<![CDATA[ABSTRACT Coffee is one of Indonesia’s leading commodities, playing a vital role in the national economy, providing employment opportunities, and serving as a primary source of income for many farmers. However, coffee productivity is often reduced due to pest and disease attacks, particularly on th leaves, such as coffee leaf rust and red spider mites. These diseases can disrupt photosynthesis, lowe plant quality, and even cause plant death if not addressed promptly. Manual identification at the farmer level is often challanging due to limited knowledge and the similarity of visual symptoms betwen diseases. This study aims to develop an image classification system for detecting healthy leaves, leaf rust, and red spider mite infestations on coffee plants automatically. The method employed is machine learning based on a Convolutional Neural Network architecture using MobileNetV2 and transfer learning. The dataset consists of 501 images of coffee leaves, divided into 456 training data and 45 testing data. The model was trained to distinguish between the three classes, achieving a training accuracy of 64% and a testing accuracy of 56%. The resulting model was then integrated into a web-based application using Streamlit, enabling easy access for farmers and the general public. This system is expected to facilitate early detection of coffee leaf diseases in a faster, more practical, and affordable way, allowing farmers to take timely action before damage spreads. In the long term, this technology is anticipated to support improved coffee plantation productivity in Indonesia. Keywords: image classification, coffee leaf, MobileNetV2, transfer learning.]]>
Copyrights © 2025
