Lightweight Convolutional Neural Network for Multi-Class Plant Disease Detection and Classification

Abstract

Globally, there is rise of farming and cultivation. Now a days due to global warming and unpredicted conditions of nature due to deforestation has become a predominant challenge for the farming and agri-communities to enrich the system. To address the need, it is necessary to save the crop from diseases at early stage for better yield. This paper presents a lightweight convolutional neural network (CNN) model for accurate and efficient plant disease classification using leaf images. Leveraging the publicly available PlantVillage dataset, the proposed model undergoes extensive preprocessing and data augmentation to enhance its robustness and generalization. The model architecture includes convolutional layers for spatial feature extraction, ReLU activations, dropout regularization, and softmax-based classification. Evaluated through metrics such as accuracy, precision, recall, and F1-score, the model achieves high classification performance with strong real-world deployment potential. Visual analysis via confusion matrices and accuracy/loss trends further affirms the model’s reliability. With practical deployment on mobile and edge devices in mind, this work contributes to the development of scalable, AI-driven solutions for early plant disease detection in smart agriculture