DEEP LEARNING-BASED FRAMEWORK FOR EARLY DIAGNOSIS, SEVERITY ASSESSMENT, AND PEST DETECTION IN COCONUT TREES

Coconut production is severely affected by diseases such as Weligama coconut leaf wilt, root wilt, and bud rot,
and by pest infestations including caterpillars, rhinoceros beetles, and red palm weevils. Traditional manual detection is
labor-intensive, subjective, and ineffective for large-scale monitoring. This paper presents an integrated deep learning
based framework for coconut tree health monitoring. The system combines GoogleNet for disease classification, CNNs
for leaf-level severity detection, and ResNet-50 for pest recognition. Results on a curated dataset show 95.6% accuracy
for disease classification, 93.8% for leaf wilt severity grading, and 92.9% for pest detection. The proposed solution
highlights the potential of AI-driven tools to support precision agriculture and sustainable coconut farming.
Keywords: Coconut Tree Diseases, Deep Learning, GoogleNet, ResNet, Severity Assessment, Pest Detection