A COMPREHENSIVE REVIEW ON DEEP LEARNING AND ARTIFICIAL INTELLIGENCE APPROACHES FOR COFFEE LEAF DISEASE AND PEST DETECTION

Abstract

Coffee production faces constant threats from leaf diseases and pest attack that reduce yield and quality. Early and accurate detection of these conditions is essential for sustainable coffee cultivation and global food security. In recent years, deep learning and computer vision have emerged as transformative tools in agricultural disease monitoring. This review critically examines recent developments in artificial intelligence (AI) based detection systems for coffee leaf and pest diseases, focusing on convolutional neural networks (CNNs), transfer learning models such as MobileNetV2 and ResNet, object detection architectures like YOLO and Mask R-CNN, and hybrid models integrating segmentation and graph neural networks. The synthesis of fifteen peer-reviewed studies reveals a consistent trend toward improved accuracy, automation, and scalability in disease identification, with models achieving up to 99% precision using curated datasets such as RoCoLe and BRACOL. Despite these advances, challenges remain in dataset diversity, real-time implementation, and adaptation to varying environmental conditions. This review highlights current gaps, evaluates comparative model performance, and outlines future research opportunities in integrating edge computing, Internet of Things (IoT), and low-cost sensing technologies for intelligent and sustainable coffee farming.