets. This dataset consists of high-quality, real-world raw images. The bestachieved results such as 99.05% accuracy, 98.85% recall, 98.35% precision, 98.60% F1-score, and 99.99% AUC are achieved using a custom DL-based Classifier with the MobileNetV2 model. The achieved performance is undoubtedly a significant one among the performances achieved by the related works.