Mask R-CNN Powered Deep Learning Model for Accurate Diagnosis and Classification of Plant Leaf Diseases

Abstract

Timely and accurate identification of plant leaf diseases is crucial in the agricultural sector, where plant health directly impacts crop yield and food security. Traditional methods for diagnosing plant diseases predominantly rely on manual inspection and expert judgment, which are often time-consuming, labor-intensive, and not scalable for large-scale monitoring. As agricultural demands continue to rise, there is a pressing need for automated, intelligent systems that can assist in the early detection and classification of plant diseases with high precision and reliability. In this paper, we propose a novel Mask R-CNN Powered Deep Learning Model for Accurate Diagnosis and Classification of Plant Leaf Diseases (MRPDL-PLD). The proposed model integrates advanced image processing techniques with deep learning architectures to create an end-to-end pipeline capable of identifying infected regions and classifying disease types from raw leaf images. To enhance the quality of the input data, a preprocessing stage using Adaptive Filtering (AF) is employed, effectively reducing image noise and preserving critical features necessary for accurate segmentation and classification.
The core of the proposed system is a Mask Region-Based Convolutional Neural Network (Mask R-CNN), which excels in both instance segmentation and object detection. This architecture allows for precise localization of diseased regions on the leaf surface, enabling a more detailed understanding of the severity and spread of infections. Feature extraction is performed on these segmented regions to generate high-quality feature vectors, which are then used to classify the specific type of disease affecting the plant. To further optimize model performance, we incorporate the ADAM optimizer for adaptive learning rate adjustment, which significantly enhances convergence speed and model accuracy. The MRPDL-PLD model is trained and evaluated on publicly available benchmark datasets consisting of diverse plant species and disease categories. Extensive simulation experiments demonstrate that our model not only achieves superior classification accuracy and segmentation performance compared to existing contemporary approaches but also generalizes well across different types of plant diseases.