Enhanced plant health monitoring with dual head CNN for leaf classification and disease identification

Abstract

This study introduces a novel Dual-Head Convolutional Neural Network (DH-CNN) designed to enhance plant health monitoring by concurrently classifying plant leaves and detecting diseases. The monitoring of plant health is essential for ensuring food security and improving agricultural productivity, underscoring the need for accurate and efficient disease detection methods. Utilizing the extensive PlantVillage dataset, the DH-CNN model proposed in this research demonstrates significant improvements compared to existing approaches. The model achieved a classification accuracy of 99.71 % for leaf classification and 99.26 % for disease identification across 13,576 previously unseen images. Featuring a shared feature extraction layer and two classifier heads, the DH-CNN architecture optimizes performance for both classification tasks. This novel approach aims to support agricultural automation by providing precise and effective plant species and disease detection, thereby contributing to better crop management and disease control strategies. • Explores the development of a dual-head convolutional neural network (DH-CNN) for simultaneous plant species classification and leaf disease identification. • Achieves remarkable accuracy with 99.71 % in leaf classification and 99.26 % in disease identification on 13,576 previously unseen images by the model. • Demonstrated the superiority of DH-CNN over ResNet-101 by modifying and enhancing the pre-trained model. • Aims to facilitate agricultural automation by automatically detecting plant species and identifying leaf diseases using an autonomous robot.